WO2013072543A1 - Method for obtaining useful information for the diagnosis of neuromuscular diseases - Google Patents

Method for obtaining useful information for the diagnosis of neuromuscular diseases Download PDF

Info

Publication number
WO2013072543A1
WO2013072543A1 PCT/ES2012/070796 ES2012070796W WO2013072543A1 WO 2013072543 A1 WO2013072543 A1 WO 2013072543A1 ES 2012070796 W ES2012070796 W ES 2012070796W WO 2013072543 A1 WO2013072543 A1 WO 2013072543A1
Authority
WO
WIPO (PCT)
Prior art keywords
fibers
type
average
typical
neighbors
Prior art date
Application number
PCT/ES2012/070796
Other languages
Spanish (es)
French (fr)
Other versions
WO2013072543A8 (en
Inventor
Luis María ESCUDERO CUADRADO
Adoración MONTERO SÁNCHEZ
Carmen PARADAS LÓPEZ
Eloy RIVAS INFANTE
Alberto Pascual Bravo
Aurora SÁEZ MANZANO
Carmen SERRANO GOTARREDONA
Begoña ACHA PIÑERO
Original Assignee
Servicio Andaluz De Salud
Universidad De Sevilla
Consejo Superior De Investigaciones Científicas (Csic)
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Servicio Andaluz De Salud, Universidad De Sevilla, Consejo Superior De Investigaciones Científicas (Csic) filed Critical Servicio Andaluz De Salud
Priority to PCT/ES2012/070796 priority Critical patent/WO2013072543A1/en
Publication of WO2013072543A1 publication Critical patent/WO2013072543A1/en
Publication of WO2013072543A8 publication Critical patent/WO2013072543A8/en

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T9/00—Image coding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/69—Microscopic objects, e.g. biological cells or cellular parts

Definitions

  • the object of the present invention is a method to obtain, from the biopsy of a patient, a set of new parameters that allow, by means of a comparative histological study, to objectively diagnose different neuromuscular diseases and their degree of involvement to the patient.
  • the method can be performed automatically by computer.
  • Skeletal muscle constitutes the voluntary muscle and is supplied by motor neurons of the somatic nervous system.
  • the skeletal muscle is made up of very long tubular cells, also called muscle fibers, whose diameters range between 10 and 100 microns and that contain many nuclei located in the periphery of the cell.
  • Muscle fibers are separated from each other by a network of loose connective tissue composed of collagen and reticular fibers called endomysium.
  • endomysium a network of loose connective tissue composed of collagen and reticular fibers
  • the muscle fibers group together forming fascicles, which are surrounded by a connective tissue called perimysium.
  • the muscle fibers can be of two types normally referred to as type I and type II, both types of fibers being arranged according to a messy mosaic along the fascicles (Helliwell, 1999; O'Ferrall and Sinnreich, 2009; Pette and Staron, 2000 ).
  • the type of muscle fiber depends on the nature of the motor neuron that innervates it:
  • Type I fibers They contain a myosin isoform that uses ATP at low speed. They are slow contraction fibers.
  • Type II fibers They contain a myosin isoform that uses ATPase at high speed. They are fast contracting fibers.
  • the diagnosis of neuromuscular diseases is mainly based on histological characterization and morphological evaluation of sections of muscle samples, usually skeletal muscle biopsies (Dubowitz and Sewry, 2006). For this diagnosis, it is vitally important to determine the size and proportions between the two types of fibers, as well as the size of the collagen fibers, since the type of problem in the components of the muscular connection (motor neuron) axon-muscle) is reflected in a characteristic pattern of muscle section.
  • Dystrophic pattern It is characterized by an increase in endomysium due to the appearance of fibrosis. The fibers acquire a more rounded morphology. The affectation is usually homogeneous within the muscle, and affects all the muscles. The selectivity of affectation between type I and type II fibers is not frequent.
  • Non-dystrophic pattern The fibers also acquire a more rounded morphology, although without an increase in endomysium. A greater disparity between fiber sizes is observed
  • C. Neurogenic atrophy pattern A large number of normal fibers with the naked eye, although small groups appear without a specific pattern of small cells. It can affect some muscles more than others. It can present selectivity in the affectation of type I and type II fibers, although it does not always occur. Reinervations may appear, which which eliminates the mosaic. In very advanced cases, early fibrosis may appear.
  • pathologists Currently, the morphological evaluation of biopsies or muscle sections to make the diagnosis is performed visually by medical specialists called pathologists. It is easy to appreciate that there is a high degree of subjectivity in these diagnoses based on the visual interpretation of the pathologists, and differences may appear in the diagnoses depending on the degree of training, experience or expertise of each pathologist. In addition, pathologists are currently able to determine the presence or absence of a certain disease and estimate the degree of involvement or evolution in the patient, but cannot quantify it.
  • the present invention describes a method that applies recent discoveries related to network theory to obtain new specific parameters capable of characterizing a biopsy of a patient's skeletal muscle tissue.
  • the comparison of the characteristic vector corresponding to the patient's biopsy with the characteristic vector corresponding to a control biopsy will allow to determine in an objective way the existence of possible muscular diseases. Moreover, depending on the difference between some parameters and others, it will even be possible to determine the degree of patient involvement.
  • the method is specifically developed for skeletal muscle tissue that, as mentioned above, is formed by elongated muscle fibers grouped into fascicles. Each fiber is interpreted as a node of the network, while the contacts between fibers constitute the junctions between the nodes of the network.
  • the method of the invention recognizes two types of nodes depending on whether the muscle fiber is type I or type II, and also takes into account the collagen stored as an endomysium at the borders between fibers.
  • the method for obtaining useful information for the diagnosis of neuromuscular diseases from a skeletal muscle tissue biopsy comprises the following steps:
  • a first antibody that acts against collagen protein VI is applied to stain the endomysium of a first color (green) and a second antibody that acts against the heavy chain of the slow myosin isoform (specifically sMyHC) to dye type I muscle fibers of a second color (red).
  • Type II muscle fibers which have not been dyed, remain a dark color close to black.
  • a fluorescence microscope is preferably used, usually connected to a computer.
  • the image is obtained from an area of the biopsy devoid of artifacts caused during its preparation.
  • the result is an image that essentially contains a mosaic of red (type I) and black (type II) muscle fibers separated by green bands (endomysium).
  • a segmentation step is then carried out to clearly separate the different muscle fibers from the endomysium.
  • different resegmentation techniques known in the art could be used, although in a preferred embodiment of the invention the Watershed transform is applied.
  • This technique requires the design of markers that identify the fibers that we want to segment by applying different mathematical morphology operators. The objective is to find the local minimums of the image that indicate the presence of muscle fibers, and then apply the Watershed Transform that will already identify exactly the contour of each fiber.
  • the fibers are identified and the image is represented as a graph formed by nodes (muscle fibers) and joints between those nodes (each fiber / node is attached to the neighboring fibers / nodes, that is, those that they are in contact with her).
  • this step further comprises graphically representing the biopsy as a graph formed by the nodes and the junctions between nodes, the nodes being located at the center of mass of each fiber. 5) Form a characteristic vector of the biopsy whose elements are chosen between geometric parameters of the fibers and network parameters. Finally, a characteristic vector is constructed comprising at least one network parameter or geometric parameter of the fibers.
  • the parameters of the characteristic vector are chosen from the following list of 26 parameters:
  • Convex envelope Proportion of pixels of the convex envelope that also belong to the fiber, convex envelope being the smallest possible convex polygon that contains the fiber.
  • Angles angle formed by the major axis with the x axis.
  • Clustering coefficient mean: quantifies how interconnected it is with its neighbors. Proportion between links connected to their neighbors divided by the number of links in a dike in which connectivity is maximum.
  • Métrica_s Sum of the products of the degrees between two cells, understanding as degree the number of neighbors.
  • “Assortattivity” Pearson's coefficient of the degree between connected nodes.
  • the selection of the subgroup of parameters that will form the characteristic vector can be performed using the SFS (Sequential Forward Selection) and SBS (Sequential Backward Selection) methods by means of a Fuzzy-ARTMAP neural network.
  • RNA Artificial Neural Networks
  • the characteristic control vector can be obtained through the biopsy of an individual whose condition is known, that is, who is known to be healthy or sick. Note also that it is not necessary to have a single characteristic vector of the patient being evaluated and a single characteristic control vector, but it is possible to construct a group of characteristic vectors of the patient from one or several biopsies and compare it with a group of vectors control characteristics corresponding to individuals of known, preferably healthy characteristics.
  • control characteristic vector may refer not only to a characteristic vector corresponding to a real biopsy of an individual, but also to a vector characteristic formed by elements considered threshold between normal values and altered values. For example, it would be possible to determine, through studies of healthy and diseased individuals, a set of threshold values of the parameters described above corresponding to the boundary between normal values and values potentially corresponding to the presence of muscle diseases.
  • the selection of the subgroup of parameters that will form the characteristic vector is performed by looking for that subgroup of parameters that maximizes the value of an ACP descriptor of the Calinski-Harabasz type.
  • the ACP descriptor used gives an idea of the goodness with which a certain subgroup of parameters manages to separate a group of vectors (corresponding to the patient sample) from another group of vectors (corresponding to the control).
  • the matrices W and B are added to obtain the dispersion matrix T of the data set. It can be expected that compact and separate subgroups have low values of W and high values of B. Consequently, the better the partition of the data, the greater the value of the relationship between B / W that constitutes the ACP descriptor.
  • the proposed method is based primarily on testing different combinations of parameters until obtaining a subset of parameters that maximizes the value of said ACP descriptor. This can be done in different ways, although in a preferred embodiment of the invention it is implemented by the following steps:
  • the described method can be carried out by means of a computer program, and therefore the invention also extends to computer programs, particularly to computer programs located on a carrier, which are adapted so that a computer Put the invention into practice.
  • the program may have the form of source code, object code, an intermediate source of code and object code, for example, as in partially compiled form, or in any other form suitable for use in the implementation of the processes according to the invention .
  • the carrier can be any entity or device capable of supporting the program.
  • the carrier could include a storage medium, such as a ROM, a CD ROM, a ROM memory of semiconductor, or a hard drive.
  • the carrier can also be a transmissible carrier, for example, an electrical or optical signal that could be transported through electrical or optical cable, by radio or by any other means.
  • the carrier may be constituted by said cable or other device or means.
  • the carrier could be an integrated circuit in which the program is included, the integrated circuit being adapted to execute, or to be used in the execution of, the corresponding processes.
  • Fig. 1 shows an example of a patient's muscle biopsy before staining.
  • Fig. 2 shows a biopsy similar to that of Fig. 1 once stained with antibodies against collagen protein VI and with antibodies against the heavy chain of the slow myosin isoform.
  • Fig. 3 shows the biopsy of Fig. 2 after the segmentation is finished.
  • Fig. 4 shows a detail of the biopsy of Fig. 3 once the network is formed.
  • Fig. 5 is a graph where each point corresponds to the characteristic vector of a patient with dystrophy (circles) or of a healthy individual (squares). Each circle, which corresponds to an individual with dystrophy, has numbers that indicate the degree of involvement according to an expert neuropathologist.
  • the first step is the acquisition of biopsies of a patient's muscle tissue, which are subsequently processed by standard freezing and cryostat cutting methods.
  • Fig. 1 shows the appearance of an example of a biopsy before staining.
  • the biopsies receive an immunohistochemical staining with a standard fluorescence staining protocol. As described above, double stains are performed, adding a different secondary antibody for each primary antibody. The result is preparations stained at the same time with two antibodies, one that detects the endomysium and the other that detects type I fibers. Specifically, in this example the biopsies are stained with a first antibody against the collagen VI protein to identify the endomysium , which is stained green, and with a second specific antibody against the heavy chain of the slow myosin isoform, specifically the sMyHC, to identify type I fibers, which are stained red.
  • an image of the biopsy is acquired using a fluorescence microscope.
  • the image is obtained from a part of the biopsy that has no artifact due to the preparation.
  • images are usually acquired with the same resolution and size.
  • the segmentation process begins that will allow the precise identification of the contours of the different fibers, both type I and type II. Once the contours of the fibers have been identified, everything other than fiber will be considered endomysium.
  • Segmentation is performed using the Watershed technique, using markers to achieve proper segmentation of muscle fibers with this technique.
  • the purpose of the markers is to identify the presence of all the fibers in the image. It is carried out through the use of morphological operators, which are detailed below.
  • the green plane is binarized, putting at a value of 1 those pixels that form related regions with a constant intensity value less than a threshold and whose edges are formed by pixels with greater intensity value.
  • the rest of the pixels in the image acquire a value of 0.
  • regions with an area smaller than a threshold are eliminated.
  • a morphological reconstruction is performed that allows filling gaps, meaning gaps those regions of pixels with value 0 surrounded by regions with value 1. It continues with an opening.
  • erosion is carried out, to prevent regions with a value of 1 that identify different fibers from being in contact.
  • regions with value 1 In order not to segment layers as fibers, several color conditions are applied to regions with value 1. These regions have to meet that their average intensity value in the G plane (Green plane) is less than a threshold, and on the other hand that its intensity value in the R plane (Red plane) is less than a threshold and its average value of the G plane when its histogram is equalized is greater than a threshold.
  • the markers are obtained, which are those regions with value 1 resulting in the binary image, and that identify the presence of fibers.
  • the Watershed transform is applied to these markers next to the gradient of the green plane obtaining the precise segmentation of the contours of the muscle fibers. The result is shown in Fig. 3, where the contours of the muscle fibers appear in white, while the fibers themselves and the endomysium still have the same tones described above in relation to Fig. 2.
  • the segmented image of Fig. 2 is processed by expanding equally the detected surface of each fiber until contacting the expansions of the neighboring fibers. It is done by directly applying the Watershed Transform to the binary image obtained from the found markers.
  • the next step is the creation of a network or graph representative of the biopsy, where the nodes correspond to the different fibers and the joints connect each fiber with its neighboring fibers.
  • This network is represented as shown in Fig. 4, where the gray lines of a lighter color represent the contours of the different fibers / nodes of the network, while the darker (almost black) lines represent the joints between some nodes / fibers and others.
  • the nodes themselves are represented by the intersections between several dark lines, and are located in the center of mass of the corresponding fibers / nodes.
  • the network created comprises nodes of two types, which correspond to the two types of fibers (type I and type II).
  • the network is formed, it is easy to calculate different parameters, either geometric or network parameters, in order to characterize the biopsy and allow its comparison with certain thresholds obtained empirically or with the parameters corresponding to control biopsies. For example, you can determine the area of the different fibers, the length of their axes, the number of neighbors each fiber has, etc. To do this, a region of interest (ROI) that meets the following conditions is previously selected:
  • the fibers of the outer limits of the ROI must be surrounded by at least two additional rows of cells.
  • the characteristic vector of the sample is constructed using, for example, the SFS (Sequential Forward Selection) and SBS (Sequential Backward Selection) methods through a Fuzzy-ARTMAP neural network, with the purpose of achieving a characteristic vector capable of discriminating between healthy biopsies and biopsies affected by some primary myopathy (dystrophic or non-dystrophic) or neurogenic atrophy.
  • SFS Sequential Forward Selection
  • SBS Simential Backward Selection
  • Fig. 5 shows an example where the selected characteristic vector is formed by the characteristics 19 (average of the relation of axes), 23 (average angles) and 25 (average fiber area ratio / average area of neighboring fibers) of the previous table.
  • ACP Principal Component Analysis
  • ACP analysis is an objective method that uses the values of the selected characteristics to make the comparison between two or more groups of images. As a result, a projection in two or three dimensions is obtained that maximizes the dispersion of each of the images.
  • This visualization allows the quantification of the differences between groups formed by each type of data.
  • Each image or point of Fig. 5 therefore corresponds to the characteristic vector of a patient with dystrophy or of a healthy individual. It can be seen how both groups appear separated in Fig. 5, with the squares corresponding to healthy individuals mainly to the left of the graph and the circles corresponding to individuals with dystrophies mainly to the right of the graph.
  • Fig. 5 it can be seen that in most cases the evaluation of the degree of affectation carried out by the pathologist, which is shown as a number next to each circle corresponding to patients with muscular dystrophies, correlates directly with the distance between said point and the center of mass of the group of squares corresponding to control patients. That is, basically the further to the right is the circle corresponding to the characteristic vector, the greater the degree of affectation of the dystrophy corresponding to that biopsy.
  • the image analysis method provides useful data for the diagnosis of muscular diseases in a fast, automatic and objective way.
  • KNN method K nearest neighbors, or nearest neighbors according to its acronym in English
  • Muscle Part 1 - Normal structure and function.

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Investigating Or Analysing Biological Materials (AREA)

Abstract

Method for obtaining useful information for the diagnosis of neuromuscular diseases. The invention describes a method for obtaining, on the basis of a patient biopsy, new parameters that allow objective diagnosis of different neuromuscular diseases and the degree to which these affect the patient, comprising: -staining the biopsy sample in order to highlight type-I muscle fibres, type-II muscle fibres and endomysium; - obtaining an image of the biopsy sample after staining; - dividing the image into segments to identify the outlines of the muscle fibres; - forming a network in which the muscle fibres constitute nodes and the contacts between muscle fibres constitute the connections between the nodes; - forming a vector characteristic of the biopsy sample, the elements of which are chosen from geometric parameters of the fibres and parameters of the network constructed; - and comparing control and affected biopsy samples by means of ACP using, in each case, the characteristic vector selected.

Description

MÉTODO PARA OBTENER INFORMACIÓN ÚTIL PARA EL DIAGNÓSTICO DE ENFERMEDADES NEUROMUSCULARES  METHOD FOR OBTAINING USEFUL INFORMATION FOR THE DIAGNOSIS OF NEUROMUSCULAR DISEASES
OBJETO DE LA INVENCIÓN OBJECT OF THE INVENTION
El objeto de la presente invención es un método para obtener, a partir de la biopsia de un paciente, un conjunto de nuevos parámetros que permiten, mediante un estudio histológico comparativo, diagnosticar de manera objetiva diferentes enfermedades neuromusculares y su grado de afectación al paciente. Además, el método puede realizarse de manera automática por ordenador. The object of the present invention is a method to obtain, from the biopsy of a patient, a set of new parameters that allow, by means of a comparative histological study, to objectively diagnose different neuromuscular diseases and their degree of involvement to the patient. In addition, the method can be performed automatically by computer.
ANTECEDENTES DE LA INVENCIÓN El músculo esquelético constituye el músculo voluntario y está ¡nervado por las neuronas motoras del sistema nervioso somático. El músculo esquelético está constituido por células tubulares muy largas, también denominadas fibras musculares, cuyos diámetros oscilan entre 10 y 100 mieras y que contienen muchos núcleos ubicados en la periferia de la célula. Las fibras musculares están separadas unas de otras por una red de tejido conectivo laxo compuesto de fibras de colágeno y reticulares denominado endomisio. A su vez, las fibras musculares se agrupan formando fascículos, que están rodeados por un tejido conectivo denominado perimisio. Las fibras musculares pueden ser de dos tipos normalmente denominados tipo I y tipo II, estando dispuestos ambos tipos de fibras según un mosaico desordenado a lo largo de los fascículos (Helliwell, 1999; O'Ferrall and Sinnreich, 2009; Pette and Staron, 2000). El tipo de fibra muscular depende de la naturaleza de la neurona motora que la inerva: BACKGROUND OF THE INVENTION Skeletal muscle constitutes the voluntary muscle and is supplied by motor neurons of the somatic nervous system. The skeletal muscle is made up of very long tubular cells, also called muscle fibers, whose diameters range between 10 and 100 microns and that contain many nuclei located in the periphery of the cell. Muscle fibers are separated from each other by a network of loose connective tissue composed of collagen and reticular fibers called endomysium. In turn, the muscle fibers group together forming fascicles, which are surrounded by a connective tissue called perimysium. The muscle fibers can be of two types normally referred to as type I and type II, both types of fibers being arranged according to a messy mosaic along the fascicles (Helliwell, 1999; O'Ferrall and Sinnreich, 2009; Pette and Staron, 2000 ). The type of muscle fiber depends on the nature of the motor neuron that innervates it:
Fibras tipo I: Contienen una isoforma de miosina que utiliza ATP a baja velocidad. Son fibras de contracción lenta. Fibras tipo II: Contienen una isoforma de miosina que utiliza ATPasa a alta velocidad. Son fibras de contracción rápida. Type I fibers: They contain a myosin isoform that uses ATP at low speed. They are slow contraction fibers. Type II fibers: They contain a myosin isoform that uses ATPase at high speed. They are fast contracting fibers.
El diagnóstico de las enfermedades neuromusculares está principalmente basado en la caracterización histológica y la evaluación morfológica de secciones de muestras musculares, normalmente biopsias del músculo esquelético (Dubowitz and Sewry, 2006). Para este diagnóstico, es de vital importancia la determinación del tamaño y de las proporciones entre los dos tipos de fibras, así como el tamaño de las fibras de colágeno, ya que el tipo de problema en los componentes de la conexión muscular (neurona motora- axón-músculo) se refleja en un patrón característico de la sección muscular. The diagnosis of neuromuscular diseases is mainly based on histological characterization and morphological evaluation of sections of muscle samples, usually skeletal muscle biopsies (Dubowitz and Sewry, 2006). For this diagnosis, it is vitally important to determine the size and proportions between the two types of fibers, as well as the size of the collagen fibers, since the type of problem in the components of the muscular connection (motor neuron) axon-muscle) is reflected in a characteristic pattern of muscle section.
Principalmente se pueden distinguir tres patrones característicos correspondientes a dos grupos de enfermedades: Mainly, three characteristic patterns corresponding to two groups of diseases can be distinguished:
1 ) Miopatía con origen en el propio músculo 1) Myopathy with origin in the muscle itself
a. Patrón distrófico: Se caracteriza por un aumento de endomisio debido a la aparición de fibrosis. Las fibras adquieren una morfología más redondeada. La afectación suele ser homogénea dentro del músculo, y afecta a todos los músculos. No es frecuente la selectividad de afectación entre fibras tipo I y tipo II.  to. Dystrophic pattern: It is characterized by an increase in endomysium due to the appearance of fibrosis. The fibers acquire a more rounded morphology. The affectation is usually homogeneous within the muscle, and affects all the muscles. The selectivity of affectation between type I and type II fibers is not frequent.
b. Patrón no distrófico: Las fibras adquieren también una morfología más redondeada, aunque sin aumento de endomisio. Se observa una mayor disparidad entre los tamaños de las fibras  b. Non-dystrophic pattern: The fibers also acquire a more rounded morphology, although without an increase in endomysium. A greater disparity between fiber sizes is observed
2) Patología neuromuscular con origen en el sistema nervioso periférico 2) Neuromuscular pathology with origin in the peripheral nervous system
c. Patrón de atrofia neurógena: Gran cantidad de fibras normales a simple vista, aunque aparecen pequeños grupos sin patrón específico de células pequeñas. Puede afectar más a unos músculos que a otros. Puede presentar selectividad en la afectación de fibras tipo I y tipo II, aunque no ocurre siempre. Pueden aparecer reinervaciones, lo cual elimina el mosaico. En casos muy avanzados, pueden aparecer principios de fibrosis. C. Neurogenic atrophy pattern: A large number of normal fibers with the naked eye, although small groups appear without a specific pattern of small cells. It can affect some muscles more than others. It can present selectivity in the affectation of type I and type II fibers, although it does not always occur. Reinervations may appear, which which eliminates the mosaic. In very advanced cases, early fibrosis may appear.
Actualmente, la evaluación morfológica de las biopsias o secciones musculares para realizar el diagnóstico es realizada visualmente por médicos especialistas denominados patólogos. Es fácil apreciar que existe un elevado grado de subjetividad en estos diagnósticos basados en la interpretación visual de los patólogos, pudiendo aparecer diferencias en los diagnósticos en función del grado de formación, de la experiencia o de la pericia de cada patólogo. Además, actualmente los patólogos son capaces de determinar la presencia o ausencia de una determinada enfermedad y de estimar el grado de afectación o de evolución en el paciente, pero no pueden cuantificarlo. Currently, the morphological evaluation of biopsies or muscle sections to make the diagnosis is performed visually by medical specialists called pathologists. It is easy to appreciate that there is a high degree of subjectivity in these diagnoses based on the visual interpretation of the pathologists, and differences may appear in the diagnoses depending on the degree of training, experience or expertise of each pathologist. In addition, pathologists are currently able to determine the presence or absence of a certain disease and estimate the degree of involvement or evolution in the patient, but cannot quantify it.
Recientemente se ha descubierto la utilidad de aplicar la teoría de redes a la caracterización de la organización de los tejidos. La teoría de redes se ha utilizado ampliamente para analizar diversos tipos de procesos o sistemas biológicos complejos a diferentes escalas, desde interacciones entre moléculas hasta interacciones entre especies. Sin embargo, su aplicación a nivel de individuo o de grupos de células ha sido lim itada hasta la publicación del artículo "Epithelial organisation revealed by a network of cellular contacts", de Luis M. Escudero et al. (Escudero et al., 201 1 ). El método descrito en este artículo está basado en el análisis de las imágenes de tejidos la mosca Drosophila y de pollo, que están básicamente constituidas por simples células poligonales de diferentes tamaños según el tipo de epitelio. De este análisis se obtienen una serie de parámetros fundamentalmente basados en el tamaño, forma y conectividad de dichas células, construyéndose un vector característico de cada muestra. Se descubre entonces que el vector característico de cada muestra permite identificar cambios en el epitelio como consecuencia de la mutación de un gen, diferenciar entre tejidos de uno u otro organismo, e incluso identificar diferentes estados de crecimiento de un mismo organismo, todo ello de forma automática y objetiva. DESCRIPCIÓN DE LA INVENCIÓN Recently, the utility of applying network theory to the characterization of tissue organization has been discovered. Network theory has been widely used to analyze various types of complex biological processes or systems at different scales, from interactions between molecules to interactions between species. However, its application at the individual or cell group level has been limited until the publication of the article "Epithelial organization revealed by a network of cellular contacts", by Luis M. Escudero et al. (Escudero et al., 201 1). The method described in this article is based on the analysis of the Drosophila and chicken fly tissue images, which are basically constituted by simple polygonal cells of different sizes depending on the type of epithelium. From this analysis, a series of parameters are obtained fundamentally based on the size, shape and connectivity of said cells, building a characteristic vector of each sample. It is then discovered that the characteristic vector of each sample makes it possible to identify changes in the epithelium as a result of the mutation of a gene, differentiate between tissues of one or another organism, and even identify different growth states of the same organism, all in a way Automatic and objective. DESCRIPTION OF THE INVENTION
La presente invención describe un método que aplica los recientes descubrimientos relacionados con la teoría de redes para obtener nuevos parámetros específicos capaces de caracterizar una biopsia de tej ido de músculo esquelético de un paciente. La comparación del vector característico correspondiente a la biopsia del paciente con el vector característico correspondiente a una biopsia control permitirá determinar de un modo objetivo la existencia de posibles enfermedades musculares. Es más, en función de la diferencia entre unos parámetros y otros, será posible incluso determinar el grado de afectación del paciente. The present invention describes a method that applies recent discoveries related to network theory to obtain new specific parameters capable of characterizing a biopsy of a patient's skeletal muscle tissue. The comparison of the characteristic vector corresponding to the patient's biopsy with the characteristic vector corresponding to a control biopsy will allow to determine in an objective way the existence of possible muscular diseases. Moreover, depending on the difference between some parameters and others, it will even be possible to determine the degree of patient involvement.
El método está desarrollado específicamente para el tejido de músculo esquelético que, como se ha mencionado anteriormente, está formado por fibras musculares alargadas agrupadas en fascículos. Cada fibra se interpreta como un nodo de la red, mientras que los contactos entre fibras constituyen las uniones entre los nodos de la red. El método de la invención reconoce dos tipos de nodos según si la fibra muscular es de tipo I o de tipo II, y además tiene en cuenta el colágeno almacenado en forma de endomisio en las fronteras entre fibras. The method is specifically developed for skeletal muscle tissue that, as mentioned above, is formed by elongated muscle fibers grouped into fascicles. Each fiber is interpreted as a node of the network, while the contacts between fibers constitute the junctions between the nodes of the network. The method of the invention recognizes two types of nodes depending on whether the muscle fiber is type I or type II, and also takes into account the collagen stored as an endomysium at the borders between fibers.
Según un primer aspecto de la invención, el método para obtener información útil para el diagnóstico de enfermedades neuromusculares a partir de una biopsia de tejido de músculo esquelético comprende los siguientes pasos: According to a first aspect of the invention, the method for obtaining useful information for the diagnosis of neuromuscular diseases from a skeletal muscle tissue biopsy comprises the following steps:
1 ) Realizar una tinción de la biopsia para resaltar las fibras musculares tipo I, las fibras musculares tipo II y el endomisio. El objeto de la tinción es poder identificar visualmente en la muestra los dos tipos de fibras musculares (tipo I y tipo II), así como el endomisio. En principio, se podría aplicar cualquier tipo de tinción capaz de conseguir este propósito, aunque preferentemente se aplican preparaciones de dos anticuerpos que tiñen de colores diferentes dos elementos de entre los tres mencionados. Es decir, se tiñen dos de ellos de colores diferentes, quedando el color del tercero sin modificar. 1) Perform a biopsy stain to highlight type I muscle fibers, type II muscle fibers and the endomysium. The purpose of staining is to be able to visually identify in the sample the two types of muscle fibers (type I and type II), as well as the endomysium. In principle, any type of staining capable of achieving this could be applied purpose, although preferably preparations of two antibodies that stain two different colors from among the three mentioned are applied. That is, two of them are dyed in different colors, leaving the color of the third unmodified.
Más concretamente, en una realización preferida de la invención se aplica un primer anticuerpo que actúa contra la proteína colágeno VI para teñir el endomisio de un primer color (verde) y un segundo anticuerpo que actúa contra la cadena pesada de la isoforma lenta de la miosina (concretamente sMyHC) para teñir las fibras musculares tipo I de un segundo color (rojo). Las fibras musculares tipo II, que no han sido teñidas, se mantienen de un color oscuro cercano al negro. More specifically, in a preferred embodiment of the invention, a first antibody that acts against collagen protein VI is applied to stain the endomysium of a first color (green) and a second antibody that acts against the heavy chain of the slow myosin isoform (specifically sMyHC) to dye type I muscle fibers of a second color (red). Type II muscle fibers, which have not been dyed, remain a dark color close to black.
2) Obtener una imagen de la biopsia tras la tinción. 2) Obtain an image of the biopsy after staining.
Para obtener la imagen se utiliza preferentemente un microscopio de fluorescencia, normalmente conectado a un ordenador. La imagen se obtiene de una zona de la biopsia carente de artefactos causados durante la preparación de la misma. El resultado es una imagen que contiene fundamentalmente un mosaico de fibras musculares de color rojo (tipo I) y negro (tipo II) separadas por bandas de color verde (endomisio). To obtain the image, a fluorescence microscope is preferably used, usually connected to a computer. The image is obtained from an area of the biopsy devoid of artifacts caused during its preparation. The result is an image that essentially contains a mosaic of red (type I) and black (type II) muscle fibers separated by green bands (endomysium).
3) Segmentar la imagen para identificar los contornos de las fibras musculares. 3) Segment the image to identify the contours of muscle fibers.
Se lleva a cabo a continuación un paso de segmentación para separar de un modo claro las diferentes fibras musculares del endomisio. En principio, se podrían utilizar diferentes técnicas resegmentación conocidas en la técnica, aunque en una realización preferida de la invención se aplica la transformada de Watershed. Esta técnica requiere del diseño de marcadores que identifiquen las fibras que deseamos segmentar mediante la aplicación de distintos operadores de morfología matemática. El objetivo es encontrar los mínimos locales de la imagen que indiquen la presencia de fibras musculares, para después aplicar la Transformada Watershed que ya sí identificará de una manera exacta el contorno de cada fibra. Para evitar que algunos capilares se confundan con dichas fibras, se aplican condiciones de color, basadas en los planos R (rojo), G (verde) y en el plano G cuando se realiza una ecualización del histograma 4) Formar una red donde las fibras musculares constituyen nodos y los contactos entre fibras musculares constituyen las uniones entre los nodos. A segmentation step is then carried out to clearly separate the different muscle fibers from the endomysium. In principle, different resegmentation techniques known in the art could be used, although in a preferred embodiment of the invention the Watershed transform is applied. This technique requires the design of markers that identify the fibers that we want to segment by applying different mathematical morphology operators. The objective is to find the local minimums of the image that indicate the presence of muscle fibers, and then apply the Watershed Transform that will already identify exactly the contour of each fiber. To prevent some capillaries from being confused with these fibers, color conditions are applied, based on the R (red), G (green) and G plane when an equalization of the histogram is performed 4) Form a network where the fibers Muscles constitute nodes and contacts between muscle fibers constitute the junctions between the nodes.
Una vez realizada la segmentación, se identifican las fibras y se representa la imagen como un grafo formado por nodos (las fibras musculares) y uniones entre esos nodos (cada fibra/nodo está unida a las fibras/nodos vecinos, es decir, aquellos que están en contacto con ella). Once the segmentation is done, the fibers are identified and the image is represented as a graph formed by nodes (muscle fibers) and joints between those nodes (each fiber / node is attached to the neighboring fibers / nodes, that is, those that they are in contact with her).
Además, puesto que existen fibras de tipo I y fibras de tipo II, también existirán nodos de tipo I y nodos de tipo II. La información acerca del tipo de nodo se utilizará para determ inar diferentes parámetros de red, según se describirá más adelante en este documento. Por ejemplo, permitirá calcular cuántos vecinos de tipo I tiene una fibra de tipo II, el tamaño medio de las fibras de tipo I, y otros parámetros similares. Según una realización preferida de la invención, este paso comprende además representar gráficamente la biopsia como un grafo formado por los nodos y las uniones entre nodos, estando los nodos ubicados en el centro de masa de cada fibra. 5) Formar un vector característico de la biopsia cuyos elementos se eligen entre parámetros geométricos de las fibras y parámetros de red. Por último, se construye un vector característico que comprende al menos un parámetro de red o parámetro geométrico de las fibras. In addition, since there are type I fibers and type II fibers, there will also be type I nodes and type II nodes. Information about the type of node will be used to determine different network parameters, as will be described later in this document. For example, it will allow you to calculate how many type I neighbors a type II fiber has, the average size of type I fibers, and other similar parameters. According to a preferred embodiment of the invention, this step further comprises graphically representing the biopsy as a graph formed by the nodes and the junctions between nodes, the nodes being located at the center of mass of each fiber. 5) Form a characteristic vector of the biopsy whose elements are chosen between geometric parameters of the fibers and network parameters. Finally, a characteristic vector is constructed comprising at least one network parameter or geometric parameter of the fibers.
En una realización preferida de la invención, los parámetros del vector característico se eligen entre la siguiente lista de 26 parámetros:  In a preferred embodiment of the invention, the parameters of the characteristic vector are chosen from the following list of 26 parameters:
1 Area media de las fibras 1 Average fiber area
2 Desviación típica del área de las fibras  2 Standard deviation of the fiber area
3 Número medio de vecinos de cada fibra  3 Average number of neighbors of each fiber
4 Desviación típica del número de vecinos de cada fibra  4 Standard deviation of the number of neighbors of each fiber
5 Area media de las fibras tipo I  5 Average area of type I fibers
6 Desviación típica del área de las fibras tipo I  6 Standard deviation of the type I fiber area
7 Area media de las fibras tipo II  7 Average area of type II fibers
8 Desviación típica del área de las fibras tipo II  8 Standard deviation of type II fiber area
9 Desviación típica del número de vecinos de las fibras tipo I  9 Standard deviation of the number of neighbors of type I fibers
10 Desviación típica del número de vecinos de las fibras tipo II  10 Standard deviation of the number of neighbors of type II fibers
1 1 Número medio de vecinos tipo I de las fibras tipo I  1 1 Average number of type I neighbors of type I fibers
12 Número medio de vecinos tipo II de las fibras tipo I  12 Average number of type II neighbors of type I fibers
13 Número medio de vecinos tipo I de las fibras tipo II  13 Average number of type I neighbors of type II fibers
14 Número medio de vecinos tipo II de las fibras tipo II  14 Average number of type II neighbors of type II fibers
15 Media del cociente A1/A2  15 Mean ratio A1 / A2
16 Desviación típica del cociente A1/A2  16 Standard deviation of the ratio A1 / A2
17 Dimensión media del eje mayor de las fibras  17 Average dimension of the major axis of the fibers
18 Dimensión media del eje menor de las fibras  18 Average dimension of the minor axis of the fibers
19 Media de la relación de ejes  19 Mean of axes ratio
20 Desviación típica de la relación de ejes  20 Standard deviation of the axis ratio
21 Media de la envoltura convexa  21 Convex sheath stocking
22 Desviación típica de la envoltura convexa  22 Standard deviation of the convex envelope
23 Angulos medios  23 Middle angles
24 Desviación típica de los ángulos  24 Standard deviation of angles
25 Media cociente área fibra/ área media fibras vecinas  25 Half ratio fiber area / average area neighboring fibers
26 Desviación típica cociente área fibra/ área media fibras vecinas Más preferentemente, se pueden añadir además los siguientes parámetros a los anteriores, de modo que número de parámetros total alcanza 82. 26 Standard deviation fiber area ratio / average area neighboring fibers More preferably, the following parameters can also be added to the previous ones, so that the total number of parameters reaches 82.
27 Media de la relación con los vecinos del eje mayor 27 Average of the relationship with the neighbors of the major axis
28 Desv. típica de la relación con los vecinos del eje mayor  28 Def. typical of the relationship with the neighbors of the major axis
29 Media de la relación con los vecinos del eje menor  29 Average of the relationship with the neighbors of the minor axis
30 Desv. típica de la relación con los vecinos del eje menor  30 Def. typical of the relationship with the minor axis neighbors
31 Media de la relación con los vecinos de la relación entre ejes  31 Average of the relationship with the neighbors of the relationship between axes
32 Des. típica de la relación con los vecinos de la relación entre ejes 32 Off typical of the relationship with the neighbors of the relationship between axes
33 Media de la relación con los vecinos de la envoltura convexa 33 Average of the relationship with the neighbors of the convex envelope
34 Des. típica de la relación con los vecinos de la envoltura convexa 34 Off typical of the relationship with the neighbors of the convex envelope
35 Media de la relación con los vecinos de ángulos medios 35 Average of the relationship with the neighbors of average angles
36 Desv. típica de la relación con los vecinos ángulos medios  36 Def. typical of the relationship with the neighbors average angles
37 Media de la relación con los vecinos de la relación A1/A2  37 Average of the relationship with the neighbors of the A1 / A2 relationship
38 Desv. típica de la relación con los vecinos de la relación A1/A2  38 Def. typical of the relationship with the neighbors of the A1 / A2 relationship
39 Media de la Suma de los pesos  39 Average of the sum of the weights
40 Desv. típica de la Suma de los pesos  40 Dev. typical of the sum of the weights
41 Media de los pesos de las fibras tipo I  41 Average of the weights of type I fibers
42 Desv. típica de los pesos de las fibras tipo I  42 Def. typical of type I fiber weights
43 Media de los pesos de las fibras tipo II  43 Average of the weights of type II fibers
44 Desv. típica de los pesos de las fibras tipo II  44 Def. typical of type II fiber weights
45 Media del "clustering coefficient"  45 Clustering coefficient average
46 Desv. típica del "clustering coefficient"  46 Def. typical of "clustering coefficient"
47 Media del "clustering coefficient" fibras tipo II  47 Average of the "clustering coefficient" type II fibers
48 Desv. típica del "clustering coefficient" fibras tipo II  48 Def. typical of "clustering coefficient" type II fibers
49 Media del "clustering coefficient" fibras tipo II  49 Average of the "clustering coefficient" type II fibers
50 Desv. típica del "clustering coefficient" tipo II  50 Def. typical of "clustering coefficient" type II
51 Media de la excentricidad  51 Eccentricity mean
52 Desv. típica de la excentricidad  52 Def. typical of eccentricity
53 Media de la excentricidad de las fibras tipo II Desv. típica de la excentricidad de las fibras tipo II 53 Average eccentricity of type II fibers Dev. typical of the eccentricity of type II fibers
Media de la excentricidad de las fibras tipo II  Average eccentricity of type II fibers
Desv. Típica de la excentricidad de las fibras tipo II  Dev. Typical of the eccentricity of type II fibers
Media de la "Betweeness Centrality"  Average of "Betweeness Centrality"
Desv. típica de la "Betweeness Centrality"  Dev. typical of the "Betweeness Centrality"
Media de la "Betweeness Centrality" de las fibras tipo II  Average of the "Betweeness Centrality" of type II fibers
Desv. típica de la "Betweeness Centrality" de las fibras tipo II Dev. typical of the "Betweeness Centrality" of type II fibers
Media de la "Betweeness Centrality" de las fibras tipo II Average of the "Betweeness Centrality" of type II fibers
Desv. típica de la "Betweeness Centrality" de las fibras tipo II Dev. typical of the "Betweeness Centrality" of type II fibers
Media de la distancia Average distance
Desv. típica de la distancia  Dev. typical of distance
Media de la distancia de las fibras tipo I l_f i bras tipo II  Mean distance of fibers type I l_f and bras type II
Desv. típica de la distancia de las fibras tipo ll_fibras tipo II Dev. typical of the distance of fibers type ll_fibres type II
Media de la distancia de las fibras tipo II a las fibras tipo IIAverage distance of type II fibers to type II fibers
Desv. típica de la distancia de las fibras tipo II a las fibras tipo IIDev. typical of the distance of type II fibers to type II fibers
Media de la distancia de las fibras fibras tipo II a las fibras tipo IIAverage distance of fibers type II fibers to type II fibers
Desv. típica de la distancia de las fibras tipo II a las fibras tipo IIDev. typical of the distance of type II fibers to type II fibers
Media de la distancia de las fibras tipo I a las fibras tipo IAverage distance of type I fibers to type I fibers
Desv. típica de la distancia de las fibras tipo II a las fibras tipo IDev. typical of the distance of type II fibers to type I fibers
Radio Radio
Diámetro  Diameter
Eficiencia  Efficiency
Coeficiente de Pearson  Pearson coefficient
Conectividad algebraica  Algebraic connectivity
Metrica_s  Metrica_s
"Assortattivity"  "Assortattivity"
Densidad de conexiones  Connection density
"Transitivity"  "Transitivity"
Modularidad maximizada donde: - Cociente A1/A2: área de cada fibra tras la segmentación partida por el área de la fibra tras la expansión proporcional Maximized modularity where: - Ratio A1 / A2: area of each fiber after segmentation split by the fiber area after proportional expansion
- Eje mayor/menor: es el eje mayor/menor de la elipse que tiene el mismo segundo momento central normalizado que la fibra.  - Major / minor axis: it is the major / minor axis of the ellipse that has the same second central normalized moment as the fiber.
- Relación de ejes: es la relación entre el eje mayor y eje menor de una fibra, entendiendo como tales los ejes mayor y menor de la elipse que tiene el mismo segundo momento central normalizado que la fibra.  - Axis relation: it is the relation between the major axis and minor axis of a fiber, understanding as such the major and minor axes of the ellipse that has the same second central normalized moment as the fiber.
- Envoltura convexa: Proporción de píxeles de la envoltura convexa que también pertenecen a la fibra, entendiendo por envoltura convexa el polígono convexo más pequeño posible que contiene a la fibra.  - Convex envelope: Proportion of pixels of the convex envelope that also belong to the fiber, convex envelope being the smallest possible convex polygon that contains the fiber.
- Ángulos: ángulo que forma el eje mayor con el eje x.  - Angles: angle formed by the major axis with the x axis.
- Media de la suma de los pesos: es la media de la suma de la distancia en píxeles entre el centro de cada célula y los centros de las células vecinas.  - Average of the sum of the weights: it is the average of the sum of the distance in pixels between the center of each cell and the centers of the neighboring cells.
- Media del "clustering coefficient": cuantifica cómo de interconectado está con sus vecinos. Proporción entre los enlaces conectados con sus vecinos dividido entre el número de enlaces existentes en un dique en el que la conectividad es máxima.  - Clustering coefficient mean: quantifies how interconnected it is with its neighbors. Proportion between links connected to their neighbors divided by the number of links in a dike in which connectivity is maximum.
- Media de la excentricidad: media de los valores máximos de las longitudes mínimas de los caminos desde una célula a cualquier otra.  - Eccentricity mean: average of the maximum values of the minimum lengths of roads from one cell to any other.
- Media de la "betweeness centrality": es la fracción entre todos los caminos de longitud mínima que pasan por una célula y el número total de caminos de longitud mínima que comienzan en esa célula.  - Average of the "betweeness centrality": it is the fraction between all the paths of minimum length that pass through a cell and the total number of paths of minimum length that begin in that cell.
- Media de la distancia: longitud (en píxeles) del camino más corto entre todos los pares de células.  - Average distance: length (in pixels) of the shortest path between all cell pairs.
- Radio: el menor camino más corto.  - Radio: the shortest shortest path.
- Diámetro: el mayor camino más corto.  - Diameter: the largest shortest path.
- Eficiencia: inversa de la media de los caminos más cortos.  - Efficiency: inverse of the average of the shortest paths.
- Coeficiente de Pearson: coeficiente de Pearson del grafo.  - Pearson coefficient: Pearson's coefficient of the graph.
- Conectividad algebraica: segundo autovalor más pequeño no nulo de la matriz laplaciana.  - Algebraic connectivity: second smallest non-zero self-assessment of the Laplacian matrix.
- Métrica_s: Suma de los productos de los grados entre dos células, entendiendo como grado el número de vecinos. - "Assortattivity": coeficiente de Pearson del grado entre nodos conectados. - Métrica_s: Sum of the products of the degrees between two cells, understanding as degree the number of neighbors. - "Assortattivity": Pearson's coefficient of the degree between connected nodes.
- Densidad de conexiones: fracción entre el número de conexiones entre las células y el número total de conexiones posibles.  - Density of connections: fraction between the number of connections between the cells and the total number of possible connections.
- "Transitivity": probabilidad de que células adyacentes a otra estén conectadas.  - "Transitivity": probability that cells adjacent to another are connected.
- Modularidad maximizada: es un estadístico que cuantifica el grado por el cual un grafo puede subdividirse en grupos claramente diferenciados. De este modo, es posible elegir un subgrupo de estos parámetros que tenga capacidad discriminatoria con relación a enfermedades musculares concretas, como miopatías o patologías neuromusculares, construyéndose así un vector característico que permite determinar, a través de la comparación con al menos un vector característico control correspondiente o equivalente a la biopsia de un individuo sano, no sólo la presencia de dichas enfermedades musculares, sino incluso su grado de afectación al paciente. En otras palabras, de todo el grupo de parámetros propuesto se descartan aquellos que no proporcionan información relevante que pueda servir para diferenciar entre tejidos sanos y tejidos afectados, por ejemplo porque su valor es constante en todas las muestras.  - Maximized modularity: it is a statistic that quantifies the degree by which a graph can be subdivided into clearly differentiated groups. In this way, it is possible to choose a subgroup of these parameters that has discriminatory capacity in relation to specific muscular diseases, such as myopathies or neuromuscular pathologies, thus building a characteristic vector that allows determining, through comparison with at least one characteristic control vector corresponding or equivalent to the biopsy of a healthy individual, not only the presence of said muscular diseases, but even their degree of involvement in the patient. In other words, those that do not provide relevant information that can be used to differentiate between healthy tissues and affected tissues are discarded from the whole set of parameters proposed, for example because their value is constant in all samples.
En una realización preferida de la invención, la selección del subgrupo de parámetros que formará el vector característico se puede realizar utilizando los métodos SFS (Sequential Forward Selection) y SBS (Sequential Backward Selection) mediante una red neuronal Fuzzy-ARTMAP. In a preferred embodiment of the invention, the selection of the subgroup of parameters that will form the characteristic vector can be performed using the SFS (Sequential Forward Selection) and SBS (Sequential Backward Selection) methods by means of a Fuzzy-ARTMAP neural network.
Una vez seleccionado un subgrupo de los parámetros anteriores para formar un vector característico, es necesario realizar la comparación entre al menos un vector característico correspondiente a un paciente y al menos un vector característico control. Esto se puede real izar em pleando diversos métodos matemáticos y estadísticos conocidos en la técnica, aunque en una realización preferida de la invención se utiliza el Análisis de la Componente Principal (ACP, o Principal Component Analysis por sus siglas en inglés) o Redes Neuronales Artificiales (RNA). Once a subset of the above parameters has been selected to form a characteristic vector, it is necessary to make the comparison between at least one characteristic vector corresponding to a patient and at least one characteristic control vector. This can be done by using various mathematical and statistical methods known in the art, although Component Analysis is used in a preferred embodiment of the invention. Principal (ACP, or Principal Component Analysis) or Artificial Neural Networks (RNA).
El vector característico control puede obtenerse a través de la biopsia de un individuo cuyo estado es conocido, es decir, que se sabe que está sano o enfermo. Nótese además que no es necesario tener un único vector característico del paciente que se está evaluando y un único vector característico control, sino que es posible construir un grupo de vectores característicos del paciente a partir de una o varias biopsias y compararlo con un grupo de vectores característicos control correspondientes a individuos de características conocidas, preferentemente sanos.. Además, en el presente documento, el término "vector característico control" puede hacer referencia no sólo a un vector característico correspondiente a una biopsia real de un individuo, sino también a un vector característico formado por elementos considerados umbral entre valores normales y valores alterados. Por ejemplo, sería posible determinar, a través de estudios de individuos sanos y enfermos, un conjunto de valores umbral de los parámetros descritos anteriormente correspondientes a la frontera entre valores normales y valores potencialmente correspondientes a la presencia de enfermedades musculares. The characteristic control vector can be obtained through the biopsy of an individual whose condition is known, that is, who is known to be healthy or sick. Note also that it is not necessary to have a single characteristic vector of the patient being evaluated and a single characteristic control vector, but it is possible to construct a group of characteristic vectors of the patient from one or several biopsies and compare it with a group of vectors control characteristics corresponding to individuals of known, preferably healthy characteristics. Furthermore, in this document, the term "control characteristic vector" may refer not only to a characteristic vector corresponding to a real biopsy of an individual, but also to a vector characteristic formed by elements considered threshold between normal values and altered values. For example, it would be possible to determine, through studies of healthy and diseased individuals, a set of threshold values of the parameters described above corresponding to the boundary between normal values and values potentially corresponding to the presence of muscle diseases.
En otra realización preferida de la invención, la selección del subgrupo de parámetros que formará el vector característico se realiza buscando aquel subgrupo de parámetros que maximiza el valor de un descriptor ACP de tipo Calinski-Harabasz. El descriptor ACP utilizado da una ¡dea de la bondad con la cual un subgrupo determinado de parámetros consigue separar un grupo de vectores (correspondiente a la muestra del paciente) de otro grupo de vectores (correspondiente al control). Brevemente, el descriptor ACP utilizado en este documento se define como sigue: =∑∑(*¿ (0 - XtXxtV) - χ,)Ί In another preferred embodiment of the invention, the selection of the subgroup of parameters that will form the characteristic vector is performed by looking for that subgroup of parameters that maximizes the value of an ACP descriptor of the Calinski-Harabasz type. The ACP descriptor used gives an idea of the goodness with which a certain subgroup of parameters manages to separate a group of vectors (corresponding to the patient sample) from another group of vectors (corresponding to the control). Briefly, the ACP descriptor used in this document is defined as follows: = ∑∑ (* ¿(0 - XtXxtV) - χ,) Ί
ii == ll 11==11  ii == ll 11 == 11
kk  kk
B = ^ Niix, - x)(¾ - x)7
Figure imgf000014_0001
B = ^ Niix, - x) (¾ - x) 7
Figure imgf000014_0001
Ratio PCA = trace(B / ) donde dado un conjunto X = {x(l), ... , x(N)} de N objetos y una partición de dichos datos en k subgrupos inconexos, N¡ es el número de objetos asignados al subgrupo ith, x¡(Z) es el lth objeto asignado a ese subgrupo, x¡ es el vector n-dimensional de las muestras de ese sub-grupo (centroide del subgrupo) y x es el vector n-dimensional de las todas las muestras (centroide de los datos). Así, las matrices W y B se suman para obtener la matriz T de dispersión del conjunto de datos. Se puede esperar que subgrupos compactos y separados tengan valores bajos de W y valores altos de B. En consecuencia, cuanto mejor sea la partición de los datos, mayor será el valor de la relación entre B/ W que constituye el descriptor ACP. Ratio PCA = trace (B /) where given a set X = {x (l), ..., x (N)} of N objects and a partition of said data into unconnected k subgroups, N ¡ is the number of objects assigned to the subgroup i th , x ¡ (Z) is the l th object assigned to that subgroup, x ¡ is the n-dimensional vector of the samples of that sub-group (centroid of the subgroup) and x is the n-dimensional vector of the all samples (centroid of the data). Thus, the matrices W and B are added to obtain the dispersion matrix T of the data set. It can be expected that compact and separate subgroups have low values of W and high values of B. Consequently, the better the partition of the data, the greater the value of the relationship between B / W that constitutes the ACP descriptor.
En consecuencia, el método propuesto se basa fundamentalmente en probar diferentes combinaciones de parámetros hasta obtener un subgrupo de parámetros que maximice el valor de dicho descriptor ACP. Esto se puede hacer de diferentes modos, aunque en una realización preferente de la invención se implementa mediante los siguientes pasos: Consequently, the proposed method is based primarily on testing different combinations of parameters until obtaining a subset of parameters that maximizes the value of said ACP descriptor. This can be done in different ways, although in a preferred embodiment of the invention it is implemented by the following steps:
Calcular el descriptor ACP para cada combinación posible de dos parámetros.  Calculate the ACP descriptor for each possible combination of two parameters.
Añadir, a las diez mejores combinaciones de dos parámetros obtenidos en el paso anterior (es decir, aquellos diez subgrupos que presenten una mayor relación B/W), un tercer parámetro y calcular el descriptor ACP para cada combinación posible. Añadir, a las cinco mejores combinaciones de tres parámetros calculadas en el paso anterior, un cuarto parámetro y calcular el descriptor ACP para cada combinación posible. Add, to the top ten combinations of two parameters obtained in the previous step (that is, those ten subgroups that have a higher B / W ratio), a third parameter and calculate the ACP descriptor for each possible combination. Add, to the five best combinations of three parameters calculated in the previous step, a fourth parameter and calculate the ACP descriptor for each possible combination.
Añadir, a las dos mejores combinaciones de cuatro parámetros calculadas en el paso anterior, un quinto parámetro y calcular el descriptor ACP para cada combinación posible.  Add, to the two best combinations of four parameters calculated in the previous step, a fifth parameter and calculate the ACP descriptor for each possible combination.
Añadir, a la mejor combinación de cinco parámetros calculada en el paso anterior, un sexto parámetro y calcular el descriptor ACP para cada combinación posible.  Add, to the best combination of five parameters calculated in the previous step, a sixth parameter and calculate the ACP descriptor for each possible combination.
- Añadir sucesivamente, a la mejor combinación de seis parámetros calculada en el paso anterior, un parámetro adicional.  - To add successively, to the best combination of six parameters calculated in the previous step, an additional parameter.
Elegir la combinación de parámetros que presente el máximo valor del descriptor ACP. Este método se lleva a cabo sucesivamente, aunque preferentemente sólo se evalúan combinaciones de entre uno y siete parámetros, ya que un subgrupo formado por más de siete parámetros podría producir inestabilidad en los posteriores análisis de ACP. Es evidente que el método descrito puede llevarse a cabo por medio de un programa de ordenador, y por tanto la invención también se extiende igualmente a los programas de ordenador, particularmente a programas de ordenador situados en una portadora, que están adaptados para que un ordenador lleve a la práctica la invención. El programa puede tener la forma de código fuente, código objeto, una fuente intermedia de código y código objeto, por ejemplo, como en forma parcialmente compilada, o en cualquier otra forma adecuada para uso en la puesta en práctica de los procesos según la invención. La portadora puede ser cualquier entidad o dispositivo capaz de soportar el programa.  Choose the combination of parameters that has the maximum ACP descriptor value. This method is carried out successively, although preferably only combinations of one to seven parameters are evaluated, since a subgroup consisting of more than seven parameters could cause instability in subsequent ACP analyzes. It is evident that the described method can be carried out by means of a computer program, and therefore the invention also extends to computer programs, particularly to computer programs located on a carrier, which are adapted so that a computer Put the invention into practice. The program may have the form of source code, object code, an intermediate source of code and object code, for example, as in partially compiled form, or in any other form suitable for use in the implementation of the processes according to the invention . The carrier can be any entity or device capable of supporting the program.
Por ejemplo, la portadora podría incluir un medio de almacenamiento, como una memoria ROM, una memoria CD ROM, una memoria ROM de semiconductor, o un disco duro. La portadora puede ser también una portadora transmisible, por ejemplo, una señal eléctrica u óptica que podría transportarse a través de cable eléctrico u óptico, por radio o por cualesquiera otros medios. Cuando el programa va incorporado en una señal que puede ser transportada directamente por un cable u otro dispositivo o medio, la portadora puede estar constituida por dicho cable u otro dispositivo o medio. For example, the carrier could include a storage medium, such as a ROM, a CD ROM, a ROM memory of semiconductor, or a hard drive. The carrier can also be a transmissible carrier, for example, an electrical or optical signal that could be transported through electrical or optical cable, by radio or by any other means. When the program is incorporated into a signal that can be directly transported by a cable or other device or medium, the carrier may be constituted by said cable or other device or means.
Como variante, la portadora podría ser un circuito integrado en el que está incluido el programa, estando el circuito integrado adaptado para ejecutar, o para ser utilizado en la ejecución de, los procesos correspondientes. As a variant, the carrier could be an integrated circuit in which the program is included, the integrated circuit being adapted to execute, or to be used in the execution of, the corresponding processes.
BREVE DESCRIPCIÓN DE LAS FIGURAS BRIEF DESCRIPTION OF THE FIGURES
La Fig. 1 muestra un ejemplo de biopsia muscular de un paciente antes de realizar la tinción. Fig. 1 shows an example of a patient's muscle biopsy before staining.
La Fig. 2 muestra una biopsia similar a la de la Fig. 1 una vez teñida con anticuerpos contra la proteína colágeno VI y con anticuerpos contra la cadena pesada de la isoforma lenta de la miosina.  Fig. 2 shows a biopsy similar to that of Fig. 1 once stained with antibodies against collagen protein VI and with antibodies against the heavy chain of the slow myosin isoform.
La Fig. 3 muestra la biopsia de la Fig. 2 una vez terminada la segmentación.  Fig. 3 shows the biopsy of Fig. 2 after the segmentation is finished.
La Fig. 4 muestra un detalle de la biopsia de la Fig. 3 una vez formada la red.  Fig. 4 shows a detail of the biopsy of Fig. 3 once the network is formed.
La Fig. 5 es una gráfica donde cada punto corresponde al vector característico de un paciente con distrofia (círculos) o de un individuo sano (cuadrados). Cada círculo, que corresponde a un individuo con distrofia, presenta unos números que indican el grado de afectación según un neuropatólogo experto.  Fig. 5 is a graph where each point corresponds to the characteristic vector of a patient with dystrophy (circles) or of a healthy individual (squares). Each circle, which corresponds to an individual with dystrophy, has numbers that indicate the degree of involvement according to an expert neuropathologist.
REALIZACIÓN PREFERIDA DE LA INVENCIÓN PREFERRED EMBODIMENT OF THE INVENTION
Se describe a continuación con mayor detalle un ejemplo de aplicación del método de la invención haciendo referencia a las figuras adjuntas. El método se realiza en su mayor parte de forma automática por ordenador, de modo que se consigue una mayor velocidad y repetitividad de los resultados en comparación con el método habitual en que el patólogo determina visualmente la presencia o ausencia de enfermedades musculares. An example of application of the method of the invention is described in greater detail below with reference to the attached figures. He The method is mostly performed automatically by computer, so that greater speed and repetitiveness of the results is achieved compared to the usual method in which the pathologist visually determines the presence or absence of muscle diseases.
El primer paso consiste en la adquisición de biopsias de tejido muscular de un paciente, que posteriormente son procesadas mediante los métodos estándar de congelación y corte con criostato. La Fig. 1 muestra el aspecto que presenta un ejemplo de biopsia antes de proceder a su tinción. The first step is the acquisition of biopsies of a patient's muscle tissue, which are subsequently processed by standard freezing and cryostat cutting methods. Fig. 1 shows the appearance of an example of a biopsy before staining.
A continuación, las biopsias reciben una tinción ¡nmunohistoquímica con un protocolo de tinción de fluorescencia estándar. Como se ha descrito anteriormente, se realizan tinciones dobles, añadiendo un anticuerpo secundario diferente para cada anticuerpo primario. El resultado son preparaciones teñidas a la vez con dos anticuerpos, uno que detecta el endomisio y el otro que detecta las fibras de tipo I. Concretamente, en este ejemplo las biopsias se tiñen con un primer anticuerpo contra la proteína colágeno VI para identificar el endomisio, que queda teñido de color verde, y con un segundo anticuerpo específico contra la cadena pesada de la isoforma lenta de la miosina, en concreto el sMyHC, para identificar las fibras tipo I, que quedan teñidas de color rojo. Next, the biopsies receive an immunohistochemical staining with a standard fluorescence staining protocol. As described above, double stains are performed, adding a different secondary antibody for each primary antibody. The result is preparations stained at the same time with two antibodies, one that detects the endomysium and the other that detects type I fibers. Specifically, in this example the biopsies are stained with a first antibody against the collagen VI protein to identify the endomysium , which is stained green, and with a second specific antibody against the heavy chain of the slow myosin isoform, specifically the sMyHC, to identify type I fibers, which are stained red.
Una vez terminada la tinción, se adquiere una imagen de la biopsia utilizando un microscopio de fluorescencia. La imagen se obtiene de una parte de la biopsia que no tenga ningún artefacto debido a la preparación. Además, normalmente las imágenes se adquieren siempre con la misma resolución y tamaño. Once the staining is finished, an image of the biopsy is acquired using a fluorescence microscope. The image is obtained from a part of the biopsy that has no artifact due to the preparation. In addition, images are usually acquired with the same resolution and size.
Puesto que actualmente la Ley de Patentes 1 1/1986 no permite el uso de colores en las figuras , la i magen adqu irida por el m icroscopio de fluorescencia ha debido representarse en blanco y negro. En la Fig. 2, el gris corresponde a las fibras tipo I, el negro corresponde a las fibras tipo II, y las bandas casi blancas que separan unas fibras de otras corresponden al endomisio. Sin embargo se debe entender que, en este ejemplo concreto, la tinción aplicada provoca que las fibras tipo I aparezcan realmente de color rojo, las fibras tipo II de color negro y el endomisio de color verde. Since currently, Patent Law 1 1/1986 does not allow the use of colors in the figures, the image acquired by the fluorescence microscope must have been represented in black and white. In Fig. 2, gray corresponds to type I fibers, black corresponds to type II fibers, and almost white bands that separate some fibers from others correspond to the endomysium. However, it should be understood that, in this specific example, the staining applied causes the type I fibers to actually appear red, the type II fibers black and the endomysium green.
Una vez adquirida la imagen, comienza el proceso de segmentación que permitirá identificar de un modo preciso los contornos de las diferentes fibras, tanto del tipo I como del tipo II. Una vez identificados los contornos de las fibras, se considerará endomisio todo lo que no sea fibra Once the image is acquired, the segmentation process begins that will allow the precise identification of the contours of the different fibers, both type I and type II. Once the contours of the fibers have been identified, everything other than fiber will be considered endomysium.
La segmentación se realiza empleando la técnica Watershed, utilizando marcadores para conseguir una correcta segmentación de las fibras musculares con esta técnica. El objetivo de los marcadores es identificar la presencia de todas las fibras existentes en la imagen. Se lleva a cabo mediante el empleo de operadores morfológicos, que se detallan a continuación. Segmentation is performed using the Watershed technique, using markers to achieve proper segmentation of muscle fibers with this technique. The purpose of the markers is to identify the presence of all the fibers in the image. It is carried out through the use of morphological operators, which are detailed below.
En primer lugar, se binariza el plano verde, poniendo a un valor de 1 aquellos píxeles que formen regiones conexas con un valor de intensidad constante menor que un umbral y cuyos bordes estén formados por píxeles con mayor valor de intensidad. El resto de los píxeles de la imagen adquieren un valor de 0. First, the green plane is binarized, putting at a value of 1 those pixels that form related regions with a constant intensity value less than a threshold and whose edges are formed by pixels with greater intensity value. The rest of the pixels in the image acquire a value of 0.
A continuación se eliminan aquellas regiones con un área menor que un umbral. Se realiza una reconstrucción morfológica que permite rellenar huecos, entendiendo por huecos aquellas regiones de píxeles con valor 0 rodeados de regiones con valor 1 . Se continúa con una apertura. Por último se realiza una erosión, para evitar que regiones con valor 1 que identifiquen a distintas fibras estén en contacto. Con el fin de no segmentar capi lares como fibras se aplican varias condiciones de color a las regiones con valor 1 . Estas regiones tienen que cumplir que su valor medio de intensidad en el plano G (plano Verde) sea menor que un umbral, y por otra parte que su valor de intensidad en el plano R (plano Rojo) sea menor que un umbral y su valor medio del plano G cuando se ecualiza su histograma sea mayor que un umbral. Después de aplicar estos operadores y condiciones, se obtienen los marcadores, que son aquellas regiones con valor 1 resultantes en la imagen binaria, y que identifican la presencia de fibras. La transformada Watershed se aplica a estos marcadores junto al gradiente del plano verde obteniendo la segmentación precisa de los contornos de las fibras musculares. El resultado se muestra en la Fig. 3, donde los contornos de las fibras musculares aparecen en color blanco, mientras que las propias fibras y el endomisio siguen teniendo los mismos tonos descritos anteriormente con relación a la Fig. 2. Next, those regions with an area smaller than a threshold are eliminated. A morphological reconstruction is performed that allows filling gaps, meaning gaps those regions of pixels with value 0 surrounded by regions with value 1. It continues with an opening. Finally, erosion is carried out, to prevent regions with a value of 1 that identify different fibers from being in contact. In order not to segment layers as fibers, several color conditions are applied to regions with value 1. These regions have to meet that their average intensity value in the G plane (Green plane) is less than a threshold, and on the other hand that its intensity value in the R plane (Red plane) is less than a threshold and its average value of the G plane when its histogram is equalized is greater than a threshold. After applying these operators and conditions, the markers are obtained, which are those regions with value 1 resulting in the binary image, and that identify the presence of fibers. The Watershed transform is applied to these markers next to the gradient of the green plane obtaining the precise segmentation of the contours of the muscle fibers. The result is shown in Fig. 3, where the contours of the muscle fibers appear in white, while the fibers themselves and the endomysium still have the same tones described above in relation to Fig. 2.
A continuación, se determina cuáles son las fibras vecinas de cada fibra de la imagen. Para ello, se procesa la imagen segmentada de la Fig. 2 expandiendo por igual la superficie detectada de cada fibra hasta contactar con las expansiones de las fibras vecinas. Se realiza mediante la aplicación directa de la Transformada Watershed a la imagen binaria obtenida a partir de los marcadores encontrados. Next, it is determined which are the neighboring fibers of each fiber of the image. For this, the segmented image of Fig. 2 is processed by expanding equally the detected surface of each fiber until contacting the expansions of the neighboring fibers. It is done by directly applying the Watershed Transform to the binary image obtained from the found markers.
El paso siguiente consiste en la creación de una red o grafo representativo de la biopsia, donde los nodos corresponden a las diferentes fibras y las uniones conectan cada fibra con sus fibras vecinas. Esta red se representa según se muestra en la Fig. 4, donde las líneas grises de un color más claro representan los contornos de las diferentes fibras/nodos de la red, mientras que las líneas más oscuras (casi negras) representan las uniones entre unos nodos/fibras y otros. Los nodos propiamente dichos están representados por las intersecciones entre varias líneas oscuras, y están ubicados en el centro de masa de las fibras/nodos correspondientes. Aunque no se ha representado gráficamente en la Fig. 4, se entiende que la red creada comprende nodos de dos tipos, que corresponden a los dos tipos de fibras (tipo I y tipo II). The next step is the creation of a network or graph representative of the biopsy, where the nodes correspond to the different fibers and the joints connect each fiber with its neighboring fibers. This network is represented as shown in Fig. 4, where the gray lines of a lighter color represent the contours of the different fibers / nodes of the network, while the darker (almost black) lines represent the joints between some nodes / fibers and others. The nodes themselves are represented by the intersections between several dark lines, and are located in the center of mass of the corresponding fibers / nodes. Although not shown graphically in Fig. 4, it is understands that the network created comprises nodes of two types, which correspond to the two types of fibers (type I and type II).
Una vez formada la red, resulta sencillo calcular diferentes parámetros, bien de tipo geométrico o bien parámetros de red, con el objeto de caracterizar la biopsia y perm itir su comparación con determ inados umbrales obtenidos empíricamente o con los parámetros correspondientes a biopsias control. Por ejemplo, se puede determinar el área de las diferentes fibras, la longitud de sus ejes, el número de vecinos que tiene cada fibra, etc. Para hacer esto, previamente se selecciona una región de interés (ROI) que cumpla las siguientes condiciones:  Once the network is formed, it is easy to calculate different parameters, either geometric or network parameters, in order to characterize the biopsy and allow its comparison with certain thresholds obtained empirically or with the parameters corresponding to control biopsies. For example, you can determine the area of the different fibers, the length of their axes, the number of neighbors each fiber has, etc. To do this, a region of interest (ROI) that meets the following conditions is previously selected:
- Tiene siempre el mismo tamaño y forma  - It always has the same size and shape
- Excluye posibles artefactos de la preparación  - Excludes possible artifacts from the preparation
- Las fibras de los límites exteriores de la ROI deben estar rodeadas al menos por dos hileras adicionales de células.  - The fibers of the outer limits of the ROI must be surrounded by at least two additional rows of cells.
A continuación, es posible seleccionar un subconjunto de parámetros, que en este ejemplo se eligen de entre los 26 u 82 parámetros mencionados anteriormente, para construir el vector característico de la muestra. Este subconjunto se selecciona empleando, por ejemplo, los métodos SFS (Sequential Forward Selection) y S B S (Sequential Backward Selection) mediante una red neuronal Fuzzy-ARTMAP, con el propósito de conseguir un vector característico capaz de discriminar entre biopsias sanas y biopsias afectadas por alguna miopatía primaria (distrófica o no distrófica) o atrofia neurógena. Next, it is possible to select a subset of parameters, which in this example are chosen from among the 26 or 82 parameters mentioned above, to construct the characteristic vector of the sample. This subset is selected using, for example, the SFS (Sequential Forward Selection) and SBS (Sequential Backward Selection) methods through a Fuzzy-ARTMAP neural network, with the purpose of achieving a characteristic vector capable of discriminating between healthy biopsies and biopsies affected by some primary myopathy (dystrophic or non-dystrophic) or neurogenic atrophy.
La Fig. 5 muestra un ejemplo donde el vector característico seleccionado está formado por las características 19 (media de la relación de ejes), 23 (ángulos medios) y 25 (media cociente área fibra/ área media fibras vecinas) de la tabla anterior. A continuación, se ha aplicado la técnica de Análisis de Componente Principal (ACP) para representar un primer grupo de 16 biopsias correspondientes a individuos sanos (representados por cuadrados negros) y un segundo grupo de 20 biopsias correspondientes a pacientes con distrofias musculares (representados por círculos negros). Todas las biopsias pertenecen a cuádriceps de personas de entre 2 y 15 años de edad. Fig. 5 shows an example where the selected characteristic vector is formed by the characteristics 19 (average of the relation of axes), 23 (average angles) and 25 (average fiber area ratio / average area of neighboring fibers) of the previous table. Next, the Principal Component Analysis (ACP) technique has been applied to represent a first group of 16 biopsies corresponding to healthy individuals (represented by black squares) and a second group of 20 biopsies corresponding to patients with dystrophies muscle (represented by black circles). All biopsies belong to quadriceps of people between 2 and 15 years old.
El análisis ACP es un método objetivo que utiliza los valores de las características seleccionadas para realizar la comparación entre dos grupos o más de imágenes. Como resultado, se obtiene una proyección en dos o tres dimensiones que maximiza la dispersión cada una de las imágenes. Esta visualización permite la cuantificación de las diferencias entre grupos formados por cada tipo de datos. Cada imagen o punto de la Fig. 5 corresponde, por tanto, al vector característico de un paciente con distrofia o de un individuo sano. Se aprecia cómo ambos grupos aparecen separados en la Fig. 5, quedando los cuadrados correspondientes a individuos sanos principalmente a la izquierda de la gráfica y los círculos correspondientes a individuos con distrofias principalmente a la derecha de la gráfica. ACP analysis is an objective method that uses the values of the selected characteristics to make the comparison between two or more groups of images. As a result, a projection in two or three dimensions is obtained that maximizes the dispersion of each of the images. This visualization allows the quantification of the differences between groups formed by each type of data. Each image or point of Fig. 5 therefore corresponds to the characteristic vector of a patient with dystrophy or of a healthy individual. It can be seen how both groups appear separated in Fig. 5, with the squares corresponding to healthy individuals mainly to the left of the graph and the circles corresponding to individuals with dystrophies mainly to the right of the graph.
Además, se descubre que la distancia entre los vectores correspondientes a pacientes afectados de distrofia (círculos) y el centro de masas de los vectores correspondiente a individuos sanos (cuadrados) se correlaciona con el grado de afectación al paciente. Para comprobarlo, un neuropatólogo experto realizó, sin conocer el resultado del análisis anterior, evaluaciones del grado de afectación de cada biopsia de los pacientes con distrofia. Se asignó a cada biopsia una gradación de entre 1 -4, correspondiendo el grado 1 a una biopsia poco afectada por la distrofia y el grado 4 a una biopsia muy afectada por la distrofia. In addition, it is discovered that the distance between the vectors corresponding to patients affected by dystrophy (circles) and the center of mass of the vectors corresponding to healthy individuals (squares) correlates with the degree of patient involvement. To verify this, an expert neuropathologist performed, without knowing the result of the previous analysis, evaluations of the degree of involvement of each biopsy of patients with dystrophy. A gradation of between 1 -4 was assigned to each biopsy, grade 1 corresponding to a biopsy poorly affected by dystrophy and grade 4 to a biopsy severely affected by dystrophy.
En la Fig. 5 se puede ver que en la mayoría de los casos la evaluación del grado de afectación realizada por el patólogo, que se muestra como un número junto a cada círculo correspondiente a pacientes con distrofias musculares, se correlaciona directamente con la distancia entre dicho punto y el centro de masas del grupo de cuadrados correspondientes a pacientes control. Es decir, básicamente cuanto más a la derecha se encuentra el círculo correspondiente al vector característico, mayor es el grado de afectación de la distrofia correspondiente a esa biopsia. Se demuestra así que el método de análisis de imagen aporta datos útiles para el diagnóstico de enfermedades musculares de una forma rápida, automática y objetiva. Para una entrada (una biopsia) nueva a este sistema, también es posible estimar la clase a la que ésta pertenece empleando el método KNN (K vecinos más cercanos, o nearest neighbours según sus siglas en inglés). En ese caso, se calcula la distancia entre los vectores característicos ya almacenados y el nuevo vector característico de la nueva biopsia, y se seleccionan los k ejemplos más cercanos. La nueva biopsia es clasificada con la clase que más se repite en los k vectores seleccionados. In Fig. 5 it can be seen that in most cases the evaluation of the degree of affectation carried out by the pathologist, which is shown as a number next to each circle corresponding to patients with muscular dystrophies, correlates directly with the distance between said point and the center of mass of the group of squares corresponding to control patients. That is, basically the further to the right is the circle corresponding to the characteristic vector, the greater the degree of affectation of the dystrophy corresponding to that biopsy. This demonstrates that the image analysis method provides useful data for the diagnosis of muscular diseases in a fast, automatic and objective way. For a new entry (a biopsy) to this system, it is also possible to estimate the class to which it belongs using the KNN method (K nearest neighbors, or nearest neighbors according to its acronym in English). In that case, the distance between the characteristic vectors already stored and the new characteristic vector of the new biopsy is calculated, and the nearest k examples are selected. The new biopsy is classified with the most repeated class in the selected vectors.
REFERENCIAS REFERENCES
• Dubowitz, V., and Sewry, C.A. (2006). Muscle Biopsy: A Practical Approach (Eselvier). • Dubowitz, V., and Sewry, C.A. (2006). Muscle Biopsy: A Practical Approach (Eselvier).
• Escudero, L.M. , da F. Costa, L, Kicheva, A. , Briscoe, J., Freeman, M. , and Babu, M.M. (201 1 ). Epithelial organisation revealed by a network of cellular contacts. Nature Communications. In press.  • Escudero, L.M. , da F. Costa, L, Kicheva, A., Briscoe, J., Freeman, M., and Babu, M.M. (201 1). Epithelial organization revealed by a network of cellular contacts. Nature Communications In press
• Helliwell, T.R. (1999). Muscle: Part 1 - Normal structure and function.  • Helliwell, T.R. (1999). Muscle: Part 1 - Normal structure and function.
Current Orthopaedics 13, 33 41 .  Current Orthopedics 13, 33 41.
• Mitiche, A., and Ben Ayed, I. (2010). Variational and Level Set Methods in Image Segmentaron. (Springer).  • Mitiche, A., and Ben Ayed, I. (2010). Variational and Level Set Methods in Image Segmented. (Springer)
• O'Ferrall, E.K., and Sinnreich, M. (2009). The role of muscle biopsy in the age of genetic testing. Curr Opin Neurol 22, 543-553.  • O'Ferrall, E.K., and Sinnreich, M. (2009). The role of muscle biopsy in the age of genetic testing. Curr Opin Neurol 22, 543-553.
• Pette, D., and Staron, R.S. (2000). Myosin isoforms, muscle fiber types, and transitions. Microsc Res Tech 50, 500-509.  • Pette, D., and Staron, R.S. (2000). Myosin isoforms, muscle fiber types, and transitions. Microsc Res Tech 50, 500-509.
• Calinsk¡& Harabasz(1974) Communications in Statistics 3: 1 -27.  • Calinsk¡ & Harabasz (1974) Communications in Statistics 3: 1-27.
• Sánchez-Gutiérrez, D., Sáez, A., Pascual, A., Escudero, L.M.* The emergence of significant differences in proliferating epithelia. In preparation. • Sánchez-Gutiérrez, D., Sáez, A., Pascual, A., Escudero, LM * The emergence of significant differences in proliferating epithelia. In preparation

Claims

REIVINDICACIONES
1 . Método para obtener información útil para el diagnóstico de enfermedades neuromusculares a partir de una biopsia de tejido de músculo esquelético, caracterizado porque comprende los siguientes pasos: one . Method for obtaining useful information for the diagnosis of neuromuscular diseases from a biopsy of skeletal muscle tissue, characterized in that it comprises the following steps:
- realizar una tinción de la biopsia para resaltar las fibras musculares tipo I, las fibras musculares tipo II y el endomisio;  - perform a biopsy stain to highlight type I muscle fibers, type II muscle fibers and the endomysium;
- obtener una imagen de la biopsia tras la tinción; - obtain an image of the biopsy after staining;
- segmentar la imagen para identificar los contornos de las fibras musculares; - segment the image to identify the contours of muscle fibers;
- formar una red donde las fibras musculares constituyen nodos y los contactos entre fibras musculares constituyen las uniones entre los nodos; y - forming a network where the muscle fibers constitute nodes and the contacts between muscle fibers constitute the junctions between the nodes; Y
- formar un vector característico de la biopsia cuyos elementos se eligen entre parámetros geométricos de las fibras y parámetros de red. - form a characteristic vector of the biopsy whose elements are chosen between geometric parameters of the fibers and network parameters.
2. Método de acuerdo con la reivindicación 1 , donde el paso de tinción de la muestra comprende aplicar preparaciones de dos anticuerpos que tiñen de colores diferentes dos elementos de entre: las fibras musculares tipo I, las fibras musculares tipo II y el endomisio. 2. Method according to claim 1, wherein the staining step of the sample comprises applying preparations of two antibodies that stain two different colors between two elements: type I muscle fibers, type II muscle fibers and the endomysium.
3. Método de acuerdo con la reivindicación 2, donde un primer anticuerpo actúa contra la proteína colágeno VI para teñir el endomisio de un primer color y un segundo anticuerpo actúa contra la cadena pesada de la isoforma lenta de la miosina para teñir las fibras musculares tipo I de un segundo color. 3. Method according to claim 2, wherein a first antibody acts against the collagen VI protein to stain the endomysium of a first color and a second antibody acts against the heavy chain of the slow myosin isoform to stain the type muscle fibers I of a second color.
4. Método de acuerdo con la reivindicación 3, donde el anticuerpo contra la cadena pesada de la isoforma lenta de la miosina es sMyHC. 4. Method according to claim 3, wherein the antibody against the myosin slow isoform heavy chain is sMyHC.
5. Método de acuerdo con cualquiera de las reivind icaciones anteriores, donde el paso de obtención de una imagen de la biopsia teñida se lleva a cabo utilizando un microscopio de fluorescencia en una zona de la biopsia carente de artefactos causados durante la preparación de la misma. 5. Method according to any of the preceding claims, wherein the step of obtaining a stained biopsy image is carried out using a fluorescence microscope in an area of the biopsy devoid of artifacts caused during the preparation of the same.
6. Método de acuerdo con cualquiera de las reivindicaciones anteriores, donde el paso de segmentación com prende el cálculo de marcadores morfológicos para fibras musculares y la aplicación de la Transformada Watershed. 6. Method according to any of the preceding claims, wherein the segmentation step comprises the calculation of morphological markers for muscle fibers and the application of the Watershed Transform.
7. Método de acuerdo con la reivindicación 6, donde el paso de formación de la red comprende identificar los vecinos de las fibras musculares mediante la expansión del contorno de cada fibra aplicando directamente la Transformada Watershed a los marcadores morfológicos para fibras musculares. 7. Method according to claim 6, wherein the network formation step comprises identifying the neighbors of the muscle fibers by expanding the contour of each fiber by directly applying the Watershed Transform to the morphological markers for muscle fibers.
8. Método de acuerdo con cualquiera de las reivind icaciones anteriores, donde el paso de formación de la red comprende además representar gráficamente la biopsia como un grafo formado por nodos y uniones entre nodos, estando los nodos ubicados en el centro de masa de cada fibra. 8. Method according to any of the preceding claims, wherein the network formation step further comprises graphically representing the biopsy as a graph formed by nodes and junctions between nodes, the nodes being located at the center of mass of each fiber .
9. Método de acuerdo con cualquiera de las reivindicaciones anteriores, donde los parámetros del vector característico se eligen del siguiente grupo: 9. Method according to any of the preceding claims, wherein the parameters of the characteristic vector are chosen from the following group:
1 Area media de las fibras 1 Average fiber area
2 Desviación típica del área de las fibras  2 Standard deviation of the fiber area
3 Número medio de vecinos de cada fibra  3 Average number of neighbors of each fiber
4 Desviación típica del número de vecinos de cada fibra  4 Standard deviation of the number of neighbors of each fiber
5 Area media de las fibras tipo I  5 Average area of type I fibers
6 Desviación típica del área de las fibras tipo I  6 Standard deviation of the type I fiber area
7 Area media de las fibras tipo II  7 Average area of type II fibers
8 Desviación típica del área de las fibras tipo II 9 Desviación típica del número de vecinos de las fibras tipo I 8 Standard deviation of type II fiber area 9 Standard deviation of the number of neighbors of type I fibers
10 Desviación típica del número de vecinos de las fibras tipo II 10 Standard deviation of the number of neighbors of type II fibers
1 1 Número medio de vecinos tipo I de las fibras tipo I 1 1 Average number of type I neighbors of type I fibers
12 Número medio de vecinos tipo II de las fibras tipo I  12 Average number of type II neighbors of type I fibers
13 Número medio de vecinos tipo I de las fibras tipo II  13 Average number of type I neighbors of type II fibers
14 Número medio de vecinos tipo II de las fibras tipo II  14 Average number of type II neighbors of type II fibers
15 Media del cociente A1/A2  15 Mean ratio A1 / A2
16 Desviación típica del cociente A1/A2  16 Standard deviation of the ratio A1 / A2
17 Dimensión media del eje mayor de las fibras  17 Average dimension of the major axis of the fibers
18 Dimensión media del eje menor de las fibras  18 Average dimension of the minor axis of the fibers
19 Media de la relación de ejes  19 Mean of axes ratio
20 Desviación típica de la relación de ejes  20 Standard deviation of the axis ratio
21 Media de la envoltura convexa  21 Convex sheath stocking
22 Desviación típica de la envoltura convexa  22 Standard deviation of the convex envelope
23 Angulos medios  23 Middle angles
24 Desviación típica de los ángulos  24 Standard deviation of angles
25 Media cociente área fibra/ área media fibras vecinas  25 Half ratio fiber area / average area neighboring fibers
26 Desviación típica cociente área fibra/ área media fibras vecinas  26 Standard deviation fiber area ratio / average area neighboring fibers
10. Método de acuerdo con la reivindicación 9, donde dicho c comprende además los siguientes parámetros: 10. Method according to claim 9, wherein said c further comprises the following parameters:
27 Media de la relación con los vecinos del eje mayor 27 Average of the relationship with the neighbors of the major axis
28 Desv. típica de la relación con los vecinos del eje mayor  28 Def. typical of the relationship with the neighbors of the major axis
29 Media de la relación con los vecinos del eje menor  29 Average of the relationship with the neighbors of the minor axis
30 Desv. típica de la relación con los vecinos del eje menor  30 Def. typical of the relationship with the minor axis neighbors
31 Media de la relación con los vecinos de la relación entre ejes 31 Average of the relationship with the neighbors of the relationship between axes
32 Des. típica de la relación con los vecinos de la relación entre ejes32 Off typical of the relationship with the neighbors of the relationship between axes
33 Media de la relación con los vecinos de la envoltura convexa33 Average of the relationship with the neighbors of the convex envelope
34 Des. típica de la relación con los vecinos de la envoltura convexa34 Off typical of the relationship with the neighbors of the convex envelope
35 Media de la relación con los vecinos de ángulos medios Desv. típica de la relación con los vecinos ángulos medios35 Average of the relationship with the neighbors of average angles Dev. typical of the relationship with the neighbors average angles
Media de la relación con los vecinos de la relación A1/A2Average of the relationship with the neighbors of the A1 / A2 relationship
Desv. típica de la relación con los vecinos de la relación A1/A2Dev. typical of the relationship with the neighbors of the A1 / A2 relationship
Media de la Suma de los pesos Average of the sum of the weights
Desv. típica de la Suma de los pesos  Dev. typical of the sum of the weights
Media de los pesos de las fibras tipo I  Average weights of type I fibers
Desv. típica de los pesos de las fibras tipo I  Dev. typical of type I fiber weights
Media de los pesos de las fibras tipo II  Average weight of type II fibers
Desv. típica de los pesos de las fibras tipo II  Dev. typical of type II fiber weights
Media del "clustering coefficient"  Clustering coefficient average
Desv. típica del "clustering coefficient"  Dev. typical of "clustering coefficient"
Media del "clustering coefficient" fibras tipo II  Average of the "clustering coefficient" type II fibers
Desv. típica del "clustering coefficient" fibras tipo II  Dev. typical of "clustering coefficient" type II fibers
Media del "clustering coefficient" fibras tipo II  Average of the "clustering coefficient" type II fibers
Desv. típica del "clustering coefficient" tipo II  Dev. typical of "clustering coefficient" type II
Media de la excentricidad  Eccentricity mean
Desv. típica de la excentricidad  Dev. typical of eccentricity
Media de la excentricidad de las fibras tipo II  Average eccentricity of type II fibers
Desv. típica de la excentricidad de las fibras tipo II  Dev. typical of the eccentricity of type II fibers
Media de la excentricidad de las fibras tipo II  Average eccentricity of type II fibers
Desv. Típica de la excentricidad de las fibras tipo II  Dev. Typical of the eccentricity of type II fibers
Media de la "Betweeness Centrality"  Average of "Betweeness Centrality"
Desv. típica de la "Betweeness Centrality"  Dev. typical of the "Betweeness Centrality"
Media de la "Betweeness Centrality" de las fibras tipo II Average of the "Betweeness Centrality" of type II fibers
Desv. típica de la "Betweeness Centrality" de las fibras tipo IIDev. typical of the "Betweeness Centrality" of type II fibers
Media de la "Betweeness Centrality" de las fibras tipo IIAverage of the "Betweeness Centrality" of type II fibers
Desv. típica de la "Betweeness Centrality" de las fibras tipo IIDev. typical of the "Betweeness Centrality" of type II fibers
Media de la distancia Average distance
Desv. típica de la distancia  Dev. typical of distance
Media de la distancia de las fibras tipo I l_f i bras tipo II  Mean distance of fibers type I l_f and bras type II
Desv. típica de la distancia de las fibras tipo ll_fibras tipo II 67 Media de la distancia de las fibras tipo II a las fibras tipo IIDev. typical of the distance of fibers type ll_fibres type II 67 Average distance of type II fibers to type II fibers
68 Desv. típica de la distancia de las fibras tipo II a las fibras tipo II 68 Dev. typical of the distance of type II fibers to type II fibers
69 Media de la distancia de las fibras fibras tipo II a las fibras tipo II 69 Average distance of fiber type II fibers to type II fibers
70 Desv. típica de la distancia de las fibras tipo II a las fibras tipo II 70 Def. typical of the distance of type II fibers to type II fibers
71 Media de la distancia de las fibras tipo I a las fibras tipo I  71 Average distance of type I fibers to type I fibers
72 Desv. típica de la distancia de las fibras tipo II a las fibras tipo I  72 Def. typical of the distance of type II fibers to type I fibers
73 Radio  73 Radio
74 Diámetro  74 Diameter
75 Eficiencia  75 Efficiency
76 Coeficiente de Pearson  76 Pearson coefficient
77 Conectividad algebraica  77 Algebraic Connectivity
78 Metrica_s  78 Metric_s
79 "Assortattivity"  79 "Assortattivity"
80 Densidad de conexiones  80 Density of connections
81 "Transitivity"  81 "Transitivity"
82 Modularidad maximizada  82 Maximized Modularity
1 1 . Método de acuerdo con cualquiera de las reivindicaciones anteriores, que además comprende comparar al menos un vector característico con al menos un vector característico control correspondiente o equivalente a una biopsia de tejido de un individuo de características conocidas. eleven . Method according to any of the preceding claims, which further comprises comparing at least one characteristic vector with at least one corresponding control vector equivalent or equivalent to a tissue biopsy of an individual of known characteristics.
12. Método de acuerdo con cualquiera de las reivindicaciones anteriores, donde los parámetros del vector característico se eligen utilizando los métodos SFS y SBS mediante una red neuronal Fuzzy-ARTMAP. 12. Method according to any of the preceding claims, wherein the parameters of the characteristic vector are chosen using the SFS and SBS methods via a Fuzzy-ARTMAP neural network.
13. Método de acuerdo con la reivindicación 12, donde la comparación del al menos un vector característico con el al menos un vector característico control se realiza utilizando la técnica de Análisis de la Componente Principal (ACP). 13. Method according to claim 12, wherein the comparison of the at least one characteristic vector with the at least one characteristic control vector is performed using the Principal Component Analysis (ACP) technique.
14. Método de acuerdo con cualquiera de las reivindicaciones 1 -1 1 , donde los parámetros del vector característico se eligen buscando aquel subgrupo de parámetros que maximiza el valor de un descriptor ACP de tipo Calinski-Harabasz. 14. Method according to any of claims 1 -1 1, wherein the parameters of the characteristic vector are chosen by looking for that subset of parameters that maximizes the value of an ACP descriptor of the Calinski-Harabasz type.
15. Método de acuerdo con la reivindicación 14, donde el descriptor ACP se define según la ecuación: 15. Method according to claim 14, wherein the ACP descriptor is defined according to the equation:
Ratio PCA = trace(B / )  PCA ratio = trace (B /)
donde: where:
B =
Figure imgf000029_0001
(0 - XiXxtV) - xJT B =
Figure imgf000029_0001
(0 - XiXxtV) - xJ T
16. Método de acuerdo con cualquiera de las reivindicaciones 14-15, donde la selección del subgrupo de parámetros que componen el vector característico comprende los siguientes pasos: 16. Method according to any of claims 14-15, wherein the selection of the subset of parameters that make up the characteristic vector comprises the following steps:
- calcular el descriptor ACP para cada combinación de dos parámetros; - calculate the ACP descriptor for each combination of two parameters;
- añadir, a las diez mejores combinaciones de dos parámetros calculadas en el paso anterior, un tercer parámetro y calcular el descriptor ACP para cada combinación posible; - add, to the top ten combinations of two parameters calculated in the previous step, a third parameter and calculate the ACP descriptor for each possible combination;
- añadir, a las cinco mejores combinaciones de tres parámetros calculadas en el paso anterior, un cuarto parámetro y calcular el descriptor ACP para cada combinación posible;  - add, to the five best combinations of three parameters calculated in the previous step, a fourth parameter and calculate the ACP descriptor for each possible combination;
- añadir, a las dos mejores combinaciones de cuatro parámetros calculadas en el paso anterior, un quinto parámetro y calcular el descriptor ACP para cada combinación posible;  - add, to the two best combinations of four parameters calculated in the previous step, a fifth parameter and calculate the ACP descriptor for each possible combination;
- añadir, a la mejor combinación de cinco parámetros calculada en el paso anterior, un sexto parámetro y calcular el descriptor ACP para cada combinación posible;  - add, to the best combination of five parameters calculated in the previous step, a sixth parameter and calculate the ACP descriptor for each possible combination;
- añadir sucesivamente, a la mejor combinación de seis parámetros calculada en el paso anterior, un parámetro adicional; y  - successively add, to the best combination of six parameters calculated in the previous step, an additional parameter; Y
- elegir la combinación de parámetros que presente el máximo valor del descriptor ACP. - choose the combination of parameters that has the maximum ACP descriptor value.
17. Método de acuerdo con la reivindicación 16, que comprende evaluar combinaciones de entre uno y siete parámetros. 17. Method according to claim 16, which comprises evaluating combinations of one to seven parameters.
18. Método de acuerdo con cualquiera de las reivindicaciones anteriores, donde un vector característico formado por los parámetros 1 918. Method according to any of the preceding claims, wherein a characteristic vector formed by parameters 1 9
(media de la relación de ejes), 23 (ángulos medios) y 25 (media cociente área fibra/área media fibras vecinas) permite determinar la presencia y grado de afectación de distrofia muscular. (average of the relation of axes), 23 (average angles) and 25 (average fiber area ratio / average area of neighboring fibers) allows to determine the presence and degree of involvement of muscular dystrophy.
19. P rog ram a de ordenador q ue com prende i nstrucciones de programa para hacer que un ordenador lleve a la práctica el método según cualquiera de las reivindicaciones 1 a 18. 19. A computer program that includes program constructions to make a computer implement the method according to any one of claims 1 to 18.
20. Programa de ordenador según la reivindicación 19, incorporado en medios de almacenamiento. 20. Computer program according to claim 19, incorporated into storage media.
21 . Programa de ordenador según la reivindicación 19, soportado en una señal portadora. twenty-one . Computer program according to claim 19, supported on a carrier signal.
PCT/ES2012/070796 2011-11-15 2012-11-15 Method for obtaining useful information for the diagnosis of neuromuscular diseases WO2013072543A1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/ES2012/070796 WO2013072543A1 (en) 2011-11-15 2012-11-15 Method for obtaining useful information for the diagnosis of neuromuscular diseases

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
ESP201131840 2011-11-15
PCT/ES2012/070796 WO2013072543A1 (en) 2011-11-15 2012-11-15 Method for obtaining useful information for the diagnosis of neuromuscular diseases

Publications (2)

Publication Number Publication Date
WO2013072543A1 true WO2013072543A1 (en) 2013-05-23
WO2013072543A8 WO2013072543A8 (en) 2015-03-12

Family

ID=52629307

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/ES2012/070796 WO2013072543A1 (en) 2011-11-15 2012-11-15 Method for obtaining useful information for the diagnosis of neuromuscular diseases

Country Status (1)

Country Link
WO (1) WO2013072543A1 (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11436452B2 (en) * 2017-08-24 2022-09-06 Toyota Motor Europe System and method for label augmentation in video data

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
BARRERA, J. ET AL.: "Morphological operators characterized by neighborhood graphs. Computer Graphics and Image Processing, 1997.", PROCEEDINGS., X BRAZILIAN SYMPOSIUM ON CAMPOS DO JORDAO, 14 October 1997 (1997-10-14), BRAZIL, pages 179 - 186 *
ESCUDERO LUIS M ET AL.: "Epithelial organisation revealed by a network of cellular contacts.", NATURE COMMUNICATIONS, vol. 2, 8 November 2011 (2011-11-08), pages 1 - 7 *
GARTON, F. ET AL.: "Validation of an automated computational method for skeletal muscle fibre morphometry analysis.", NEUROMUSCULAR DISORDERS, vol. 20, no. 8, 1 August 2010 (2010-08-01), pages 540 - 547 *
SERTEL, O. ET AL.: "Microscopic image analysis for quantitative characterization of muscle fiber type composition.", COMPUTERIZED MEDICAL IMAGING AND GRAPHICS, vol. 35, no. 7, 27 January 2011 (2011-01-27), pages 616 - 628 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11436452B2 (en) * 2017-08-24 2022-09-06 Toyota Motor Europe System and method for label augmentation in video data

Also Published As

Publication number Publication date
WO2013072543A8 (en) 2015-03-12

Similar Documents

Publication Publication Date Title
Sigal et al. Mapping synaptic input fields of neurons with super-resolution imaging
Santella et al. A hybrid blob-slice model for accurate and efficient detection of fluorescence labeled nuclei in 3D
ES2374443T3 (en) AUTOMATED SEGMENTATION OF IMAGE STRUCTURES.
Colonna et al. Segmentation of corneal nerves using a U-Net-based convolutional neural network
Zhang et al. Label-free colorectal cancer screening using deep learning and spatial light interference microscopy (SLIM)
Mannam et al. Machine learning for faster and smarter fluorescence lifetime imaging microscopy
US11501446B2 (en) Segmenting 3D intracellular structures in microscopy images using an iterative deep learning workflow that incorporates human contributions
ES2804770T3 (en) Vessel analysis in multiplexed images
CN116188423B (en) Super-pixel sparse and unmixed detection method based on pathological section hyperspectral image
JP6756339B2 (en) Image processing device and image processing method
ES2754075T3 (en) Machine vision-based method to extract features related to the development stage of trichuris spp eggs
JP2016154810A (en) Image processor and image processing method
Fishman et al. Practical segmentation of nuclei in brightfield cell images with neural networks trained on fluorescently labelled samples
Mathew et al. Deep learning‐based automated mitosis detection in histopathology images for breast cancer grading
Herold et al. Automated detection and quantification of fluorescently labeled synapses in murine brain tissue sections for high throughput applications
WO2013072543A1 (en) Method for obtaining useful information for the diagnosis of neuromuscular diseases
Rathore et al. A novel approach for colon biopsy image segmentation
ES2408137B1 (en) METHOD FOR OBTAINING USEFUL INFORMATION FOR THE DIAGNOSIS OF NEUROMUSCULAR DISEASES
Pho et al. Segmentation-driven hierarchical retinanet for detecting protozoa in micrograph
Ishaq et al. Automated quantification of zebrafish tail deformation for high-throughput drug screening
Al Mouiee et al. Classifying retinal degeneration in histological sections using deep learning
Shimoni et al. Normalized polarization ratios for the analysis of cell polarity
EP3591574A2 (en) Method and computer program for clustering large multiplexed spatially resolved data of a biological sample
JARA‐WILDE et al. Optimising adjacent membrane segmentation and parameterisation in multicellular aggregates by piecewise active contours
Tang et al. Fast post-processing pipeline for optical projection tomography

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 12849613

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

WPC Withdrawal of priority claims after completion of the technical preparations for international publication

Ref document number: P201131840

Country of ref document: ES

Date of ref document: 20140514

Free format text: WITHDRAWN AFTER TECHNICAL PREPARATION FINISHED

122 Ep: pct application non-entry in european phase

Ref document number: 12849613

Country of ref document: EP

Kind code of ref document: A1