BR102012031953A2 - method for determining the viability and quality of embryos - Google Patents

Abstract

method for determining the viability and quality of embryos. The invention is described for a method for determining embryo viability and quality that allows rapid evaluation with minimal interference with embryonic development using a microscopy system associated with digital image capture, providing a set of values for each embryo it represents. how much the embryo can be considered in each of the four possible degrees, using the technique of artificial neural networks, providing objectivity and reproducibility.

Description

FIELD OF THE INVENTION The present invention describes a method for determining the viability and quality of embryos. More specifically, it comprises a method for determining the quality of embryos by morphologically analyzing a two-dimensional figure using the artificial neural network (RNA) technique, providing an objective and reproducible quality index.

BACKGROUND OF THE INVENTION

Since the development of the first successful techniques of artificial fertilization and embryo transfer in mammalian embryos, a direct relationship of embryonic quality with the success rate of embryo transfer in recipient females has become visible, and morphologically classified embryos of quality. have a higher success rate (HR Tervit, MW Cooper, Pamela G. Goold, GM Haszard, Non-surgicalembryotransfer in cattle, Theriogenology, Volume 13, Issue 1, January 1980, Pages 63-71, ISSN 0093-691X, 10.1016 (0093-691X (80) 90015-1) and (HJ Schneider Jr., Castleberry RS, JL Griffin, Commercial Aspects of Bovine Embryo Transfer, Theriogenology, Volume 13, Issue 1, January 1980, Pages 73-85, ISSN 0093- 691X, 101016 / 0093-691X (80) 90016-3). Embryonic morphological classification is of great importance for many laboratory techniques, from basic research to applied in assisted reproduction. It is from this that success rates (pregnancy), the use of associated biotechniques (cryopreservation, biopsy, bipartition, microinjection, etc.) or even the standardization of embryos used in scientific experiments can be inferred.

Therefore, methods of standardization of the elements that categorize the embryos in different degrees according to their quality (and therefore also viability) have been developed, and the four-degree system is currently widely used: Excellent, Good ( Good), Fair and Poor (Gary M. Lindner, Raymond W. Wright Jr., Bovineembryomorphologyandevaluation, Theriogenology, Volume 20, Issue4, October 1983, Pages 407-416, ISSN 0093-691X, 10.1016 / 0093-691 X (83) 90201 -7) and (RW Wright Jr., J. Ellington, Morphological and physiological differences between in vivo- and in vitro-produced preemplantation from live stock species, Theriogenology, Volume 44, Issue 8, December 1995, Pages 1167-1189, ISSN 0093-691X, 10.1016 / 0093-691X (95) 00327-5). This system is based on the visual morphological qualitative analysis of the embryo, commonly performed by optical microscopy (stereomicroscope). The technique depends on the embryologist's experience and accuracy in analyzing and categorizing from the obvious factors to the nuances that make an embryo more or less fit for development. In this classical morphological analysis, such factors are not objectively measured, so that they become subjective and of low repeatability (BalázsBényei, IstvánKomlósi, Anna Pécsi, GeoffryPollott, Cruvinel Heraldo Marcos, Maida Paula Lemes, The effect of internai and external factors on bovine embryo transfer results in a tropical environment, Animal Reproduction Science, Volume 93, Issues 3-4, July 2006, Pages 268-279, ISSN 0378-4320, 10.1016 / j.anireprosci.2005.07.012).

As a result, the same embryo measured by different experts can obtain different degrees of classification. Such disagreement occurs mainly between close grades, such as between good and excellent embryos (PW Farin, JH Britt, DW Shaw, BD Slenning, Theriogenology, Volume 44, Issue3, August 1995, Pages 339-349, ISSN 0093-691X, 10.1016 / 0093-691X (95) 00189-F).

Seeking solutions to the subjectivity issue of morphological analysis, several alternative methods were developed. These include in vitro embryo culture, blastomeric membrane integrity, embryonic metabolism analysis (Eric W. Overström, In vitro assessment of embryo viability, Theriogenology, Volume 45, Issue 1, 1 January 1996, Pages 3-16 , ISSN 0093-691X, 10.1016 / 0093-691 X (96) 84625-5.), Measurement of cellular respiration (HiroyoshiHoshi, In vitro production of bovine embryos and their application for embryo transfer, Theriogenology, Volume 59, Issue 2, 15 January 2003, Pages 675-685, ISSN 0093-691X, 10.1016 / S0093-691X (02) 01247-5), Electron Microscopic Analysis (López-Damián E. P, Galina CS, Merchant H., Cedillo-Peláez C. , Aspron M., Assessment of Bostaurus embryos comparing stereoscopic microscopy and transmission electron microscopy Journal of Cell and Animal Biology Vol. 2 (3), pp. 072-078, March 2008, ISSN 1996-0867 © 2008 AcademicJournals) and use of birefringence indices of the zona pellucida (Held, Eva, Mertens, Eva-Maria, Mohammadi-S angcheshmeh, Abdollah, Salilew-Wondim, Dessie, Besenfelder, Urban, Havlicek, Vitezslav, Herrler, Andreas, Tesfaye, Dawit, Schellander, Karl, and Holker, Michael (2011). Zonapellucida birefringence correlates with developmental capacity of bovine oocytes classified by maturational environment, COC morphology and G6PDH activity. Playback Fertile. Dev. 24, 568-579).

However, no method has so far presented a definitive solution for measuring quality and viability, and it is still necessary to develop methods that are fast, non-invasive and objective.

Another factor to be analyzed is the high cost of some methods, which prevents their wide use. Thus, despite subjectivity and low repeatability, visual morphological analysis is still widely used to determine embryonic quality.

Therefore, the present invention describes a method for determining the viability and quality of embryos that allows rapid evaluation and minimal interference with embryonic development, as a microscopy system associated with digital image capture is required, providing a set of values for each embryo. embryo that represents how well the embryo can be considered in each of the four possible degrees, using the Artificial Neural Networks technique, providing objectivity and reproducibility.

It is characteristic of the invention a method for determining embryo viability and quality from the morphological analysis of a two-dimensional figure using the Artificial Neural Networks (RNA) technique. A method for determining embryo viability and quality that provides an objective and reproducible embryonic quality index is a feature of the invention. A method for determining the viability and quality of embryos with potential application in scientific research involving mouse / rat embryos, in the classification of quality of bovine embryos produced for commercial use (in vitro production), as well as for the evaluation of the invention is characteristic of the invention. of embryonic quality of human embryos in assisted reproduction clinics. A method for determining the viability and quality of embryos based on the objectivity of the analysis is independent of the microscope quality, the embryologist's experience, and the restricted capacity of the embryonic quality grades used for evaluation. It is a feature of the invention that a method for determining embryo viability and quality that is non-invasive, non-invasive and can be performed in a short time, allowing the evaluator to "interpret" the response, so that individual clinical experience can maximize the result obtained by the method.

BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 shows the representation of the measurement marking of the first embryo diameter (DE1). Figure 2 shows the representation of the measurement marking of the second diameter of the embryo (DE2). Figure 3 shows the representation of the marking of the first zone of the zona pellucida (DZP1). Figure 4 shows the representation of the marking of the second diameter of the zona pellucida (DZP2). Figure 5 shows the representation of the embryo area marking (AE). Figure 6 shows the representation of the zona pellucida area (AZP) marking. Figure 7 shows the representation of dead cell area (ACM) marking. Figure 8 shows the Pellucid Zone Color Density (DCZP) marking. Figure 9 shows the flowchart of the method used to evaluate RNA architecture.

DETAILED DESCRIPTION OF THE INVENTION The method for determining the viability and quality of embryos, object of the present invention, allows the objective morphological evaluation of embryos by extracting information from embryo images and further analyzing them by means of a computer program. computer based on Artificial Neural Networks (RNA), an Artificial Intelligence technique indicated for solving nonlinear problems with interconnected variables (Guoqiang Zhang, B. EddyPatuwo, Michael Y. Hu, Forecastingwith artificial neural networks: The State of the Art , International Journal of Forecasting, Volume 14, Issue 1, March 1, 1998, Pages 35-62, ISSN 0169-2070, 10.1016 / S0169-2070 (97) 00044-7) and (Eidon Y. Li, Artificial neural networks and their business applications, Information & Management, Volume 27, Issue 5, November 1994, Pages 303-313, ISSN 0378-7206, 10.1016 / 0378-7206 (94) 90024-8).

Basically, an Artificial Neural Network is a system that acts in solving problems, simulating the functioning of a set of biological neurons. Such RNA neurons (also called perceptrons) need to be exposed to training data in order to learn how to generalize an output from an input data set. Once properly trained, RNA is able to make predictions without output data (Guoqiang Zhang, B. Eddy Patuwo, Michae! Y. Hu, Forecasting with artificial neural networks: The State of the Art, International Journal of Forecasting, Volume 14, Issue 1, March 1, 1998, Pages 35-62, ISSN 0169-2070, 10.1016 / SO169-2070 (97) 00044-7) and (Haykin, S., Neural Networks: The Comprehensive Foundation, 2nd edition, Prentice Hall, 1999 ISBN: 0132733501).

In the present method to determine the viability and quality of embryos, the information is extracted from the embryo image by means of a protocol, and then processed by a computer program that works the received data, transforming them into input variables for RNA and performs the processing, resulting in an embryonic quality score in the same standards as an embryologist would provide ("Excellent", "Good", "Regular" and "Bad" grades).

In a first step of the method to determine embryo viability and quality, embryo images are obtained, and preferably images of blastocysts are selected, from the initial blastocyst to expanded blastocyst phase.

Necessarily, the image must have the embryo in its entirety, with clearness of the zona pellucida and the delimitation of the blastocele should be visible.

In a second step, the embryo's image is processed by an image processing software, and measurements are obtained from the selection of image points.

The variables obtained using the computer image processing program include: a) First embryo diameter (DE1), which comprises the measurement of the diameter of the embryo obtained through a line joining the most extreme points of the embryo, disregarding the diameter of the zona pellucida (Figure 1); (b) Second Embryo Diameter (DE2), which comprises the measurement of the diameter of the embryo obtained through a line approximately perpendicular to the DE1 line, which connects the most extreme points of the embryo, disregarding the diameter of the zona pellucida (Figure 2 ); (c) first zone of the zona pellucida (DZP1), comprising the measurement of the diameter of the embryo obtained through a line connecting the most extreme points of the embryo, taking into account the diameter of the zona pellucida (Figure 3); (d) Second Pellucid Zone Diameter (DZP2), which comprises the measurement of the diameter of the embryo obtained through a line perpendicular to DZP1, connecting the most extreme points of the embryo and taking into account the diameter of the pellucid zone (Figure 4); e) Area of the embryo (AE), being a contour that surrounds the whole area of the embryo (not considering the area of the zona pellucida). The accuracy of this selection is important because various information is extracted from the area measurement (embryo area, embryo color density, embryo circularity) (Figure 5); f) Area of the zona pellucida (AZP), being made a contour that involves the whole area of the embryo, being considered the area of the zona pellucida. The accuracy of this selection is important because various information is extracted from the area measurement (Pellucid Zone Area, Total Color Density, Pellucid Zone Roundness) (Figure 6); g) Dead cell area (MCA), with a contour that surrounds the entire dead cell area, including both dead cells that are still attached and loose from the embryo (Figure 7); (h) Color Density of the zona pellucida (DCZP), a contour being drawn that involves only the area of the zona pellucida, not including the area of the embryo. This variable need not be obtained if the perivitelline space is absent, as in this particular case it can be calculated by AE (embryo area) and AZP (zona pellucida area) (Figure 8).

In a third step, the variables obtained by the computer program for image analysis are processed to serve as input to the RNA, being stored in a database. Data on the diameter of the embryo (DE) is obtained by formula 1. Formula 1: diameter of the embryo (DE) Data on diameter of the zona pellucida (DZP) is obtained from formula 2. Formula 2: diameter of the zona pellucida ( DZP) Live cell area (ACV) data is obtained from formula 3. Formula 3: ACV (live cell area) ACV = AE -ACM Embryo color density (DCE) data is obtained from the first computer program along with the area of the embryo (EDE) and the ideal stage for the time elapsed since fertilization. The Group Mean Relationship (RMG) data comprises the relationship between the development phase (EDE) and the Group Average, determined by formula 5. Thus, values greater than 1 will indicate that the embryo is ahead of values of less than 1 will indicate that the embryo is behind its group. Formula 5: relationship to group mean (RMG): Where: EDE group corresponds to the average embryonic phase of all other embryos in the same collection. The degree of embryo roughness or granulation (EDG) is obtained through a macro for the computer program, allowing to identify the contrast regions in the embryo and to count these regions in order to numerically represent this visual characteristic. The Sharp Edges (EDG) macro uses the basic Sharpen and FindEdges operations of an image analysis software, consisting of the operations: run ("UnsharpMask ...", "radius = 10 mask = 0.50"); run ("FindEdges");

Then the FindMaxima option (Noise: 100; No light background) should be used. The average diameter ratio (RDM) data of the embryo is obtained by formula 6, where DE (embryo diameter) and DZP (pellucid zone diameter) are obtained by averaging the largest and smallest diameters, respectively. embryo and the zona pellucida, favored by the large constancy of the diameter of the zona pellucida when compared with that of the embryo itself. Formula 6: Average Diameter Ratio (RDM) Data on Living Cell Ratio (RCV) is obtained from Formula 7. Formula 7: Living Cell Ratio (RCV) Data on Dead Cell Ratio (RCM) is obtained. using formula 8. The higher the SPC value, the higher the proportion of dead cells in the embryo, which will have a negative impact on its quality. Formula 8: Dead Cell Ratio (SPC) Data on the embryo color density ratio (RDC) is obtained from formula 9. Embryo color is an important factor to analyze because it is directly influenced by its density. cells and how viable they are. This variable is highly dependent on the conditions under which the image was obtained. However, these variations are compensated when an embryo color (DCE) to pellucid zone (DCZP) relationship is established. Formula 9: Embryo Color Density Ratio (RDC) RDC values less than 1 indicate an embryo lighter than its pellucid zone, while values greater than 1 indicate a darker embryo. Color intensity (DCE and DCZP) is measured by averaging the brightness values of each pixel in the area in question. This value ranges from 0 for completely dark to 255 for completely white. The circularity squared ratio (RCE2) data is obtained from formula 10. Formula 10: circularity ratio (RCE) An ideal circle will have a value of 1. The closer to zero, the less the measured shape resembles a circle. The circularity is defined as RCE = CE / CZP, where CE is the circularity of the embryo and CZP is the circularity of the zona pellucida.

Since the zona pellucida is constant and very circular, values close to 1 indicate high circularity of the embryo, while values close to 0 indicate very low circularity.

However, for the purpose of attributing different characteristics to circular or non-circular embryos by RNA, given that the values are always very close to 1, the RCE2 (squared circularity ratio) is used to highlight the numerically small differences in circularity, where the most circular embryos continue tending to 1. The data on embryonic quality (Q) varies on an Excellent, Good, Regular and Bad scale, being used as a template for the training of the Artificial Neural Network. . RNA has 4 neurons in its output layer, each representing one of the possible quality classes (neurons with values 0 or 1).

In a fourth step, the RNA architecture is developed, defining the number of layers, number of neurons in the layers, transfer functions between neurons and training functions. The structure of a Neural Network is composed of several elements, such as the number of neuron layers, the number of neurons in each layer, their transfer functions, and the training function of the network. Although the correct determination of these factors is of paramount importance for the better development of RNA, there is no fixed protocol for determining the best architecture (Xin Yao, Yong Liu, Applied Mathematics and Computation, Volume 91). , Issuel, April 1998, Pages 83-90, ISSN 0096-3003, 10.1016 / S0096-3003 (97) 10005-4). The program was executed with a stopping condition of 10,000 cycles, a range of 5 to 20 neurons in the first and second layers, a draw of transfer functions between tansig, logig and purelin and a draw of training functions between trainlm, trainscg and traingdx. The error was calculated with confusion matrix (percentage of misclassifications). The RNA was developed using the backpropagation algorithm.

Thus, the best RNA architecture was obtained for this specific case, being a network with 18 neurons in the first layer and purelin transfer function (linear function), 13 neurons in the second layer and logsig transfer function (logistic function). The training function, chosen by the algorithm, was trainscg (ScaledConjugateGradientAlgorithm).

As shown in figure 9, the inputs comprise variables used to create the different RNA configurations. The computer program draws values for all variables and creates an RNA that is trained with the database defined in the third step of this method containing the embryonic data, using the data split into training, validation and testing according to the choice at the beginning of the program. This process occurs cyclically, and at each cycle the program compares the error of the current network with the past networks, saving the RNA with the smallest error. Once any of the stop variables are reached, the program terminates the cycle and displays the best result. Next, a graphical interface with fields for inputting the variables obtained through the image analysis program is used, and the variables are standardized and a previously trained RNA simulation is performed.

RNA input variables include EDE (Embryonic Developmental Stage), CPD (Post-Copulation Days), RMG (Developmental Stage to Group Average), RDM (Mean Embryo-Diameter Ratio). zona pellucida), RCV (Ratio of Living Cell Area to Total Area), RCM (Ratio of Dead Cell Area to Living Cell Area), RDC (Embryo Color Density Ratio), RCE2 (Relationship between Embryo Circularity and Squared Area), EDG (Macro Sharp EDGes), RAB (Relationship between Blastocele Area and Embryo Area), RDCB (Relationship between Blastocele Color Density and embryo color), the RCB2 (Relationship between Blastocele Circularity and squared embryo circularity). The embryonic quality assessment by RNA can be presented by a bar graph, which represents each of the RNA outputs (the 4 degrees of quality), with the height of each bar determined by the magnitude of the output value; and / or through a quality score according to the highest output value of the network, the results being preferably presented as Excellent, Good, Fair and Bad; and / or a descriptive vector, which is the actual RNA output vector, representing the values of the four output layer neurons (“Excellent”, “Good”, “Regular” and “Bad”, respectively). Table 1 presents the RNA results for the Test data. The ID column represents the embryo ID in the database. The error column was calculated by subtracting the quality provided by RNA from the quality provided as a template (embryologist's assessment), so that 0 indicates a hit, +1 that the RNA gave a lower quality than the template and -1 that the RNA attributed a higher quality than the template. TABLE 1: RNA Results for Test Data. The four RNA outputs represent the quality grades of the analyzed embryo, being 1 (excellent), 2 (good), 3 (regular) and 4 (bad).

Claims (4)

Method for determining the viability and quality of embryos characterized by comprising the steps of: a) obtaining the digital image of the embryo; b) processing the digital image of the embryo in an image processing computer program, obtaining measurements from the selection of image points; c) obtaining the variables: c.1) first diameter of the embryo (DE1), which comprises the measurement of the diameter of the embryo obtained through a line joining the most extreme points of the embryo; c.2) second diameter of the embryo (DE2), comprising the measurement of the diameter of the embryo obtained through a line approximately perpendicular to the line DE1, which connects the most extreme points of the embryo; c.3) first zone of the zona pellucida (DZP1), which comprises the measurement of the diameter of the embryo obtained through a line connecting the most extreme points of the embryo, taking into account the diameter of the zona pellucida; c.4) second diameter of the zona pellucida (DZP2), comprising the measurement of the diameter of the embryo obtained through a line perpendicular to DZP1, connecting the most extreme points of the embryo and taking into account the diameter of the zona pellucida; c.5) area of the embryo (AE), without considering the area of the zona pellucida; c.6) area of the zona pellucida (AZP), being considered the area of the zona pellucida; o.7) dead cell area (ACM); ç. 8) color zone color density (DCZP); d) processing of the variables obtained to serve as input to the Artificial Neural Network, as follows: d. 1) diameter of the embryo (DE) obtained by the formula d.2) diameter of the zona pellucida (DZP) obtained by the formula d.3) area of living cells (ACV) obtained by the formula ACV = AE - ACM; d.4) embryo color density (DCE) obtained along with the embryo area (AE); d.5) total color density (DCTotal) obtained together with the area of the zona pellucida (AZP); d.6) color density of the zona pellucida (DCZP) obtained by the formula d.7) circularity of the embryo (EC) obtained together with area of the embryo (AE); d.8) pellucid zone circularity (CZP) obtained together with Pellucid Zone Area (AZP); d.9) Embryonic Development (EDE) phase obtained by the visual morphological analysis of the embryo, where EDE = 0.0 in the case of initial blastocyst, EDF = 0.5 in case of blasticite and EDE = 1.0 in case of blastocyte expanded; d. 10) Post-copulation (DPC) days comprises the time (in days) that has elapsed since the embryo-generating mammal's copulation; d. 11) Relationship with the group average (RMG) comprises the relationship between the development phase (EDE) and the group average, determined by the formula d. 12) degree of embryo roughness or granulation (EDG) obtained by means of a macro for the computer image processing program; d. 13) embryo mean diameter ratio (RDM) is obtained by formula d.14) living cell ratio (RCV) obtained by formula d. 15) Dead Cell ratio (SPC) obtained by the formula d. 16) embryo color density ratio (RDC) obtained by formula d. 17) squared circularity relation (RCE2) obtained by the formula e) development of the Artificial Neural Network architecture, defining the number of layers, number of neurons in the layers, transfer functions between neurons and training functions; f) standardization of variables and simulation with previously trained RNA, using a graphical interface with fields for input of variables obtained through the computer program for image analysis; g) embryonic quality assessment by Artificial Neural Network. Method for determining the viability and quality of embryos according to claim 1, characterized in that preferably the digital image is selected from blastocysts. Method for determining the viability and quality of embryos according to Claim 1, characterized in that the digital image of the embryo clearly shows the zona pellucida and the delimitation of the blastocele. Method for determining the viability and quality of embryos according to claim 1, characterized in that the evaluation of embryonic quality by Artificial Neural Network is presented by means of a bar graph representing each of the RNA outputs. height of each bar determined by the magnitude of the output value; and / or through a quality score according to the highest output value of the network, the results being preferably presented as Excellent, Good, Fair and Bad; and / or a descriptive vector, which is the actual RNA output vector, representing the values of the four output layer neurons (“Excellent”, “Good”, “Regular” and “Bad”, respectively).