JP2013174823A - Image processing device, microscope system and image processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image processing device or the like in which matching between thinly cut specimen images mutually different in a cut piece, a dyeing method or the like can be performed with high accuracy.SOLUTION: An image processing device includes: a component separation part 143 which separates and extracts specimen components photographed in each image from two images obtained by imaging two cut pieces collected from one specimen based on first foresight information on the specimen components of the specimen; a contribution rate calculation part 144 which calculates the degree of contribution contributed to alignment processing by the separated and extracted specimen components based on second foresight information on the specimen component; and a matching processing part 145 which specifies the relative position between the two images by calculating an appreciation value based on the degree of contribution.

Description

  The present invention relates to an image processing apparatus, a microscope system, and an image processing method for processing an image of a biological tissue specimen acquired by a microscope.

  Conventionally, diagnosis of biological tissue specimens including pathological specimens has been performed by slicing block specimens obtained by organectomy or specimens obtained by needle biopsy to a thickness of about several microns, and slicing the sliced specimens with a microscope. This is done by observing an enlarged observation image. Transmission observation using an optical microscope is one of the most widespread observation methods, since the equipment is relatively inexpensive and easy to handle, and has been performed historically. In recent years, it has become the mainstream to make a diagnosis on an image acquired by capturing an observation image with a camera attached to an optical microscope.

  By the way, a sliced biological tissue specimen (hereinafter referred to as a sliced specimen) hardly absorbs or scatters light and is nearly colorless and transparent. For this reason, generally, prior to microscopic observation, a sliced specimen is stained. Various dyeing techniques have been proposed, and the total number reaches 100 or more. Among them, especially for pathological specimens, hematoxylin-eosin staining (hereinafter referred to as HE staining) using two pigments of blue-violet hematoxylin and red eosin is used as standard.

  Further, clinically, when it is difficult to visually recognize a biological tissue to be observed by HE staining, or when morphological diagnosis of a biological tissue is interpolated, a special staining different from HE staining is applied to the specimen, A technique that visually enhances the color of the tissue may be used. Further, in histopathological diagnosis, for example, immunostaining using various marker proteins for visualizing antigen-antibody reaction of cancer tissue may be used.

  Thus, when observing a pathological specimen, a plurality of types of staining methods are often used. Therefore, in general, a plurality of sliced specimens collected from one block specimen are stained differently from each other, and images (thin slice specimen images) obtained by imaging them with a microscope apparatus Thus, comparative observation of the state of tissues and structures is performed. Therefore, it is important to accurately perform alignment (matching) between a plurality of sliced specimen images.

  As a technique related to matching of sliced specimen images, for example, Patent Document 1 discloses matching of sliced slice specimen images that have been subjected to HE staining and immunostaining (IHC) by phase-only correlation or the like. It is disclosed.

International Publication No. 2008/108059

The matching method disclosed in Patent Document 1 is based on the premise that thin sliced samples to be matched have little change in state, that is, similar continuous sections.
However, even though the samples were collected from the same block specimen, the sections of the sliced specimens are different, so that the state of the tissue or the like in the sliced specimen does not completely match. In particular, between sliced specimens with separated sections, the tissue or the like that appeared in one sliced specimen may disappear in the other sliced specimen, or the shape of the tissue or the like may change greatly. In addition, since the appearance of tissues and the like is greatly different between sliced specimens that have been subjected to different types of staining, it is very difficult to find tissues that are common between sliced specimen images. In these cases, since the similarity between the sliced specimen images is low, the sliced specimens that can be aligned are limited.

  The present invention has been made in view of the above, and an image processing apparatus, a microscope system, and an image processing method capable of performing high-precision alignment between sliced specimen images having different sections and staining methods. The purpose is to provide.

  In order to solve the above-described problems and achieve the object, the image processing apparatus according to the present invention provides a sample of the specimen from two images obtained by imaging two sections taken from one specimen. Based on the first foresight information relating to the constituent elements, the specimen constituent element separating unit for separating and extracting the specimen constituent elements shown in each image, and the sample separated and extracted based on the second foresight information relating to the specimen constituent elements A contribution calculation unit that calculates a contribution that a component contributes to the alignment process; an alignment processing unit that specifies a relative position between the two images by calculating an evaluation value based on the contribution; and It is characterized by providing.

  In the image processing apparatus, the first look-ahead information is information representing a color distribution characteristic of the sample component, and the second look-ahead information is information representing a shape characteristic of the sample component. It is characterized by.

  The image processing apparatus further includes an attribute information setting unit that sets information related to the attribute of the pathological specimen, and a foresight information acquisition unit that acquires the first and second foresight information based on the information related to the attribute. It is characterized by that.

  In the image processing apparatus, the information related to the attribute of the pathological specimen is information representing a part in the living body from which the pathological specimen is collected, and the foresight information acquisition unit includes the part in the living body and the specimen included in the part. The first and second foresight information is acquired through information associated with a component.

  In the image processing apparatus, the sample component separation unit calculates a color feature amount of a plurality of pixels constituting each image, and based on the color feature amount and the first look-ahead information, A pixel group corresponding to the sample component is extracted from the pixel.

  In the image processing apparatus, the contribution calculation unit refers to the second look-ahead information, acquires the feature of the shape of the sample component separated and extracted by the sample component separation unit, and the feature of the shape The contribution is calculated based on the above.

  In the image processing device, the sample component separation unit corresponds to the sample component by extracting a pixel group corresponding to the sample component from a plurality of pixels constituting each image and performing a labeling process. A label image including a label area is generated, and the alignment processing unit performs alignment between the label images corresponding to the two images, respectively.

  In the image processing apparatus, the specimen component includes at least one of a main tissue including elastic fibers, a cell membrane, a cytoplasm, a red blood cell, a cell nucleus, and a cavity.

  A microscope system according to the present invention is provided to face the stage, the image processing apparatus, a stage on which the sliced specimen is placed, an illumination optical system that irradiates illumination light toward the stage, and the stage. An objective optical system, a moving means for changing the position of the stage relative to the imaging field of the objective optical system, and an imaging unit that images the observation light transmitted through the objective optical system and generates image information, The image processing apparatus further includes an image acquisition unit that acquires image information generated by the imaging unit.

  In the microscope system, the image processing device is generated by the imaging unit imaging the first image acquired in advance and the sliced specimen placed on the stage, and the image acquisition unit Control for changing the imaging field of view by performing alignment processing with the second image corresponding to the image information acquired through the control, and controlling the moving means according to the processing result of the alignment processing It further has a section.

  The image processing method according to the present invention is based on first look-ahead information relating to the sample component of the specimen from two images obtained by imaging two sections taken from one specimen respectively. Based on the sample component separation step for separating and extracting the sample components reflected in the above and the second look-ahead information relating to the sample components, the degree of contribution that the sample components extracted and contributed to the alignment processing is calculated The method includes a contribution calculation step, and an alignment processing step that specifies a relative position between the two images by calculating an evaluation value based on the contribution.

  According to the present invention, the sample components captured in each image are separated and extracted based on the foresight information, the degree of contribution that each sample component contributes to the alignment processing between images is calculated, and based on this contribution By calculating the evaluation value, the relative position between the two images is specified, so that it is possible to perform highly accurate alignment even between sliced specimen images having different sections and staining methods.

FIG. 1 is a block diagram showing a configuration of an image processing apparatus according to Embodiment 1 of the present invention. FIG. 2 is a table showing foreseeing information in which organs and organs from which thin sliced specimens are collected are associated with main specimen constituent elements included in each organ. FIG. 3 is a table showing foreseeing information of the shape characteristics of the specimen constituent elements. FIG. 4 is a table showing foresight information in which specimen constituent elements, staining types, and color distribution information are associated with each other. FIG. 5 is an xy chromaticity diagram in the CIExy color system, which is an example of color distribution characteristics. FIG. 6 is a graph illustrating an example of a spectrum of spectral transmittance that is an example of color distribution characteristics. FIG. 7 is an example of a distribution diagram of the amount of pigment, which is an example of color distribution characteristics. FIG. 8 is a flowchart showing the operation of the image processing apparatus shown in FIG. FIG. 9 is a schematic diagram illustrating an example of a sliced specimen image that is an image processing target. FIG. 10 is a schematic diagram illustrating an example of a sample attribute setting screen. FIG. 11 is a flowchart showing the separation and extraction of the sample components and the label image creation process. FIG. 12 is a schematic diagram illustrating an example of a label image. FIG. 13 is a flowchart showing a contribution rate calculation process. FIG. 14 is a histogram showing an example of the ratio of circular to circular ratio distribution. FIG. 15 is a flowchart showing the matching process. FIG. 16 is a diagram for explaining the matching process. FIG. 17 is a diagram for explaining an example of matching processing in each section. FIG. 18 is a schematic diagram illustrating a display example of matching results. FIG. 19 is a diagram showing a configuration of a microscope system according to Embodiment 2 of the present invention. FIG. 20 is a flowchart showing the operation of the microscope system shown in FIG. FIG. 21 is a schematic diagram for explaining a microscope observation method according to the second embodiment of the present invention.

  Hereinafter, embodiments of an image processing apparatus, a microscope system, and an image processing method according to the present invention will be described in detail with reference to the drawings. Note that the present invention is not limited to these embodiments. Moreover, in description of each drawing, the same code | symbol is attached | subjected and shown to the same part.

(Embodiment 1)
FIG. 1 is a block diagram showing a configuration of an image processing apparatus according to Embodiment 1 of the present invention.
As shown in FIG. 1, the image processing apparatus 10 according to Embodiment 1 includes a microscopic observation image (hereinafter referred to as a sliced specimen image or a sliced specimen image) of a plurality of sections of one biological tissue specimen such as a pathological specimen. An input unit 11 that receives an instruction and information input to the image processing apparatus 10, and displays a sliced sample image and other information on a screen. A display unit 12, a storage unit 13 storing a program used in the main image processing, a calculation unit 14 that performs matching processing between sliced specimen images, and a control unit 15 that controls operations of these units. Is provided. The image processing apparatus 10 is provided with a recording apparatus 20 having an image database 21.

  The input unit 11 includes input devices such as a keyboard, various buttons, and various switches, and pointing devices such as a mouse and a touch panel. The input unit 11 receives a signal input through these devices and inputs the signal to the control unit 15. In addition, when the user selects a desired image file from among the image files recorded in the image database 21 using the device, the input unit 11 inputs a selection signal for selecting the file to the control unit 15. An image file designating unit 111 for performing the operation.

  The display unit 12 is configured by a display device such as an LCD, an organic EL display, or a CRT display, for example, and displays various screens on which various information and images are arranged in a predetermined format in accordance with a control signal output from the control unit 15. indicate.

  The storage unit 13 includes a semiconductor memory such as flash memory, RAM, and ROM that can be updated and recorded, a recording medium such as a hard disk, MO, CD-R, and DVD-R that is built in or connected by a data communication terminal, and the recording medium. It is comprised by the reader etc. which read the recorded information. The storage unit 13 records various programs executed by the calculation unit 14 and the control unit 15 and various setting information, and temporarily stores an image file read from the recording device 20. Specifically, the storage unit 13 stores an image processing program 131 that performs a matching process between a plurality of sliced specimen images, and a foresight in which foresight information used by the calculation unit 14 during execution of the image processing program is stored. And an information storage unit 132.

Here, the foresight information refers to elements of the sliced specimen, elements such as tissues and structures constituting the sliced specimen (hereinafter collectively referred to as constituent elements or specimen constituent elements), and each specimen. This is general information regarding the characteristics of the constituent elements, the type and characteristics of staining applied to the sliced specimen, or information (teacher data) acquired in advance by experiments. These pieces of foreseeing information are used when separating and extracting specimen constituent elements from the sliced specimen image and calculating the contribution rate of each specimen constituent element to the matching process. Note that the specimen constituent elements are roughly classified and include main tissues and general tissues constituting a living body, cavities, and specific areas. The main tissue is a tissue that is abundant in organs, and includes elastic fibers, collagen fibers, blood vessels, and the like. A general tissue is an element other than the main tissue, and includes intracellular components such as cytoplasm, cell membrane, cell nucleus (hereinafter also simply referred to as nucleus), blood cells (blood cells) such as red blood cells and white blood cells, and the like. A cavity is a region where there is no main tissue or general tissue, and corresponds to, for example, a cross-section of a tubular tissue or a gap between cytoplasms. The singular region is a region that exhibits a specific phenomenon in which the light applied to the region adjacent to the elastic fiber or the elastic fiber in the microscopic device is not attenuated under certain conditions and conversely increases the light intensity. . Such a specific phenomenon is considered to occur when a collection of fine fibers such as elastic fibers is in the form of a band or bundle.
Specific contents of the foresight information will be described later.

  The calculation unit 14 is configured by hardware such as a CPU, for example, and executes a matching process between a plurality of sliced sample images read from the image database 21 by reading the image processing program 131.

  More specifically, the calculation unit 14 sets a sample attribute setting unit 141 that sets an attribute of a sliced sample captured in two sliced sample images to be processed, and a foresight from the storage unit 13 based on the set attribute. A foresight information acquisition unit 142 for acquiring information, a sample component separation unit 143 for separating and extracting the sample components in the sliced specimen image based on the foresight information, and a matching process for each sample component extracted and extracted A contribution rate calculation unit 144 that calculates the contribution rate based on the foresight information, and a matching processing unit that specifies a relative position between the two sliced sample images by calculating a matching evaluation value based on the calculated contribution rate 145.

  The control unit 15 is configured by hardware such as a CPU, for example, and by reading various programs stored in the storage unit 13, to each unit configuring the image processing apparatus 10 according to an operation signal input from the input unit 11. Instruction, data transfer, and the like, and overall operation of the image processing apparatus 10 is controlled.

  Such an image processing apparatus 10 can be configured by a general-purpose device such as a personal computer or a workstation.

  The recording device 20 includes, for example, a recording medium such as a hard disk, MO, CD-R, and DVD-R connected to the image processing apparatus 10 through a data communication terminal, and a reading device that reads information recorded on the recording medium. The Note that the recording device 20 may be built in the image processing device 10. In this case, in addition to the various recording media and reading devices listed above, update-recordable flash memory, semiconductor memory such as RAM and ROM, etc. Can be used.

  The image database 21 records a plurality of image files generated by imaging thin sliced specimens in a plurality of sections of one biological tissue specimen with a microscope apparatus. Each image file includes image data of the sliced specimen image and accompanying information. The incidental information includes, for example, the type of organ from which the biological tissue specimen was collected, the type of staining applied to the sliced specimen, the microscopic method when the sliced specimen image was imaged, the magnification of the image, and the thickness of the sliced section Etc. are included.

Next, the content of the foresight information stored in the foresight information storage unit 132 will be described.
FIG. 2 is a table showing foreseeing information in which organs and organs from which thin sliced specimens are collected are associated with main specimen constituent elements included in each organ. For example, in a sliced specimen collected from the lung, elastic fibers, bronchioles and blood vessels are the main specimen components.

  FIG. 3 is a table of foreseeing information representing the shape characteristics of each specimen component. The feature of the shape is, for example, the thickness n in the case of the nucleus, the diameter rd in the case of red blood cells, the diameter bd of the cross section perpendicular to the longitudinal direction in the case of blood vessels, and the diameter ed of the same cross section in the case of elastic fibers. Each is represented.

  FIG. 4 is a table of look-ahead information representing the types of staining that can stain each specimen component and data representing the color distribution characteristics of each pixel determined according to the combination of the specimen components and the type of staining. The color distribution characteristics can be represented by, for example, an xy chromaticity diagram in the CIExy color system, a spectrum of spectral transmittance, a distribution of estimated dye amount, and the like.

  FIG. 5 is an xy chromaticity diagram in the CIExy color system. The regions P1, P2,... on the xy chromaticity diagram represent color distribution characteristics according to the sample constituent elements and the type of staining. When the xy chromaticity diagram is used as the color distribution characteristic, coordinate data representing the regions P1, P2,... As illustrated in FIG. 5 is stored as the color distribution characteristic data 1, 2,.

  FIG. 6 is a graph showing an example of a spectrum of spectral transmittance at a point on the sliced sample corresponding to the pixel position of the sliced sample image. When an estimated spectrum of spectral transmittance is used as the color distribution characteristic, spectral data as illustrated in FIG. 6 is stored as the color distribution characteristic data 1, 2,... Shown in FIG.

The spectrum of the spectral transmittance can be estimated from the sliced sample image corresponding to each band (wavelength band) imaged in multiband by the following calculation.
Spectral transmittance t (x _n, λ _k), the pixel of each pixel in the background image value _{I 0 (x n, λ k} ) and the pixel value I of each pixel in the thin section sample image (x _n, λ _k) And is given by the following equation (1). Here, the background image is a microscopic image obtained without placing a specimen. Further, x _n is a position vector indicating the pixel position of each pixel constituting the sliced specimen image, and n = 1 to N (N is the number of pixels of the sliced specimen image). λ _k indicates the center wavelength of each band (k is the number of bands).

  FIG. 7 is an example of a distribution diagram of a pigment amount at a point on a thin slice sample corresponding to a pixel position of the thin slice sample image. Sample components (nucleus, cytoplasm, fiber, blood) subjected to HE staining are shown in FIG. The pigment amount distribution is shown. When the dye amount is used as the color distribution characteristic, the dye amount distribution data illustrated in FIG. 7 is stored as the color distribution characteristic data 1, 2,... Shown in FIG.

式（２）において、係数ｋ_H（λ_k）及びｋ_E（λ_k）は、色素Ｈ及び色素Ｅにそれぞれ対応する係数であり、波長に依存して決まる物質固有の値である。また、値ｄ_Hは色素Ｈの色素量であり、値ｄ_Eは色素Ｅの色素量である。なお、ＨＥ染色以外の色素についても、上述した方法と同様にして算出することができる。 The amount of dye on the sliced specimen can be estimated from the sliced specimen image of each band (wavelength band) imaged in multiband by the following calculation.
First, the spectral transmittance t (x _n , λ _k ) is calculated from the equation (1). After that, according to the Lambert-Beer rule, a simultaneous equation in which the following equation (2) that gives the spectral transmittance t (x _n , λ _k ) is simultaneous with respect to a plurality of wavelength components λ _k is solved.

In Expression (2), coefficients k _H (λ _k ) and k _E (λ _k ) are coefficients corresponding to the dye H and the dye E, respectively, and are values specific to the substance determined depending on the wavelength. The value d _H is the amount of dye H, and the value d _E is the amount of dye E. In addition, it can calculate similarly to the method mentioned above also about pigments other than HE dyeing | staining.

Next, the operation of the image processing apparatus 10 will be described. FIG. 8 is a flowchart showing the operation of the image processing apparatus 10.
First, prior to image processing, the image processing apparatus 10 causes the display unit 12 to display a plurality of image files recorded in the image database 21. The user selects two thin slice sample image files to be aligned from the displayed image files using an input device such as a mouse. The input unit 11 receives selection signals for two image files and inputs them to the control unit 15.
In addition, the image displayed here is an image each image | photographed with respect to the 2 section | slice (thin slice sample) extract | collected from one pathological sample.

  In step S <b> 10, the control unit 15 acquires the two selected image files from the image database 21 and temporarily stores them in the storage unit 13.

FIG. 9 is a schematic diagram illustrating an example of a sliced specimen image that is an image processing target. In the sliced specimen image M1 shown in FIG. 9, specimen constituent elements such as an elastic fiber a10, other main tissues a11, nuclei a12, erythrocytes a13, singular regions a14, and cavities a15 are shown. The four corners of the image M1 are background areas a16 that are outside the field of view of the microscope. In addition to the above-described specimen constituent elements, FIG. 9 also shows a region a17 that is not a target of matching processing, which will be described later, such as the inside of a blood vessel or the medulla.
The presence / absence, position, shape, and the like of the various specimen components described above are likely to be different between the two sliced specimen images acquired in step S10.

  In subsequent step S11, the sample attribute setting unit 141 reads the auxiliary information of the image data from the acquired image file, and sets the attribute of the sliced sample based on the auxiliary information. Furthermore, the sample attribute setting unit 141 creates a screen on which the attributes of the set sliced sample are displayed, and causes the display unit 12 to display the screen.

  FIG. 10 is a schematic diagram illustrating an example of a sample attribute setting screen displayed on the display unit 12. The specimen attribute setting screen B1 shown in FIG. 10 includes two sliced specimen images M2a and M2b selected by the user, the types of organs (for example, liver) reflected in the sliced specimen images M2a and M2b, Types of staining (for example, HE staining and DAB staining) applied to the sliced specimens shown in the sliced specimen images M2a and M2b, and a spectroscopic method when imaging the sliced specimen images M2a and M2b (for example, bright field) Observation, bright field observation), magnifications of sliced specimen images M2a, M2b (eg, 10 times, 40 times), thickness of sliced specimens (eg, 4 μm), slices of sliced specimen images M2a, M2b (slices) Attribute information such as a distance (for example, discontinuity, 12 μm), an OK button b1, and a cancel button b2 are shown. Note that the thickness of the sliced specimen is 4 μm and the distance between the slices of the sliced specimen image M2a and the sliced specimen image M2b is 12 μm, which means that the two images are separated by three slices. .

  The user confirms the attribute information on the sample attribute setting screen B1, and if there is no mistake, selects the OK button b1 using an input device such as a mouse. In response to this, the input unit 11 receives a selection signal for the OK button b1 and inputs it to the control unit 15. The sample attribute setting unit 141 determines the sample attribute with the content displayed on the sample attribute setting screen B1 according to the control signal from the control unit 15.

  In the sample attribute setting screen B1, the display contents of each attribute information can be changed by, for example, a pull-down menu, and the user changes the contents of desired attribute information using an input device such as a mouse. The attribute information can be changed by selecting the OK button b1. In this case, the input unit 11 receives the change information of the attribute information and the selection signal of the OK button b1 and inputs them to the control unit 15. The sample attribute setting unit 141 determines the sample attribute with the changed content according to the control signal from the control unit 15.

  In subsequent step S12, the foresight information acquisition unit 142 acquires the foresight information from the foresight information storage unit 132 based on the set sample attribute. Specifically, first, based on the organ (for example, liver) from which the sliced specimen is collected among the sample attributes, main sample components (for example, veins, hepatocytes, capillaries) from the table shown in FIG. ) Is acquired. Further, based on the acquired specimen components, the shape characteristics of each specimen component are acquired from the table shown in FIG. Furthermore, data representing color distribution characteristics is acquired based on the type of staining (for example, HE staining, DAB staining) among the sample attributes and the sample components acquired from the table shown in FIG.

  In subsequent step S13, the sample component separation unit 143 separates and extracts the sample components from each of the two sliced sample images to be processed, and creates a label image corresponding to each sample component. FIG. 11 is a flowchart showing the separation and extraction of the sample components and the label image creation process. In the following, as an example, processing for the sliced sample image M1 shown in FIG. 9 will be described.

First, in step S131, the sample component separation unit 143 removes the background region a16 from the image M1.
In subsequent step S132, the specimen component separation unit 143 extracts the cavity a15 from the image M1. Here, since the cavity is an area where no specimen component exists, an area having a luminance value higher than a predetermined threshold may be extracted from the image M1 as the cavity 15a. The luminance value Y is given by the following equation (3) from the pixel values (R value, G value, B value) of each pixel, for example.
Y = 0.30R + 0.59G + 0.11B (3)

  In step S133, the specimen component separation unit 143 extracts a pixel group corresponding to the main tissue region and the general tissue region from the region excluding the cavity a15 based on the foresight information acquired by the foresight information acquisition unit 142. For example, these pixel groups calculate color feature amounts (for example, coordinates on the xy chromaticity diagram in the CIExy color system, estimated spectrum of spectral transmittance, and estimated dye amount) from the pixel values of each pixel in the image M1. Then, it can be extracted by performing principal component analysis from the color distribution characteristics for each sample component in the foresight information. Alternatively, for each color feature amount of each pixel in the image M1, by performing learning discrimination processing using a discriminator such as a support vector machine (SVM) using the color distribution characteristics of the foresight information as teacher data, each sample You may extract the pixel group applicable to a component. Thereby, in the case of the image M1, the elastic fiber a10, the other main tissue a11, the nucleus a12, and the red blood cell a13 are extracted.

  In step S134, the specimen component separation unit 143 extracts the singular region a14 from the remaining region after extracting the cavity a15, the main tissue, and the general tissue from the image M1. More specifically, first, the remaining area is binarized by the luminance value of each pixel, and an area having a luminance value equal to or greater than a predetermined value is extracted as a specific pixel. When there are a plurality of specific pixels adjacent to each other, these pixel regions are extracted as specific regions. On the other hand, the scattered specific pixels are determined as noise and are not extracted. Note that the singular region a14 may be extracted using the estimated pigment amount calculated from the pixel value of each pixel instead of the luminance value.

  In step S135, the specimen component separation unit 143 is publicly known for the regions (elastic fiber a10, other main tissue a11, nucleus a12, red blood cell a13, singular region a14, and cavity a15) extracted in steps S132 to S134. In this way, a label image is generated in which a group of connected pixels is used as one label area. Thereby, fine information in each extracted region is excluded, and each region can be characterized only by shape information. In the case of a fiber structure such as the elastic fiber a10, the connectivity of the pixels constituting the line segment is determined, and the pixels determined to be connected are defined as one label region. Each label area is indexed for each type of label area (that is, for each type of corresponding sample component).

FIG. 12 is a schematic diagram illustrating an example of a label image corresponding to the image M1. 12 includes an elastic tissue label region (hereinafter referred to as an elastic tissue label) b10, a main tissue label region (hereinafter referred to as a main tissue label) b11-1, b11-2,. Label regions (hereinafter referred to as nuclear labels) b12-1, b12-2,..., Red blood cell label regions (hereinafter referred to as red blood cell labels) b13-1, b13-2,. .., And cavity label areas (hereinafter referred to as cavity labels) b15-1, b15-2,.
Thereafter, the process returns to the main routine.

  In step S14 following step S13, the contribution rate calculation unit 144 calculates the contribution rate that each label region contributes to the matching process based on the look-ahead information (see FIG. 3) of the sample component corresponding to each label region. . FIG. 13 is a flowchart showing a contribution rate calculation process. In the following, as an example, processing for the label image M3 shown in FIG. 12 will be described.

  First, in step S141, the contribution rate calculation unit 144 sets the contribution rates of the unique region labels b14-1, b14-2,... Existing in the label image M3 to zero. This is because the singular region has a high possibility of being included in only one of the two images to be matched, and is not used for the matching process.

  In the subsequent step S142, the contribution rate calculation unit 144 selects the sample component whose thickness is less than or equal to the inter-slice distance of the sliced sample based on the foresight information (see FIG. 3) regarding the shape characteristics of the sample component. Set the contribution of the label area to zero. Here, the distance t between slices is expressed by t = d × s by the thickness d of the sliced specimen and the section interval s. For example, in the case of the nuclear labels b12-1, b12-2,..., The thickness of the nucleus is n μm, and therefore the contribution ratio of the nuclear label b12 is set to zero when n ≦ t. This is because a sample component whose thickness is equal to or less than the distance between slices is likely to be included in only one of the two images to be matched, so that it is not used for the matching process. .

  In step S143, the contribution rate calculation unit 144 calculates shape feature amounts for the remaining label regions (that is, the contribution rate is not set to zero). Examples of the shape feature amount of the label area include area, perimeter length, circularity, maximum fillet diameter, minimum fillet diameter, and extension degree.

In step S144, the contribution rate calculation unit 144 calculates the ratio of the label area to the circle from the shape feature amount calculated in step S143. The ratio ρ to circular is given by the following equation (4) using the circularity e, the maximum fillet diameter f _max , and the minimum fillet diameter f _min .
ρ = e × (f _min / f _max ) (4)

In step S145, the contribution rate calculation unit 144 creates a distribution of the ratio of circularity calculated in step S144, and calculates the contribution rate of each label area from this distribution. FIG. 14 is a histogram showing an example of the ratio of circular to circular ratio distribution. The contribution rate V _CR of each label area is obtained by normalizing the label number L of the ratio to circularity ρ calculated for the label area by the maximum label number L _max in the histogram, and V _CR = L / L _max Given.
Thereafter, the process returns to the main routine.

  In subsequent step S15, the matching processing unit 145 executes matching processing between the label images respectively created from the two sliced sample images based on the contribution ratio of each label region. As the matching process, there is a template matching (for example, geometric pattern matching) that allows a pattern to be deformed, that is, is not affected by a rotation or a change in scale, and can select and match a characteristic feature of the pattern. Used. The specific method of the pattern matching process is not particularly limited, and a known method can be used.

FIG. 15 is a flowchart showing the matching process. FIG. 16 is a schematic diagram for explaining the matching process.
First, in step S151, the matching processing unit 145 sets a template image or a template region. At this time, when the magnifications of the two label images to be matched are different from each other, the matching processing unit 145 sets the label image having the larger magnification as the template image. At the same time, the size of the image with the larger magnification serving as the template image is reduced to match the magnification of the label image to be matched. FIG. 16A is a schematic diagram showing a matched label image M3, and FIG. 16B is a schematic diagram showing a label image (hereinafter referred to as a template image) T set as a template image.

  Further, when the two label images to be matched have the same magnification, the matching processing unit 145 sets one of the label images as a template image. Note that the template image may be set such that the user can manually select an image or select a partial area in the image. In this case, the matching processing unit 145 sets the selected image or region as a template image according to the input signal from the input unit 11.

  In subsequent step S152, the matching processing unit 145 determines the matching order of the label areas included in the template image (area) T. For the matching order, basically, a label area having a large area is given priority. In addition, when there is a variation in area for each type of label area, a continuous matching order is given to the same label area as the label area having the largest area. For example, if the areas of three nuclear labels are 30, 15, 5 (units are arbitrary) and the areas of three cytoplasmic labels are 50, 20, 10 (same as above), the areas of the three cytoplasmic labels are Matching order is given in descending order, then in order of increasing area of the three nuclear labels.

In step S153, the matching processing unit 145 calculates a feature amount _ACH that serves as an index when matching is performed for each label region in the template image T and the label image M3. Examples of the feature amount A _CH include geometric parameters, label spectrum analysis (spectral components calculated by fast Fourier transform (FFT)), and the like.

Subsequently, as shown in FIG. 16C, the matching processing unit 145 scans the template image T on the label image M3, and loops A for each section in the label image M3 that overlaps the template image T. Execute the process. In the following steps S154 and S155, as shown in FIG. 17, processing between the template image T and one section M3-i (i = 1, 2,...) In the label image M3 will be described. As shown in FIG. 17A, the template image T shows cytoplasm labels C _1-1 and C _1-2 and nuclear labels N _1-1 , N _1-2 and N _1-3. Yes. As shown in FIG. 17 (b), the compartment M3-i has cytoplasmic labels _C2-1 , _C2-2 , _C2-3 and nuclear labels _N2-1 , _N2-2. It is shown.

In step S154, the matching processing unit 145 calculates a matching error for each label area between the template image T and the section M3-i. For this purpose, the matching processing unit 145 first calculates a feature amount F indicating the degree to which each label region contributes to matching. This feature F has a contribution ratio V _CR calculated in step S14, given by the product between the feature amount A _CH calculated in step S153. For example, the cytoplasmic label C contributes feature quantity matching _1-1 F (C _1-1), the characteristic quantity of cytoplasmic label C _1-1 A _CH (C _1-1) the contribution rate V _CR (C _{1- 1} )) A _CH (C _1-1 ) × V _CR (C _1-1 )

Subsequently, the matching processing unit 145 extracts label regions in the template image T in the matching order determined in step S152, and calculates a matching error with the same type of label region in the section M3-i. The matching error is given as a value that minimizes the absolute value of the difference between the feature amounts F. For example, in the case of FIG. 17, the matching error M (C1) with the cytoplasmic labels C _2-1 , C _2-3 , C _2-3 in the compartment M3-i of the cytoplasmic label C _1-1 is expressed by the following equation (5) ).

Also, cytoplasmic label C _2-1 cytoplasmic label C _1-2 the compartment M3-i, C _2-3, matching error M (C2) of the C _2-3 becomes as the following equation (6) .

Nuclear matching error M (N1) of the label N _1-1 ~N _1-3 ~M (N3) are similarly calculated by the following equation (7) to (9).

Further, the matching processing unit 145 detects an outlier value of a matching error that should be excluded in calculation of a matching score, which will be described later, for each type of label region. For example, for the nuclear labels N _{1-1 to} N _1-3 , the average A (N) and standard deviation σ (N) of the matching errors M (N1) to M (N3) are expressed by the following equations (10) and (10): 11), a matching error that does not fall within the range of average A (N) ± standard deviation σ (N) is defined as an outlier.

In this case, a matching error M (Nj) that satisfies the following expression (12) or (13) is excluded (j = 1 to 3).
M (Nj)> (A (N) + σ (N)) (12)
M (Nj) <(A (N) −σ (N)) (13)

Here, when the matching accuracy of some label regions is extremely small even though the matching accuracy of most label regions is high in the same type of label region, include such label region matching errors. When the calculation is performed, the matching score is lowered. This is because, for example, the nuclear labels N _2-1 and N _2-2 in the section M3-i match well with the nuclear labels N _1-1 and N _1-2 in the template image T shown in FIG. This corresponds to the case where the label corresponding to the nuclear label N _1-3 in the template image T does not exist by chance in the section M3-i. Therefore, as described above, the outlier value for the matching score is calculated by excluding the outlier value of the matching error for each type of label area and performing the calculation based on the average matching error of the label area of that type. Can be reduced.

In subsequent step S155, the matching processing unit 145 calculates a matching score (matching evaluation value) in the section M3-i. For this purpose, the matching processing unit 145 first adds a matching error for each type of label region from the above formulas (5) to (9), thereby matching the matching error M (C) of the entire cytoplasmic label and the entire nuclear label. The matching error M (N) is calculated. At this time, the matching processing unit 145 performs the calculation by excluding the matching error (for example, the matching error M (N3)) detected as the outlier.
M (C) = M (C1) + M (C2)
M (N) = M (N1) + M (N2)

The matching processing unit 145 calculates a matching score M in the section M3-i from the matching errors M (C) and M (N) by the following equation (14).

The matching processing unit 145 performs the process of the loop A on the section in the label image M3, and then selects a section having the maximum matching score M in step S156. The selected section represents the relative position of the label image (template image) T with respect to the label image M3.
Thereafter, the process returns to the main routine.

  In step S16 following step S15, the calculation unit 14 outputs a matching result. In response to this, the control unit 15 displays the matching result on the display unit 12 in a predetermined format. FIG. 18 is a schematic diagram illustrating a display example of matching results. The matching result display screen B2 shown in FIG. 18 has a format in which a matching area display frame b3, a superimposed image display field b4, and an image deformation information display field b5 are added to the sample attribute setting screen B1 shown in FIG. It has become.

  The matching area display frame b3 represents a corresponding area between the sliced specimen images M2a and M2b. For example, when the entire sliced sample image M2b having the larger magnification is used as the template image, a frame indicating an area in the sliced sample image M2a corresponding to the sliced sample image M2b is on the sliced sample image M2a. Is displayed.

  The superimposed image display column b4 is a column in which an image in which corresponding regions are overlapped between the image M2a and the image M2b is displayed.

  The image deformation information display column b5 is a column in which the content of the change is displayed when there is a change in the corresponding areas between the images M2a and M2b. Specifically, information such as image scale information, rotation angle information when any image is rotating, and similarity indicating the likelihood of matching are displayed.

  As described above, according to the first embodiment, the contribution degree that the label region corresponding to each sample constituent element contributes to the matching process is calculated based on the foresight information of the sample constituent element, and the contribution degree and each label Since pattern matching is performed based on the feature amount of the region, sample components that change greatly between sections of the biological tissue sample can be positively excluded from the calculation of the matching score. Therefore, even when the slices are separated between the sliced specimen images to be processed or the stains applied to the sliced specimens are different from each other, highly accurate matching processing can be performed.

  In the first embodiment, the image processing apparatus 10 acquires the image file from the image database 21 stored in the recording apparatus 20, but the means for acquiring the image is not particularly limited, and for example, via a network It is good also as acquiring an image file by. In this case, the configuration of the storage device 20 is not necessary.

In the first embodiment, the calculation unit 14 acquires the foresight information from the storage unit 13, but the means for acquiring the foresight information is not particularly limited. For example, the foresight information may be attached to the image file, and the foresight information may be acquired at the same time as the image file is read, or the foresight information is acquired by performing a search via the network based on the attribute information of the image. It is also good.
In the present embodiment, the alignment process for two images has been described as an example. However, two or more images can be processed at a time.

(Embodiment 2)
Next, a second embodiment of the present invention will be described.
FIG. 19 is a diagram illustrating a configuration of a microscope system according to the second embodiment. As shown in FIG. 19, the microscope system 1 according to the second embodiment generates a microscopic device 30 that can be observed by a plurality of spectroscopic methods, and generates and outputs image data by capturing an observation image in the microscopic device 30. An image capturing unit 40, an image processing device 10-2 that processes image data output from the image capturing unit 40, and a microscope control unit 50 that controls the operation of the microscope device 30 under the control of the image processing device 10-2. Is provided.

The microscope apparatus 30 includes a substantially C-shaped arm part 31, an epi-illumination light source 32 and an epi-illumination optical system 32a provided on the arm part 31, a trans-illumination light source 33 and a trans-illumination optical system 33a, and an arm part. An objective lens unit 35 including an objective lens 351 disposed on the observation optical path L ₀ so as to face the specimen stage 34, and a cube unit 36 provided on the observation optical path L _0. A trinocular tube unit 37 provided on the observation optical path L ₀ , an eyepiece unit 38 provided via the trinocular tube unit 37, and an imaging lens unit 39 connected to the trinocular tube unit 37; Have An imaging unit 40 is provided at the end of the imaging lens unit 39.

Epi-illumination optical system 32a includes various optical members for guiding incident illumination light emitted from the epi-illumination light source 32 in the direction of the observation optical path L ₀ by condensing (filter unit, a shutter, a field stop, an aperture stop, etc.) . On the other hand, transmission illumination optical system 33a is various optical members (a collector lens for guiding the direction of the observation optical path L ₀ the transmitted illumination light emitted from the transmissive illumination light source 33 is focused, the filter unit, field stop, shutter, aperture Aperture, condenser optical element unit, top lens unit, etc.).

On the specimen stage 34, a specimen (sliced specimen) S to be observed is placed. The specimen stage 34 is driven by a driving device (not shown) and moves two-dimensionally in a plane orthogonal to the observation optical path L ₀ , thereby changing the region of the specimen S with respect to the field of view of the objective lens 351.

  The specimen stage 34 is provided with a motor drive unit (not shown) that operates under the microscope and the microscope control unit 50, and controls the position of the specimen stage 34 from the image processing apparatus 10-2 via the microscope control unit 50. be able to. The sample stage 34 is provided with an encoder (not shown). By inputting an output signal of the encoder to the image processing apparatus 10-2 via the microscope control unit 50, the sample stage is set in the image processing apparatus 10-2. The position of 34 can be grasped.

The objective lens unit 35 includes a plurality of objective lenses 351 and 352 having different magnifications, and a revolver 353 that holds these objective lenses 351 and 352. The magnification of the microscope image can be changed by rotating the revolver 353 and switching the objective lenses 351 and 352 disposed at positions facing the sample stage 34 on the observation optical path L ₀ . FIG. 19 shows a state where the objective lens 351 is arranged on the observation optical path L ₀ . The revolver 353 is provided with an encoder (not shown) and a sensor capable of individually detecting the objective lenses 351 and 352, and an output signal from the encoder and sensor is sent to the image processing apparatus 10-2 via the microscope control unit 50. by inputting, in the image processing device 10-2 can grasp what magnification of the objective lens current is provided in the observation light path L ₀ in.

The cube unit 36 includes a plurality of optical cubes 361 and 362 and a cube switching unit 363 that holds the optical cubes 361 and 362 in a switchable manner, and is arranged on the observation optical path L ₀ according to various spectroscopic methods. The optical cubes 361 and 362 are switched.

The trinocular tube unit 37 branches the observation light (transmitted light or fluorescent light) of the sample S incident from the objective lens 351 into the direction of the eyepiece unit 38 and the direction of the imaging lens unit 39.
The eyepiece unit 38 is used when the user directly observes the specimen S.

  The imaging lens unit 39 is provided with a zoom unit including a plurality of zoom lenses and a drive unit (not shown) that changes the positions of these zoom lenses. The zoom unit enlarges or reduces the imaging target in the imaging field by adjusting the position of the zoom lens.

  The imaging unit 40 includes an imaging element such as a CCD, for example, and is configured by a multiband camera that can capture color images having pixel levels (pixel values) in a plurality of different wavelength bands (bands) in each pixel.

  The imaging unit 40 includes a light receiving surface 40a that receives light (observation light) emitted from the objective lens 351 and incident through the imaging lens unit 39, and generates image data corresponding to the observation light to generate an image. The data is output to the processing device 10-2.

  The image processing apparatus 10-2 has a configuration further including an image acquisition unit 16 with respect to the image processing apparatus 10 shown in FIG. The image acquisition unit 16 is an interface that receives input of image data output from the imaging unit 40. The configuration and operation of the image processing apparatus 10-2 other than the image acquisition unit 16 are the same as those in the first embodiment.

  The microscope control unit 50 controls operations of the specimen stage 34 and the zoom unit in the imaging lens unit 39 in the microscope apparatus 30 according to various control signals output from the control unit 15. In addition, the microscope control unit 50 receives an output signal output from each unit of the microscope apparatus 30 (such as an encoder provided in the specimen stage 34 or the revolver 353), and outputs the output signal to the image processing apparatus 10-2. The image processing apparatus 10-2 can recognize the state of the microscope apparatus 30 such as the position of the sample stage 34 and the magnification of the objective lens in use from these output signals.

Next, a microscope observation method according to the second embodiment will be described. FIG. 20 is a flowchart showing the operation of the microscope system 1 according to the second embodiment.
First, prior to image processing, the user places the sliced specimen S collected from the biological specimen to be analyzed on the specimen stage 34 of the microscope apparatus 30 and rotates the revolver 353 to obtain a low-magnification objective lens. Set. Here, the low magnification means that when a sliced specimen image is acquired, the image is usually captured at a magnification of about 20 to 60 times, whereas the magnification is about 4 times, for example. Further, the user selects an image file of a sliced specimen collected from the same biological tissue specimen as the sliced specimen S from the image files displayed on the display unit 12 using an input device such as a mouse. The input unit 11 receives an image file selection signal and inputs it to the control unit 15.

  In step S <b> 20, the control unit 15 acquires the selected image file from the image database 21 and temporarily stores it in the storage unit 13. FIG. 21A is a schematic diagram illustrating an example of a sliced sample image reproduced from a selected image file.

  In subsequent step S21, the microscope system 1 images the sliced specimen S at a low magnification, and generates image data and incidental information of the sliced specimen image. The generated image data and incidental information are input to the image processing apparatus 10-2 via the image acquisition unit 16. As a result, a sliced specimen image in which the sliced specimen S is shown as a whole is acquired. FIG. 21B is a schematic diagram illustrating an example of a low-magnification sliced sample image that is newly captured.

In step S <b> 22, the specimen attribute setting unit 141 sets the attributes of the sliced specimen based on the image file acquired from the image database 21, newly generated image data, and incidental information.
The subsequent steps S12 to S15 are the same as those in the first embodiment.

  When the user rotates the revolver 353 after the matching process in step S15 is finished and changes the imaging magnification of the microscope apparatus 30 to a high magnification (for example, about 20 to 60 times), the image processing apparatus 10-2 Based on the matching result, an imaging region of the sliced specimen S corresponding to the imaging magnification is set (step S23). For example, when the image processing apparatus 10-2 determines that the matching similarity of the section M7 in the low-magnification image M6 is high with respect to the image M5, the region in the sliced sample S corresponding to the section M7 is the objective lens 351. A control signal for driving the specimen stage 34 is output so as to be in the imaging field of view.

  In step S <b> 24, the microscope system 1 generates image data and incidental information of a sliced sample image representing detailed information of the section M <b> 7 by imaging the set imaging region at a high magnification. The generated image file including the image data and the accompanying information is taken into the image processing apparatus 10-2 and displayed on the display unit 12, or appropriately subjected to image processing and stored in the image database 21.

  As described above, according to the second embodiment, the user can acquire an image of a region corresponding to the selected sliced specimen image in real time by imaging another sliced specimen. Become. Therefore, if the user wants to observe the area near or related to the sliced specimen image of interest in detail with a further enlarged image, or using a different staining or another microscopic method, Even when it is desired to observe, a desired image can be easily acquired.

  In the second embodiment, the user manually changes the magnification of the objective lens. However, the revolver 353 that holds the objective lenses 351 and 352 is provided with a rotation driving unit, and the operation of the rotation driving unit is controlled by a microscope. The imaging magnification in the microscope apparatus 30 may be automatically changed by controlling the image processing apparatus 10-2 via the unit 50. Alternatively, by controlling the operation of the zoom unit provided in the imaging lens unit 39 on the side of the image processing device 10-2 via the microscope control unit 50, the imaging magnification in the microscope device 30 is automatically changed. Also good.

  Embodiments 1 and 2 described above are not limited as they are, and various inventions can be formed by appropriately combining a plurality of components disclosed in the embodiments and modifications. . For example, some components may be excluded from all the components shown in the embodiment. Or you may form combining the component shown in different embodiment suitably.

DESCRIPTION OF SYMBOLS 1 Microscope system 10, 10-2 Image processing apparatus 11 Input part 111 Image file designation | designated part 12 Display part 13 Storage part 131 Image processing program 132 Foresight information storage part 14 Calculation part 141 Specimen attribute setting part 142 Foresight information acquisition part 143 Specimen structure Element separation unit 144 Contribution rate calculation unit 145 Matching processing unit 15 Control unit 16 Image acquisition unit 20 Recording device 21 Image database 30 Microscope device 31 Arm unit 32 Epi-illumination light source 32a Epi-illumination optical system 33 Trans-illumination light source 33a Trans-illumination optical System 34 Specimen stage 35 Objective lens unit 351, 352 Objective lens 353 Revolver 36 Cube unit 361, 362 Optical cube 363 Cube switching unit 37 Trinocular tube unit 38 Eyepiece unit 39 Imaging lens unit 4 Imaging unit 40a receiving surface 50 microscope control unit

Claims (11)

Based on the first look-ahead information on the sample components of the sample, the sample components shown in each image are separated and extracted from two images obtained by imaging two sections taken from one sample. A specimen component separation unit to perform,
Based on the second foresight information related to the sample component, a contribution calculation unit that calculates a contribution that the sample component extracted and contributed to the alignment process;
An alignment processing unit that specifies a relative position between the two images by calculating an evaluation value based on the contribution;
An image processing apparatus comprising: The first foreseeing information is information representing a color distribution characteristic of the sample component,
The image processing apparatus according to claim 1, wherein the second look-ahead information is information representing a feature of a shape of the specimen component. An attribute information setting unit for setting information on attributes of the pathological specimen;
A foresight information acquisition unit that acquires the first and second foresight information based on information about the attribute;
The image processing apparatus according to claim 1, further comprising: Information on the attribute of the pathological specimen is information representing a part in the living body where the pathological specimen is collected,
The said foresight information acquisition part acquires the said 1st and 2nd foresight information via the information which linked | related the site | part in a biological body and the sample component contained in this site | part. An image processing apparatus according to 1.   The sample component separation unit calculates a color feature amount of a plurality of pixels constituting each image, and based on the color feature amount and the first look-ahead information, the plurality of pixels are converted into sample component elements. The image processing apparatus according to claim 1, wherein a corresponding pixel group is extracted.   The contribution calculation unit refers to the second look-ahead information, acquires the shape feature of the sample component separated and extracted by the sample component separation unit, and based on the shape feature, the contribution degree The image processing apparatus according to claim 1, wherein: The sample component separation unit extracts a group of pixels corresponding to the sample component from a plurality of pixels constituting each image and performs a labeling process, whereby a label image including a label region corresponding to the sample component Produces
The image processing apparatus according to claim 1, wherein the alignment processing unit performs alignment between the label images corresponding to the two images, respectively.   The image according to any one of claims 1 to 7, wherein the specimen component includes at least one of a main tissue including elastic fibers, a cell membrane, a cytoplasm, a red blood cell, a cell nucleus, and a cavity. Processing equipment. The image processing apparatus according to any one of claims 1 to 8,
A stage on which the sliced specimen is placed;
An illumination optical system that emits illumination light toward the stage;
An objective optical system provided facing the stage;
Moving means for changing the position of the stage relative to the imaging field of view of the objective optical system;
An imaging unit that images the observation light transmitted through the objective optical system and generates image information;
With
The microscope system further includes an image acquisition unit that acquires image information generated by the imaging unit. The image processing device is generated when the imaging unit captures the first image acquired in advance and the sliced specimen placed on the stage, and is acquired via the image acquisition unit. Perform alignment processing with the second image corresponding to the image information,
The microscope system according to claim 9, further comprising a control unit that changes the imaging field of view by controlling the moving unit according to a processing result of the alignment process. Based on the first look-ahead information on the sample components of the sample, the sample components shown in each image are separated and extracted from two images obtained by imaging two sections taken from one sample. A sample component separating step,
A contribution calculation step of calculating a contribution that the sample component separated and extracted contributes to the alignment process based on the second foresight information regarding the sample component;
An alignment processing step for specifying a relative position between the two images by calculating an evaluation value based on the contribution;
An image processing method comprising:

Images