Identification of a section of bodily tissue for pathology tests
Embodiments are provided for identification of a section of bodily tissue as either a candidate or a non-candidate for pathology tests. In some embodiments, a system can include a processor that executes computer-executable components stored in memory. The computer-executable components can include a feature composition component that generates a feature vector representing a physical model describing dye dynamics that determines a group of multispectral images of a section of bodily tissue. The computer-executable components also can include a classification component that generates a classification attribute for the section of bodily tissue by applying a classification model to the feature vector. The classification attribute designates the section of bodily tissue as one of biopsy-candidate or non-biopsy-candidate.
1 . A system, comprising:
a processor that executes computer-executable components stored in memory, the computer-executable component comprising:
medical imaging equipment that generates a group of multispectral images of a section of bodily tissue by emitting excitation light onto the bodily tissue and collecting emitted light from a dye substance transported through the bodily tissue;
a feature composition component that generates a feature vector representing a physical model describing on-tissue dye dynamics based on the multispectral images, wherein the feature composition component:
extracts a spatially-resolved time series of intensities of the emitted light over a grid of points in a two-dimensional region corresponding to the section of bodily tissue; and
comprises estimates of parameters of an advection-diffusion equation describing the on-tissue dye dynamics, the parameters including an advection coefficient, an isotropic diffusion coefficient, and a source distribution, the estimates being obtained by inverting the advection-diffusion equation based on the spatially-resolved time series; and
a classification component that generates a classification attribute for the section of bodily tissue by applying a classification model to the feature vector, wherein the classification attribute designates the section of bodily tissue as one of biopsy-candidate or non-biopsy-candidate for pathology testing.
2 . The system of claim 1 , wherein the feature composition component generates a second feature vector including second parameters defining the physical model representing the on-tissue dye dynamics that determines a second time series of imaging data representing a second group of multispectral images of the section of bodily tissue; and
the classification component updates the classification attribute by applying the classification model to the second feature vector.
3 . The system of claim 1 , wherein the classification attribute comprises a label, and wherein the classification component further generates a confidence score for the label, wherein the confidence score represents a probability of the label being accurate.
4 . The system of claim 1 , further comprising a report component that:
determines, using the confidence score, that a reporting criterion is satisfied; and
causes a display device to present the classification attribute.
5 . The system of claim 1 , further comprising an extractor component that generates the time series of imaging data using the group of multispectral images of the section of bodily tissue, the time series of imaging data including spatially-resolved time series of fluorescence intensities in a grid of points in a two-dimensional region relative to a coordinate system at the section of bodily tissue.
6 . The system of claim 5 , wherein the extractor component selects a spatial resolution and determines brightness intensities for respective points in a two-dimensional grid of points within a multispectral image of the section of bodily tissue at a defined time, the two-dimensional grid of points having a pitch defined by the spatial resolution, and wherein determining the brightness intensities comprises averaging observed brightness values within the two-dimensional grid of points in an multispectral image of the section of bodily tissue at a defined time.
7 . The system of claim 5 , further comprising a modeling component that generates estimates of a source distribution in the two-dimensional region of the physical model, the source distribution determining, at least partially, the spatially-resolved time series of fluorescence intensities.
8 . A computer-implemented method, comprising:
generating, by a computer system operatively coupled to a processor, a group of multispectral images of a section of bodily tissue by emitting excitation light onto the bodily tissue and collecting emitted light from a dye substance transported through the bodily tissue;
generating, by the computing system, a feature vector representing a physical model describing on-tissue dye dynamics based on the multispectral images, the generating comprising:
extracting a spatially-resolved time series of intensities of the emitted light over a grid of points in a two-dimensional region corresponding to the section of bodily tissue; and
estimating parameters of an advection-diffusion equation describing the on-tissue dye dynamics, the parameters including an advection coefficient, an isotropic diffusion coefficient, and a source distribution, the estimating comprising inverting the advection-diffusion equation based on the spatially-resolved time series to obtain the feature vector; and
generating, by the computing system, a classification attribute for the section of bodily tissue by applying a classification model to the feature vector, wherein the classification attribute designates the section of bodily tissue as one of biopsy-candidate or non-biopsy-candidate for pathology testing.
9 . The computer-implemented method of claim 8 , further comprising:
generating, by the computing system, a second feature vector representing the on-tissue dye dynamics that determines a second group of multispectral images of the section of bodily tissue; and
updating, by the computing system, the classification attribute by applying the classification model to the second feature vector.
10 . The computer-implemented method of claim 8 , wherein the classification attribute comprises a label, the method further comprising generating a confidence score for the label, wherein the confidence score represents a probability of the label being accurate.
11 . The computer-implemented method of claim 8 , wherein the generating comprises generating a time series of imaging data using the group of multispectral images of the section of bodily tissue, the time series of imaging data including spatially-resolved time series of fluorescence intensities over a grid of points in a two-dimensional region relative to a coordinate system fixed to the section of bodily tissue.
12 . The computer-implemented method of claim 11 , wherein generating the time series comprises:
selecting a spatial resolution; and
determining brightness intensities for respective points in a two-dimensional grid of points within a multispectral image of the section of bodily tissue at a defined time, the two-dimensional grid of points having a pitch defined by the spatial resolution, wherein determining the brightness intensities comprises averaging observed brightness values within the two-dimensional grid of points.
13 . The computer-implemented method of claim 11 , further comprising generating estimates of a source distribution in the two-dimensional region of the physical model, the source distribution determining, at least partially, the spatially-resolved time series of fluorescence intensities.
14 . A system, comprising:
a processor that executes computer-executable components stored in memory, the computer-executable component comprising:
medical imaging equipment that generates a first group of groups of multispectral images of a section of bodily tissue by emitting excitation light onto the bodily tissue and collecting emitted light from a dye substance transported through the bodily tissue;
an ingestion component that receives multiple feature vectors representing respective physical models describing on-tissue dye dynamics, wherein a first feature vector of the multiple feature vectors corresponds to a first group of multispectral images and is generated by:
extracting a spatially-resolved time series of intensities of the emitted light over a grid of points in a two-dimensional region corresponding to the section of bodily tissue; and
estimating parameters of an advection-diffusion equation describing the on-tissue dye dynamics, the parameters including an advection coefficient, an isotropic diffusion coefficient, and a source distribution, the estimating comprising inverting the advection-diffusion equation based on the spatially-resolved time series to obtain the first feature vector, wherein each feature vector of the multiple feature vectors is obtained by performing the extracting and estimating operations on a respective group of the groups of multispectral images; and
a constructor component that trains, using the multiple feature vectors, a classification model to classify a particular group of the groups of multispectral images as pertaining to one of a biopsy-candidate category or a non-biopsy-candidate category.
15 . The system of claim 14 , wherein the ingestion component receives user profile data indicative of a medical history of a subject corresponding to the section of bodily tissue; and
wherein the constructor component trains, using the multiple feature vectors and the user profile data, a second classification model to classify the particular group as pertaining to one of the biopsy-candidate category or the non-biopsy-candidate category.
16 . The system of claim 14 , wherein the physical model is defined, at least partially, by an advection-diffusion equation in two-dimensional space, and wherein the receiving the multiple feature vectors comprises receiving a first parameter defining an advection vector-field coefficient, a second parameter defining an isotropic diffusion coefficient, and a third parameter defining a source of the advection-diffusion equation.
17 . A computer-implemented method, comprising:
receiving, by a computing system operatively coupled to a processor, multiple feature vectors representing respective physical models describing on-tissue dye dynamics, wherein a first feature vector of the multiple feature vectors corresponds to a first group of groups of multispectral images of a section of bodily tissue generated by emitting excitation light onto the bodily tissue and collecting emitted light from a dye substance transported through the bodily tissue, and wherein generating the first feature vector comprises:
extracting a spatially-resolved time series of intensities of the emitted light over a grid of points in a two-dimensional region corresponding to the section of bodily tissue; and
estimating parameters of an advection-diffusion equation describing the on-tissue dye dynamics, the parameters including an advection coefficient, an isotropic diffusion coefficient, and a source distribution, the estimating comprising inverting the advection-diffusion equation based on the spatially-resolved time series to obtain the first feature vector,
and wherein each feature vector of the multiple feature vectors is obtained by performing the extracting and estimating operations on a respective group of the groups of multispectral images; and
training, by the computing system and using the multiple feature vectors, a classification model to classify a particular group of multispectral images as pertaining to one of a biopsy-candidate category or a non-biopsy-candidate category.
18 . The computer-implemented method of claim 17 , further comprising:
receiving, by the computing system, user profile data indicative of a medical history of a subject corresponding to the section of bodily tissue;
training, by the computing system, using the multiple feature vectors and the user profile data, a second classification model to classify the particular group as pertaining to one of the biopsy-candidate category or the non-biopsy-candidate category.
19 . The computer-implemented method of claim 17 , wherein the physical model is defined, at least partially, by an advection-diffusion equation in two-dimensional space, and wherein the receiving the multiple feature vectors comprises receiving a first parameter defining an advection vector-field coefficient, a second parameter defining an isotropic diffusion coefficient, and a third parameter defining a source of the advection-diffusion equation.
20 . A computer program product for identification of a section of a bodily tissue as either a candidate or a non-candidate for a pathology test, the computer program product comprising a computer-readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to:
generate, by the processor, a group of multispectral images of the section of bodily tissue by emitting excitation light onto the bodily tissue and collecting emitted light from a dye substance transported through the bodily tissue;
generate, by the processor, a feature vector representing a physical model describing on-tissue dye dynamics based on the multispectral images, the generating comprising:
extracting a spatially-resolved time series of intensities of the emitted light over a grid of points in a two-dimensional region corresponding to the section of bodily tissue; and
estimating parameters of an advection-diffusion equation describing the on-tissue dye dynamics, the parameters including an advection coefficient, an isotropic diffusion coefficient, and a source distribution, the estimating comprising inverting the advection-diffusion equation based on the spatially-resolved time series to obtain the feature vector; and
generate, by the processor, a classification attribute for the section of bodily tissue by applying a classification model to the feature vector, wherein the classification attribute designates the section of bodily tissue as one of biopsy-candidate or non-biopsy-candidate.
21 . The computer program product of claim 20 , wherein the program instructions are further executable by the processor to cause the processor to:
generate, by the processor, a second feature vector representing the physical model describing the dye dynamics that determines a second group of multispectral images of the section of bodily tissue; and
update, by the processor, the classification attribute by applying the classification model to the second feature vector.
22 . The computer program product of claim 20 , wherein the classification attribute comprises a label, and wherein the program instructions are further executable by the processor to cause the processor to generate a confidence score for the label, wherein the confidence score represents a probability of the label being accurate.
23 . The computer program product of claim 22 , wherein the program instructions are further executable by the processor to cause the processor to:
determine, by the processor, using the confidence score, that a reporting criterion is satisfied; and
cause, by the processor, a display device to present the classification attribute.
24 . The computer program product of claim 20 , wherein the program instructions are further executable by the processor to cause the processor to generate a time series of imaging data using the group of multispectral images of the section of bodily tissue, the time series of imaging data including spatially-resolved time series of fluorescence intensities over a grid of points in a two-dimensional region relative to a coordinate system fixed to the section of bodily tissue.
25 . The computer program product of claim 24 , wherein the program instructions are further executable by the processor to cause the processor to generate estimates of a source distribution in the two-dimensional region of the physical model, the source distribution determining, at least partially, the spatially-resolved time series of fluorescence intensities.