Whole tissue classifier for histology biopsy slides
View Patent ↗Disclosed is a computer implemented method for fully automated tissue diagnosis that trains a region of interest (ROI) classifier in a supervised manner, wherein labels are given only at a tissue level, the training using a multiple-instance learning variant of backpropagation, and trains a tissue classifier that uses the output of the ROI classifier. For a given tissue, the method finds ROIs, extracts feature vectors in each ROI, applies the ROI classifier to each feature vector thereby obtaining a set of probabilities, provides the probabilities to the tissue classifier and outputs a final diagnosis for the whole tissue.
1. A computer-implemented method of whole tissue classification steps of:
training a Multi-Layer Perceptron (MLP) classifier in a supervised manner wherein labels are given only at a tissue level, the training using a multiple-instance learning variant of backpropagation, wherein an input feature vector that generates the largest output value within all regions of interest (ROI) is back-propagated;
training a tissue classifier with an output of the MLP classifier;
for a given tissue image:
segmenting the tissue image into ROIs;
extracting a vector of features from each of the ROIs;
applying the MLP classifier to the vector of features of each ROI thereby obtaining a set of probabilities;
providing the probabilities to a tissue classifier; and
outputting a diagnosis of the whole-tissue.
2. The method of claim 1 wherein tissue classifier comprises a support vector machine (SVM) which receives as input a histogram of individual ROI probabilities.
3. The method of claim 1 wherein the tissue classifier comprises a support vector regression (SVR) which receives as input a histogram of the ROI probabilities and outputs a tissue histological grade.
4. The method of claim 1 , wherein the probabilities obtained from the MLP classifier are provided to a downstream system.
5. A system for performing whole-tissue classification, said system comprising a computing device including a processor and a memory coupled to said processor, said memory having stored thereon computer executable instructions that upon execution by the processor cause the system to:
train a Multi-Layer Perception (MLP) classifier in a supervised manner wherein labels are given only at a tissue level, the training using a multiple-instance learning variant of backpropagation, wherein an input feature vector that generates the largest output value within all regions of interest (ROI) is back-propagated;
train a tissue classifier with an output of the MLP classifier;
for a given tissue image:
segment the tissue image into ROIs
extract a vector of features from each of the ROIs;
apply the MLP classifier to the vector of features of each ROI thereby obtaining a set of probabilities;
provide the probabilities to a tissue classifier; and
output a diagnosis of the whole-tissue.
6. The system of claim 5 wherein tissue classifier comprises a support vector machine (SVM) which receives as input a histogram of individual ROI probabilities.
7. The system of claim 5 wherein the tissue classifier comprises a support vector regression (SVR) which receives as input a histogram of the ROI probabilities and outputs a tissue histological grade.
8. The system of claim 5 , wherein the probabilities obtained from the MLP classifier are provided to a downstream system.