Automatic detection and differentiation of pancreatic cystic lesions in endoscopic ultrasonography

A computer-implemented method capable of automatically detecting and differentiating pancreatic cystic lesions in endoscopic ultrasonography image/videos by classifying the pixels as cystic lesions, both mucinous and serous, comprising selecting an architecture combination and fully training such architecture for predicting cystic lesions with means of output validation and storage capabilities, wherein the method:

• selects a number of subsets of all endoscopic ultrasonography images/videos, each of said subsets considering only images from the same patient;
• selects another subset as validation set, wherein the subset does not overlap chosen images on the previously selected subsets;
• pre-training (8000) each of the chosen subsets with one of a plurality of combinations of image feature extraction components, followed by a subsequent classification neural network component for pixel classification as cystic lesions wherein said pre-training;
• early stops when scores of each of the chosen subsets do not improve over a given number of epochs, namely three;
• evaluates the performance of each of the combinations;
• is repeated on new, different subsets, with another combination of networks and training hyperparameters, wherein such new combination considers a higher number of dense layers if an f1-metric is low and fewer dense layers if the f1-metric suggests overfitting;
• selects (400) the architecture combination that performs best during pre-training;
• fully trains and validates during training (9000) the selected architecture combination using the entire set of endoscopic ultrasound images to obtain an optimized architecture combination;
• predicts (6000) cystic lesions using said optimized architecture combination for classification;
• receives the classification output (270) of the prediction (6000) by an output collect module with means of communication to a third-party capable of performing validation by interpreting the accuracy of the classification output and of correcting a wrong prediction, wherein the third-party comprises at least one of:
• another neural network, any other computational system adapted to perform the validation task or, optionally, a physician expert in endoscopic ultrasound imagery;
• stores the corrected prediction into the storage component.