Multi-scale tumor cell detection and classification

A method for detecting cancer, comprising:

• generating, from a scanned tissue sample, a pair of input images, a first input image representing a part of the scanned tissue sample, and a second input image, centered at the same location as the first, representing a larger part of the sample and resized to a size of the first input image, using a hardware processor;
• locating cells within the first input image using a detection model;
• classifying detected cells in the first input image, using a classification model, to determine a respective cancer probability for each cell, based on cell location information generated by the detection model;
• segmenting the second input image, using a segmentation model, to assign a cancerous-structure probability to arrangements of cells;
• determining an updated cancerous probability of each detected cell in the first pixel patch, using a weighted sum of the cancer probability for the cell and the cancerous-structure probability;
• and
• determining a tumor cell ratio based on the updated cancerous probability of each detected cell in the first pixel patch.

A system for training a machine learning model, comprising:

• a hardware processor;
• and
• a memory that stores a computer program product, which, when executed by the hardware processor, causes the hardware processor to:
• generate pairs of training pixel patches from a dataset of training images, each pair including a first patch representing a part of a respective training image, and a second patch, centered at the same location as the first, representing a larger part of the training image, being resized to a same size of as the first patch, using a hardware processor;
• train a detection model using the first pixel patches in the pairs of training patches, to detect and locate cells in the images;
• train a classification model using the first pixel patches in the pairs of training patches, to classify cells according to whether the detected cells are cancerous, based on cell location information generated by the detection model;
• and
• train a segmentation model using the second pixel patches in the pairs of training patches, to locate and classify cancerous arrangements of cells in the images.