Morphometric genotyping of cells in liquid biopsy using optical tomography

A method for stepwise isolation of a plurality of cancer mutation drivers, the method comprising:

• providing a plurality of 3D reconstruction images to a first morphological classifier, where the 3D reconstruction images represent a plurality of cell types;
• operating the first morphological classifier to isolate the plurality of cell types into normal and malignant cell types or dysplastic cell types;
• operating a second morphological classifier on the malignant cell types to isolate small cell lung cancer cells from adenocarcinoma and squamous cancer cells;
• operating a third morphological classifier on the malignant cell types to isolate squamous cancer cells from adenocarcinoma cells;
• and
• operating a fourth morphological classifier on the malignant cell types to further isolate a mutation subtype based on morphological factors.

A classification training system for training classifiers adapted to identify specific mutations associated with cancer, the system comprising:

• a control system including a processor and a memory, the memory arranged to store computer instructions that are executable by the processor, and cause the control system to:
• identify 3D image feature data from a plurality of first cells;
• generate a first set of 3D cell imaging data from the plurality of first cells and from a plurality of other malignant cells, where the first set of cell imaging data includes a plurality of first individual cell images;
• generate a second set of 3D cell imaging data from a set of normal cells where the plurality of driver mutations is expected to occur, where the second set of cell imaging data includes a plurality of second individual cell images;
• operate supervised learning based on cell line status as ground truth;
• and
• generate a classifier from the supervised learning.