A SYSTEM AND A METHOD FOR INTERPRETABLE LEUKEMIC DIAGNOSTICS FROM MICROSCOPY IMAGES

Disclosed is a system (100) for automated and interpretable cellular diagnostics from a microscopy image. The system (100)
includes a processor (132) and a memory (134) communicatively coupled to the processor (132), such that the memory (134)
stores instructions to receive a microscopy image of a biological cell. The system (100) is configured to generate a binary
representation of the biological cell and transform the binary representation into a geometric field representation comprising a
signed distance field tensor. The system (100) further derives a plurality of biophysical parameters, including cortical stiffness,
nuclear stiffness, cytoplasmic viscosity, and internal pressure, using an inverse mechanics solver module (110) in conjunction
with a physics-informed forward model unit (108). The system (100) extracts morphological features using a morphological
transformer network module (114), generates a feature vector, and combines the biophysical parameters and the feature vector to
form a fused multimodal representation for classification.