As a part of XilinxNexus program initiated under Xilinx WIT awardee 2021, the ML model development and its porting in Xilinx hardware is started.
Breast cancer is more prevalent among Indian women due to poor prognosis at an early stage. Most cancer is diagnosed only at the advanced stage. But our country has only 10,000 practicing radiologists, 2,200 certified pathologists and nearly 10,000 diagnostic laboratories. Therefore, we may need more radiologists and pathologists for screening and diagnosis of cancer disease. Diagnosis procedure is operator dependent and needs an experienced pathologist. The requirement of radiologist, pathologist and trained technicians are less in number for our population size. Moreover, some human factors like exhaustion and insufficient concentration may also cause misdetection.
Motivated by the need for development of a computer assisted dialogistic system for the women folk in early detection of breast cancer and the most prevalent disease, this project intends to provide an effective screening and diagnostic system. It is based on recent development in Deep Learning DL methods / networks namely capsule network and transfer learning.
The developed model has the ability to provide a screening and diagnosis report on digital mammogram and histopathology image. The Xilinx embedded hardware developer kit can be interfaced to a digital mammogram DM unit as well as to a camera unit of microscope where a tissue specimen is kept for investigation. The output of the DL model for an image acquired through the DM unit gives its predictions / finding of Benign (B), Malignant (M).