PocketOnco: A Novel CoreML-Based iOS App for the Diagnosis and Prognosis of Colorectal, Breast, and Skin Cancer Using Multilayered Convolutional Neural Network Algorithms

摘要

Colorectal, breast, and skin cancer are among the most common and deadly diseases in the United States, according the American Cancer Society. While many researchers focus on the diagnosis of such cancers, the current treatment pipeline takes several weeks and requires a large team of pathologists, oncologists, radiologists, and more. Despite the pressing need for a highly accurate and data-driven approach to not only diagnosis cancer, but deliver rapid prognoses, a current solution does not exist. This project presents PocketOnco, a novel mobile app for iOS that uses multilayered convolutional neural network algorithms (CNN) to automatically diagnose and prognose colorectal, breast, and skin cancer within seconds through tumor feature segmentation and prediction. Users can either import histopathological tissue images or take a picture of an external dermoscopic image and select crop for the region of interest.