Mammography based Image Diagnosis & Analysis System for Breast Cancer

Project

A deep learning–based automatic breast cancer recognition system designed to support radiologists and clinicians by reducing their workload and enhancing diagnostic accuracy.

MIDAS is an explainable AI system capable of detecting and highlighting both benign and malignant findings.

It is trained on screening mammographies labelled with American College of Radiology (ACR)’s BI-RADS categories for breast cancer.It automatically classifies mammography images and offers advanced image processing capabilities.

MIDAS can be seamlessly integrated into existing PACS systems or operated via a standalone user interface.

Highlights & Successes

• The first DL study conducted on a huge volume of national imaging data (mammograms with MLO & CC projections)
• The precision of the MIDAS model exceeds radiologists’ performance & shows an upward trend with novel trainings (ROC AUC* value is 0.85 and the PR AUC** value is 0.89 (CI: 95%) in the final decision model.)
• FIRST PRIZE in the “Social Responsibility” category in the “Artificial Intelligence Arf Awards” competition organized by the Türkiye Informatics Summit

Features