AI-Powered Framework for Real-Time Detection of Prostate Cancer from MRI Scans

Abstract

Since abnormalities at an early stage of prostate cancer may be hard to spot from manual examination of MRI images, timely and accurate detection is essential. The aim of this thesis was to create an AI-based system for the real-time detection of prostate cancer from MRI images using a deep learning model based on YOLOv11 and a comprehensive clinical workflow. The proposed system will also allow suspicious cancer regions to be detected as well as providing a practical patient-to-physician review process. The framework starts by preparing the MRI images, re-structuring the MRI datasets, annotations handling and image pre-processing for object detection. A model called YOLOv11 is used to detect cancerous areas in prostate MRI images and output bounding boxes and confidence scores. The trained model is deployed using a FastAPI backend, where uploaded MRI images are processed, validated, fed into the inference pipeline and transformed into clinically relevant output like annotated images and data for the diagnostic report. For the system to be more practical, there are a separate patient portal and a separate physician portal. Patients register and login to the system, upload an MRI report, or a physician can log in and view the patient's uploaded MRI report, activate the AI model, review the annotated reports, approve and send the MRI report to the patient. The safe access of users is enabled using Firebase Authentication, users, cases, images, results of the detections and report records are stored using MongoDB and the final report delivery is automated using n8n after the physician has given his approval via email. The framework developed shows how all these technologies deep learning, backend services, authentication, database storage, workflow automation, and clinical review can be integrated into a single AI-powered diagnostic system. Real time detection, explainable visual output, doctor control over when reports are released, and deployment readiness are the key highlights of the project. The system aims to assist radiologists and physicians in minimizing manual workload, optimizing workflow, and creating a structured environment for the AI-driven prostate cancer screening process.