Docker-Based Web Service to Aid Prostate Cancer Diagnosis Using Deep-Learning Techniques

Docker-Based Web Service to Aid Prostate Cancer Diagnosis Using Deep-Learning Techniques

연구 분야: Software Development

학회: International Conference on Simulation Tools and Techniques

Artificial intelligence and deep learning are increasingly developed and better integrated into the medical environment to facilitate and assist in the most routine and time-consuming tasks for doctors. Image classification with convolutional neural networks has demonstrated a high degree of maturity and very high effectiveness rates. In this work, we present a solution for doctors to use these techniques in an efficient and simple way. Focusing on the detection of prostate cancer by classifying biopsied tissue as cancerous or not, and providing the stage of the disease according to the Gleason scale, the work focuses on the creation of a web-based service that executes the requested algorithm on a machine specifically configured to minimize execution times. This machine uses containers to facilitate replication of the solution to new machines, or replicas of it and uses a Django database manager to manage users, images to be analyzed and diagnostics obtained. So that users can consult these solutions when required. These results are displayed as heat maps overlaid on the original image using the open-source QuPath tool. Our solution, which also includes execution on a hardware system specifically designed to optimize system performance, can be used not only by experienced medical pathologists, but also by junior doctors who are in the process of learning the subject.