Free and open-source software for automated cell detection

Details

Supervisors

Jodogne, Sébastien

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil en informatique, à finalité spécialisée

Abstract

enArtificial Intelligence is revolutionizing healthcare. To face the ever-growing amount of medical images, healthcare professionals resort to using computer-aided diagnosis. In the field of pathology, such systems are pricey. Pathology students’ education at academic centers with limited funds may be less engaging due to lack of access to these tools. This thesis aims to explore the feasibility of designing a web application to support learning of digital pathology. The project focuses on a central task in pathology: cell detection. To this effect, an open-source prototype was developed using Python, Flask, OpenLayers and Orthanc. It integrates two cell detection methods: marker-based watershed and StarDist. The performance of the tool was evaluated according to response time and accuracy of detections. The web-based nature and open-source foundation of the application make it particularly suited for educational environments, where cost, accessibility, and transparency are critical. This work discusses the performances of the tool as well as future developments and improvements that can be carried out. It successfully demonstrates that developing digital pathology tools for education can be achieved through an open-source approach. We published the code under the GNU Affero General Public License, and it is available on the following repository: https://github.com/farah-ard/Master_Thesis.