Open Multi-Approach Software Tools for Prostate Carcinoma Prognosis and Diagnosis

At a time where bio-engineering is sufficiently capable of better and more reliable results, oncologists (and pathologists in general) still resort to classic biopsy examinations to build the prognosis or the diagnosis of their patients. These diagnostic assessments lie on a rather subjective observation of cell structure and distribution whereas computational and automated techniques can actually be used since they provide a more objective diagnosis with a quantitative approach and a rigorous training based on previously diagnosed biomedical images. In this work, we aim to develop precise, generic and automated decision making software tools for prostate carcinoma prognosis and diagnosis. We study how image analysis tools and techniques can be used for cell segmentation and cell analysis. We develop our own image processing pipelines with a multi-approach based view and we test them on a set of already diagnosed images. From there, we design tools for learning and decision aimed to help the oncologists in their work of object recognition. The aims of this project are to supply the oncologists with powerful and integrated tools they can actually use in their diagnosis workflow, to provide an open-source basis for further development, refinement and improvement and to centralise inputs for this area of knowledge.