Evaluation of a Prototype Image- Medical Decision Support System Based- Convolutional Neural Network

International audience In the realm of image-assisted tumor diagnosis, the collaboration between human and computer interactions has reached a sophisticated level of maturity. Inspection propelled by improvements in deep learning presents initial opportunities for a more comprehensive, automated, and secure decision-making process. Nonetheless, the ethical implications of employing CNN as a newly developed technology in healthcare decisions remain inadequately substantiated, resulting in the Image Decision Support System (IMDSS) based on CNN technologies not completely achieving human-computer interactions in practice as an image-aided diagnostic system. This paper summaries the purposes and elucidates the principles of CNN in IMDSS, analyses the challenges of CNN in medical decision-making, and offers a reference framework for potential uses of CNN in image-based decisions, addressing this limitation by designing, implementing, and evaluating a comprehensive deep-learning-facilitated image-based decision support system. We created an intelligent healthcare system for supported assessment and decision-making based on a convolutional neural network (CNN). This technology evaluates patients’ medical information to aid in diagnosis stage and offers suggested treatment regimens to doctors. This study utilized data from the medical histories of 250,000 individuals, collected from three medical centers in Tunisia during a five-year span for instruction and evaluation of the system. Upon reaching 6000 case instances, the algorithm attained a precision rate of 0.87, approaching the physician’s precision of 0.88. The trial results demonstrated that the algorithm can rapidly and precisely analyses patient data, offering decision guidance for doctors.