AUTOMATED PREDICTION OF DOSE DISTRIBUTION IN PROTON THERAPY USING ARTIFICIAL INTELLIGENCE TECHNIQUES

Abstract

ONLINE ART REPRESENTS THE FUTURE OF RADIATION ONCOLOGY, AS IT WILL ALLOW EACH PATIENT¿S TREATMENT TO BE ADAPTED TO THEIR ANATOMY AT THE TIME OF THE TREATMENT, REDUCING RADIATION INDUCED SIDE EFFECTS ON OARS AND INCREASING THE EFFICACY ON THE TUMOR. HOWEVER, ACHIEVING THIS REQUIRES FASTER CLINICAL WORKFLOWS FOR RE-PLANNING IN REAL TIME. DOSE CALCULATION IS A MAJOR CHALLENGE IN THIS PROCESS AS TRADITIONAL MC COMPUTATION TIMES ARE TOO LONG FOR AN ONLINE APPROACH. AI HAS EMERGED AS A POWERFUL TOOL TO SIGNIFICANTLY REDUCE THESE COMPUTATION TIMES WHILE REACHING A HIGH PERFORMANCE. FOR THIS REASON, THE OVERALL OBJECTIVE OF THIS WORK IS THE DEVELOPMENT OF A NEURAL NETWORK-BASED MODEL THAT PREDICTS THE DOSE DISTRIBUTION IN PT FROM THE CT IMAGE AND OPTIONALLY FROM THE CTV AND OAR MASKS, DATA COMMONLY USED BY DOSIMETRISTS AND MEDICAL PHYSICISTS DURING TRADITIONAL TREATMENT PLANNING.