Graph Neural Network based multi-criteria group decision-making for optimizing radiation therapy plans in a Pythagorean fuzzy environment

Radiation therapy treatment planning involves complex decision-making between optimal tumor coverage and normal tissue sparing. The paper suggests a theoretical framework of a Multi-Criteria Group Decision-Making (MCGDM) model based on Graph Neural Networks (GNNs) and Pythagorean Fuzzy Sets (PFS) for facilitating optimal treatment planning. The model learns interdependencies between clinical criteria using GNNs and handles uncertainty in expert judgments within the PFS framework. A Three-Way Decision-Making (TWD) framework is integrated to decide on treatment options to be classified as Accept, Defer, or Reject based on computed performance scores. The theoretical formulation demonstrates clear improvements in transparency, flexibility and interpretability. The suggested model offers a good foundation for the development of intelligent, expert-based decision support in radiation therapy and other complex medical domains.