Multidimensional Hesitant Fuzzy Transformer Networks: A Deep Learning Approach for Intelligent Medical Diagnosis

Intelligent medical diagnostic technologies play a crucial role in enhancing the accuracy and efficiency of disease identification and diagnosis. However, their integration into patient-doctor interactions, which involve a complex mix of fuzzy information and quantitative data, remains underexplored. This paper addresses this gap by introducing hesitant fuzzy set (HFS), which are adept at handling subjective inputs, and by developing an intelligent diagnostic framework based on advanced Transformer networks. Specifically, we define the concepts of ascending hesitant fuzzy set (AHFS) and multidimensional hesitant fuzzy set (MHFS), based on HFS, to better capture and refine the dynamic interactions between patients and clinicians. We then introduce an intelligent diagnostic methodology that effectively processes both fuzzy data and real numbers within the multidimensional hesitant fuzzy Transformer (MHF- Transformer) networks. This approach offers notable advantages, including improved interpretability of diagnostic results, enhanced adaptability to complex medical scenarios, and robustness in handling uncertainty and incomplete data. To further enhance the adaptability of this method and reduce the risk of misdiagnosis in intelligent diagnostics, an expert review mechanism is incorporated into the output layer. Additionally, we design a comprehensive optimization algorithm for the MHF-Transformer networks, complete with detailed execution protocols and pseudocode, strengthening the reliability of intelligent diagnostics. Finally, we apply the MHF-Transformer networks to two clinical scenarios: the diagnosis of mental disorders and the recurrence assessment of thyroid cancer. The empirical findings and comparative analyses provide strong evidence of the efficacy, practicality, and superior performance of the proposed intelligent diagnostic method.