OICC Press International Journal of Mathematical Modelling & Computations 2228-6233 16 2 2026 06 30 TL–FL–QWOA: A Quantum‑Optimized Transfer‑Federated Framework for Privacy‑Preserving EEG‑Based Emotion Recognition in IoMT 10.57647/ijm2c.2026.1602.14 EN Wasan Abdallah Alawsi Institute of Artificial Intelligence and Social and Advanced Technologies, Isf.C., Islamic Azad University, Isfahan, Iran Mahdi Mosleh Institute of Artificial Intelligence and Social and Advanced Technologies, Isf.C., Islamic Azad University, Isfahan, Iran Hadab Khalid Obayes College of Information Technology, University of Babylon, Babylon , Iraq Mohammad Mosleh Department of Computer Engineering, Dez. C., Islamic Azad University, Dezful, Iran Journal Article 2026 06 30 Accurate emotion recognition from electroencephalogram (EEG) signals faces persistent challenges in modern Internet of Medical Things (IoMT) environments, including non‑independent client data, privacy risks, and unstable convergence during federated aggregation. Existing deep or fuzzy‑ensemble models such as LSTM, GRU, and Sugeno‑integral frameworks often fail to maintain high accuracy under heterogeneous client conditions and limited communication bandwidth. To address these limitations, this study proposes a hybrid optimization‑aware architecture—TL–FL–QWOA (Transfer Learning–Federated Learning–Quantum Whale Optimization Algorithm)—designed for reliable and privacy‑preserving EEG‑based emotion classification across distributed IoMT nodes. In the proposed pipeline, Transfer Learning (TL) accelerates model adaptation through pre‑trained representation reuse, Federated Learning (FL) ensures secure collaboration without centralizing raw EEG data, and Quantum Whale Optimization (QWOA) introduces dynamic probabilistic weight adaptation to stabilize convergence under non‑IID distributions. Comprehensive experiments on GAMEEMO and DEAP datasets verify the superiority of TL–FL–QWOA over existing centralized and federated baselines. The framework achieved up to 94.87 % Accuracy 94.71% F₁ Score on GAMEEMO and 96.24 % Accuracy 95.87 % F₁ Score on DEAP, corresponding to 8–10 % improvements relative to fuzzy ensemble FL and asynchronous FedProx variants. These results confirm that TL–FL–QWOA effectively balances precision, privacy, and learning stability, delivering a scalable foundation for emotion‑aware IoMT systems. Future research will extend its integration toward real‑time, cross‑modal affective computing and adaptive personalization in decentralized healthcare.