Patient-twin for sarcoidosis: Cohort-calibrated sIL-2R thresholding via nature-inspired optimization
Intelligence-Based Medicine, cilt.15, 2026 (Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 15
- Basım Tarihi: 2026
- Doi Numarası: 10.1016/j.ibmed.2026.100411
- Dergi Adı: Intelligence-Based Medicine
- Derginin Tarandığı İndeksler: Scopus, EMBASE, Directory of Open Access Journals
- Anahtar Kelimeler: Digital twin, Gray wolf optimizer, Metaheuristic algorithms, Sarcoidosis diagnosis, sIL-2R cutoff optimization, Youden’s index
- İstanbul Gelişim Üniversitesi Adresli: Evet
Özet
This work frames serum soluble interleukin-2 receptor (sIL-2R) cutoff selection for suspected sarcoidosis as a patient-oriented Digital Twin calibration problem. The twin uses three nature-inspired metaheuristics, Puma Optimizer (PO), Gray Wolf Optimizer (GWO), and Bald Eagle Search (BES), to estimate an interpretable cohort-calibrated diagnostic cutoff by maximizing Youden’s Index. Data from 189 suspected cases were analyzed under a standardized nested cross-validation protocol. Performance was evaluated using Youden’s Index, sensitivity, specificity, convergence iterations, and computation time. GWO achieved the highest Youden’s Index (0.56 ± 0.02) and sensitivity (0.76 ± 0.03), with stable specificity (0.80 ± 0.01). PO required the fewest convergence iterations (25 ± 2) and the lowest computation time (0.10 ± 0.01 s). Statistical testing showed significant differences among optimizers for Youden’s Index, sensitivity, convergence iterations, and runtime. These results suggest that Digital Twin-based calibration can support interpretable sIL-2R cutoff selection, with GWO favored when diagnostic performance is prioritized and PO favored when computational efficiency is prioritized. External validation and multi-center testing are still needed before broad clinical use.