Classification of lung cancer tissue using bioimpedance spectroscopy and fractional-order circuit modeling.

Lung cancer's lethality underscores the need for accurate, real-time detection methods. While bioimpedance spectroscopy (BIS) detects electrical differences between healthy and malignant tissues, prior studies relied on raw impedance values, limiting diagnostic insight. This study introduces a novel framework using fractional-order circuit modeling to extract physiologically relevant features from lung tissue. Ex-vivo BIS measurements (50 kHz-5 MHz) were performed on 328 resected specimens using a tetrapolar probe. Eight circuit models were fitted to the data, including classical Cole models and a newly proposed Parallel Fractional Cole (PFC) model. Although the Double Cole model achieved the best curve-fitting accuracy (mean NRMSE: 1.95%), features from the PFC model enabled superior classification. A sixth-degree polynomial SVM classifier using PFC-derived parameters distinguished tumorous from healthy tissue with 90.00% accuracy, 93.33% sensitivity, 86.67% specificity, and 0.87 AUC. This demonstrates that fractional-order models with biologically aligned topologies not only have high-fitting accuracy but also enhance diagnostic utility. The PFC model's parallel architecture effectively captures the microstructural heterogeneity of lung tumors, offering a pathway to real-time, non-invasive nodule localization during surgery.