Prediction of epidermal growth factor receptors mutation in lung adenocarcinoma using clinical-morphology and CT-image features
نویسندگان
doi
10.48309/jmpcr.2026.574237.2001چکیده
Tyrosine kinase inhibitors (TKIs) improve 5-year progression-free survival in lung adenocarcinoma patients with EGFR mutation. However, detecting these mutations requires biopsy, which is invasive and may yield insufficient samples. CT scans remain the primary imaging modality, offering detailed tumor characterization through radiologist interpretation and pixel-based analysis. This study aimed to develop a non-invasive predictive model for EGFR mutation using combined clinical-morphological and CT image features. A retrospective cohort of 179 lung adenocarcinoma patients (2018–2023) from Saiful Anwar General Hospital, East Java, Indonesia, was analyzed. Clinical data included age and sex, while CT-based morphological features covered tumor size, ground-glass opacity, interstitial thickening, margins, air bronchogram, necrosis, emphysema, nodules, lymph node characteristics, metastasis, and pleural involvement. Standard thoracic CT images were also used for feature extraction. The combined model (clinical-morphology + imaging) outperformed individual feature sets, achieving a sensitivity of 54.76%, specificity of 82.24%, and AUC of 0.76 (95% CI: 0.69–0.83). In conclusion, integrating clinical-morphological and CT image features improves the non-invasive prediction of EGFR mutation in lung adenocarcinoma compared to single-feature approaches.