MRI-based intratumoral and peritumoral radiomics predicting neoadjuvant chemotherapy response in osteosarcoma

Objectives: To investigate the predictive value of a nomogram prediction model based on MRI intratumoral and peritumoral radiomics combined with clinical factors for neoadjuvant chemotherapy(NAC) in osteosarcoma(OS) patients. Methods : This retrospective study included 93 patients with surgically confirmed OS who underwent NAC. Lesions were manually delineated on axial T2-weighted fat-suppressed (T2WI-FS) images using ITK-SNAP to define the intratumoral region. Peritumoral regions were semi-automatically segmented with 2 mm, 4 mm, and 6 mm expansions. Random forest was used to construct intratumoral, peritumoral (2 mm, 4 mm, 6 mm), and intratumoral + peritumoral radiomics models. The most effective model was integrated with the clinical model to construct a predictive nomogram. Model performance was evaluated using the area under curve (AUC) and F1 score, and clinical utility was assessed via decision curve analysis (DCA). Results: Multivariate analysis showed that ALP (OR = 1.033, 95% CI: 1.000 ~ 1.006, P = 0.031) and pathological fracture (OR = 2.575, 95% CI: 1.036 ~ 6.401, P=0.042) were independent predictors of OS in patients receiving NAC. Among all the radiomics models, the Model_rad-intra+peri ^2mm model had the best efficacy, with AUCs of 0.888 and 0.765 in the training and test sets, respectively. The integrated nomogram achieved AUCs of 0.990 and 0.815 in the training and test sets, respectively, demonstrating strong predictive ability. Conclusion: This study developed a nomogram prediction model integrating intratumoral and peritumoral MRI radiomics with clinical characteristics, which exhibited outstanding predictive performance in both the test and test sets. This model holds promise for assisting clinicians in the early assessment of NAC efficacy in OS patients and is anticipated to inform clinical decision-making.