Purpose:
Selecting lobectomy versus parenchyma-sparing limited resection for early-stage lung cancer remains challenging because clinicians must balance oncologic adequacy with preservation of pulmonary function. Current guideline-based decisions rely largely on tumor size and clinical judgment, and individualized quantitative assessment is limited. We developed and internally validated a CT radiomics–clinical nomogram to estimate the probability of lobectomy versus limited resection as selected in routine practice.

Material and methods:
We retrospectively enrolled 225 patients with early-stage lung cancer (training cohort, n = 180; test cohort, n = 45). From preoperative CT, 1,834 radiomic features were extracted, and five features were selected using LASSO with 10-fold cross-validation to construct a radiomics signature. Candidate clinical variables available before surgery were assessed using multivariable logistic regression to derive a clinical signature. The radiomics and clinical signatures were then integrated into a logistic regression nomogram. Model performance was evaluated using area under the receiver operating characteristic curve (AUC), calibration, decision curve analysis (DCA), and DeLong’s test.

Results:
In the test cohort, the nomogram achieved an AUC of 0.876, outperforming the radiomics-only model (AUC = 0.806; p = 0.0008) and the clinical-only model (AUC = 0.753; p < 1×10⁻⁹). Calibration showed good agreement between predicted and observed resection selection (Hosmer–Lemeshow p = 0.72). DCA indicated that the nomogram yielded a higher net benefit than the alternative models across threshold probabilities of 0-50%.

Conclusions:
This radiomics-clinical nomogram provides an interpretable estimate of the likelihood of lobectomy versus limited resection and may support preoperative planning and multidisciplinary discussion. External and multicenter validation is warranted.