BACKGROUND: Computed tomography (CT) is the gold standard for assessing the severity of pectus excavatum; however, it involves high radiation exposure and substantial costs. This study aimed to develop a statistical model based on measurements from posteroanterior and lateral chest radiographs (2D) to derive a novel X-ray Index (XI) for predicting the CT-derived Haller Index, thereby reducing the need for CT examinations in a subset of patients. METHODS: A total of 305 patients with pectus excavatum scheduled for Nuss procedure were retrospectively enrolled. All patients underwent both posteroanterior/lateral chest radiography and CT examination preoperatively, with an interval of no more than 6 months between the two imaging modalities. The Haller Index (HI) was measured on CT images, while the X-ray Index (XI) was assessed on radiographs. The dataset was randomly divided into a training set (n = 216) and a validation set (n = 89) at a 7:3 ratio. A predictive model was developed using LASSO regression, with the optimal penalty parameter selected via 10-fold cross-validation. Model performance was evaluated using R², root mean square error (RMSE), Bootstrap internal validation (with 1,000 resamples), calibration curves, and decision curve analysis (DCA). Using HI > 3.25 as the reference standard for surgical indication, diagnostic performance metrics including area under the curve (AUC), sensitivity, and specificity were calculated. RESULTS: The XI demonstrated a strong correlation with the HI (r > 0.9). The final predictive model was: Predicted HI = 0.1333 + 1.0922 × XI - 0.027 × Age. This model exhibited excellent performance in the validation set, achieving an R² of 0.885, an RMSE of 0.504, and an AUC of 0.966 (95% CI: 0.936-0.996). At the optimal cut-off value of 3.316, the model yielded a sensitivity of 0.904 and a specificity of 0.892. Bootstrap validation confirmed the model's stability, with a mean AUC of 0.958 (95% CI: 0.954-0.960). Calibration was satisfactory, as indicated by a calibration intercept of 0.0083 and a calibration slope of 1.1699. Decision curve analysis demonstrated a positive net clinical benefit for the model across a threshold probability range of 0.1 to 0.6. Subgroup analyses further revealed consistent and excellent predictive performance across different age and sex subgroups, with AUC values ranging from 0.943 to 1.000. CONCLUSION: In conclusion, the LASSO regression model established in this study accurately estimates the CT-derived Haller Index based on measurements from chest radiographs. Demonstrating robust predictive performance and favorable clinical applicability, this model may serve as an effective tool for the screening and follow-up of pectus excavatum, thereby reducing unnecessary exposure to ionizing radiation from CT.
Han et al. (Sun,) studied this question.