This retrospective cohort study assessed the diagnostic performance of modified thin-slab volume rendering (tsVR) with densitometry-derived thresholds compared to conventional CT feature-based diagnosis. The analysis included 383 patients for model development and 61 patients for external validation, totaling 444 patients with pneumonic-type lung cancer or inflammatory pneumonia. The study setting was not reported, and follow-up duration was not reported.
internal validation results demonstrated an AUC of 0.86, sensitivity of 0.90, specificity of 0.82, and accuracy of 0.86. The external validation cohort yielded an AUC of 0.81, sensitivity of 0.90, specificity of 0.73, and accuracy of 0.82. Both internal and external validations showed improved performance relative to the original CT feature-based diagnosis, though p-values or confidence intervals were not reported.
Secondary outcomes included inter-observer agreement, which was reported as kappa=0.713. Robustness across scanners, scanning doses, and reconstruction protocols was also evaluated. No adverse events, serious adverse events, discontinuations, or tolerability data were reported. Funding or conflicts of interest were not reported.
The study demonstrates clinical applicability and robust stability. However, because the design is observational and key statistical details like p-values are missing, causal claims cannot be made. These results should be interpreted with caution until validated in prospective trials.
View Original Abstract ↓
ObjectivesPneumonic-type lung cancer (PTLC) poses significant diagnostic challenges owing to its overlapping computed tomography (CT) imaging appearance with inflammatory pneumonia. We developed and validated a novel diagnostic model that integrates densitometry-derived thresholds into modified thin-slab volume rendering (tsVR) to differentiate PTLC from inflammatory pneumonia. We further evaluated and compared the diagnostic performance of this tsVR approach against conventional CT feature-based diagnosis.Materials and methodsThis retrospective study enrolled 383 patients (193 PTLC, 190 pneumonia) for model development (training/internal validation cohorts, 7:3 ratio) and 61 patients as external validation cohort. Peri-lesional densitometric analysis identified voxel-level and individual-level CT density thresholds translated into modified tsVR parameters to enhance the peri-lesional visualization. Discriminating capabilities of tsVR were assessed by four radiologists against original CT feature-based diagnosis using the area under curve (AUC), sensitivity, specificity, accuracy, positive predictive value (PPV), negative predictive value (NPV), F1-score, F2-score and Matthews correlation coefficient (MCC), followed by logistical regression analyses assessing independent influencing factors for diagnostic accuracy.ResultsThe tsVR model achieved superior performance over original CT feature-based diagnosis across all metrics, with internal validation AUC 0.86 (sensitivity 0.90, specificity 0.82, accuracy 0.86, F2 score 0.89) and external validation AUC 0.81 (sensitivity 0.90, specificity 0.73, accuracy 0.82, F2 score 0.87). Logistical regression analyses confirmed robustness of tsVR in diagnostic accuracy across scanners, scanning doses and reconstruction protocols, with good inter-observer agreement (κ=0.713).ConclusionThe tsVR significantly improves the diagnostic efficacy in discriminating PTLC from inflammatory pneumonia, demonstrating clinical applicability and robust stability.