The value of constructing a nomogram model based on arterial-phase imaging radiomics features combined with clinical-CT features in differentiating lung squamous carcinoma and adenocarcinoma

Objective: To investigate the value of constructing a nomogram model based on arterial-phase imaging radiomics features combined with clinical-CT features for the differentiation between squamous lung cancer (SCC) and adenocarcinoma of the lung (ADC). Methods: Retrospectively, 85 patients with lung cancer who underwent puncture pathology biopsy or surgery in our hospital, from August 2021 to September 2023, were collected as a training set. Concurrently, 40 patients with pathologically confirmed lung cancer in our hospital, from May 2023 to June 2024, were collected as a validation set. All patients underwent chest CT enhancement. The training set was divided into the SCC group (n=29) and the ADC group (n=56) based on the pathology. General clinical data and CT image characteristics of the two groups of patients were compared and differences were identified. Independent predictors were screened using unifactorial and multifactorial logistic regression analyses, and a clinical-CT model was constructed. ITK Snap software was applied to extract the radiomics features of the arterial-phase images, and the intragroup correlation coefficient (ICC), Roruta feature screening, and least absolute shrinkage and selection operator (LASSO) were sequentially used to downsize the extracted radiomics features, screen out the meaningful features, construct the arterial-phase image radiomics model using Logistic regression, and compute the model's image radiomics score (Rad-score). A multifactor logistic regression analysis was used to screen the independent variables with meaningful clinical-CT characteristics and Rad-score to construct a joint model, and a nomogram graph was plotted. ROC curves, calibration curves, H-L test, Delong test, and clinical decision curves (DCA) were applied to evaluate the clinical-CT, radiomics, and nomogram models. Results: The results of univariate analysis showed that there were more lobular signs and necrotic cavity signs, and fewer carcinoembryonic antigen (CEA), vascular cluster signs, pleural pulling, and burr signs in the SCC than in the ADC group (all P < 0.05). The above independent variables were included in the multifactorial Logistic analysis for further screening, and the results showed that CEA, lobular sign, pleural pull, and burr sign were independent risk factors. The area under the curve (AUC) values for the training and validation sets of the clinical-CT model constructed based on this were 0.623 and 0.786, respectively. A total of eight meaningful features were screened after dimensionality reduction of the radiomics features, which were three first-order features and five second-order features. The ROC curve analysis showed that the AUC values for the training and validation sets of the radiomics model were 0.830 and 0.846, respectively; and the AUC values for the training and validation sets of the nomogram model were 0.913 and 0.922, respectively. The Delong test showed that the AUC values of the nomogram model were all significantly higher than those of the clinical-CT model and the radiomics model (all P < 0.05); the Hosmer-Lemeshow test showed that the clinical-CT model, the radiomics model, and the nomogram model were all well fitted; calibration curve analysis showed that the predictive probability curve of the nomogram model was closest to the ideal curve, with better predictive accuracy; and DCA analysis showed that the AUC of the nomogram model was the largest, with the highest net clinical benefit. Conclusion: Constructing a nomogram model based on arterial-phase imaging radiomics features combined with clinical-CT features has some diagnostic value in differentiating SCC from ADC, providing a new diagnostic modality for noninvasive differentiation.