Computed-tomography-based radiomic nomogram for predicting the risk of indeterminate small (5–20 mm) solid pulmonary nodules

Abstract

This is a PDF file of a peer-reviewed, preliminarily formatted and unedited paper that has been accepted for publication in Diagnostic and Interventional Radiology. Copyediting of the text and figures and proof review of the the paper will be finished before the paper is published in its final form. Please note that errors may be discovered which could affect the content of the paper during the production process. All legal disclaimers apply. u n c o r r e c t e d p r o o f ABSTRACT PURPOSE: This study aimed to develop a diagnostic model combining computed tomography (CT) images and radiomic features to differentiate indeterminate small (5-20 mm) solid pulmonary nodules (SSPNs).METHODS: This study retrospectively enrolled 413 patients with SSPNs surgically removed and histologically confirmed from 2017 to 2019. The SSPNs included solid malignant pulmonary nodules (n=210) and benign pulmonary nodules (n=203). The least absolute shrinkage and selection operator (LASSO) was used for radiomic feature selection, and random forest (RF) algorithms were used for radiomic model construction. The clinical model and nomogram were established using univariate and multivariable logistic regression analyses combined with clinical symptoms, subjective CT findings and radiomic features. The area under the receiver operating characteristic curve (AUC) was used to evaluate the performance of the models. RESULTS:The AUC for the clinical model was 0.77 in the training cohort (n=289; 95% confidence interval [CI]: 0.71-0.82; p = 0.001) and 0.75 in the validation cohort (n=124; 95% CI: 0.66-0.83; p = 0.016); the AUCs for the nomogram were 0.92 (95% CI: 0.89-0.95;p < 0.001) and 0.85 (95% CI: 0.78-0.91;p < 0.001), respectively. The Radscore, sex, pleural indentation and age were found independent predictors and were used to build the nomogram. CONCLUSIONS: The radiomic nomogram derived from clinical features, subjective CT signs and the radiomic score can potentially identify the risk of indeterminate SSPNs and aid in the patient's preoperative diagnosis.

Development of a combined radiomics and CT feature-based model for differentiating malignant from benign subcentimeter solid pulmonary nodules

“…Clinical and CT features is the important basis for radiologists to diagnose the malignant from benign SSPNs. This study found that the mean diameter, nodule-lung interface, spiculation, vacuole, pleural indentation, and air bronchogram were independent predictors of SSPNs, in agreement with the findings of previous studies [ 13 , 16 , 23 , 24 ]. This study found that the differences between benign and malignant nodules were statistically significant at pleural indentation, and the coefficient of this feature was the highest among CT features in the combined model.…”

“…Radiomics can extract quantitative features from medical images with high throughput and reflect the internal heterogeneity of tumor tissues that cannot be observed by human eyes through objective and quantitative methods [ 16 ]. Several studies have found that radiomics models perform well in classifying malignant from benign solid pulmonary nodules; however, these studies only focused on specific pathological types, i.e.…”

“…Several studies have found that radiomics models perform well in classifying malignant from benign solid pulmonary nodules; however, these studies only focused on specific pathological types, i.e. , lung adenocarcinoma and tuberculoma with nodule diameters less than 3 or 4 cm [ 11 , 13 , 15 , 16 ]. Zhang et al [ 16 ] proposed a diagnostic model combining CT and radiomic features and achieved an AUC of 0.85 (95% CI, 0.78–0.91) in the validation cohort, but they focused on solid nodules ranging from 5 to 20 mm.…”

Development and external validation of a multimodal integrated feature neural network (MIFNN) for the diagnosis of malignancy in small pulmonary nodules (≤10 mm)

“…In the domain of small pulmonary nodule diagnosis, several models have been developed but lack external validation [13][14][15]. Chae et al [14] achieved an AUC of 0.85 using deep learning for nodules 20 mm, while Zhang et al [15] reported an internal validation AUC of 0.85 for nodules ranging from 5-20 mm.…”

“…Recent studies focusing on the diagnosis of small pulmonary nodules, including the nomograms proposed by Chen et al [11] and Zhao et al [12], as well as the machine learning models introduced by Zhang et al [13] and Chae et al [14], primarily address nodules smaller than 20 mm in diameter. Mao et al's work [15], in particular, targets nodules ranging from 6 to 15 mm.…”