Predicting the histological invasiveness of pulmonary adenocarcinoma manifesting as persistent pure ground-glass nodules by ultra-high-resolution CT target scanning in the lateral or oblique body position

Super Resolution of Pulmonary Nodules Target Reconstruction Using a Two-Channel GAN Models

RATIONALE AND OBJECTIVES

To develop a Dual generative-adversarial-network (GAN) Cascaded Network (DGCN) for generating super-resolution computed tomography (SRCT) images from normal-resolution CT (NRCT) images and evaluate the performance of DGCN in multi-center datasets.

MATERIALS AND METHODS

This retrospective study included 278 patients with chest CT from two hospitals between January 2020 and June 2023, and each patient had all three NRCT (512×512 matrix CT images with a resolution of 0.70 mm, 0.70 mm,1.0 mm), high-resolution CT (HRCT, 1024×1024 matrix CT images with a resolution of 0.35 mm, 0.35 mm,1.0 mm), and ultra-high-resolution CT (UHRCT, 1024×1024 matrix CT images with a resolution of 0.17 mm, 0.17 mm, 0.5 mm) examinations. Initially, a deep chest CT super-resolution residual network (DCRN) was built to generate HRCT from NRCT. Subsequently, we employed the DCRN as a pre-trained model for the training of DGCN to further enhance resolution along all three axes, ultimately yielding SRCT. PSNR, SSIM, FID, subjective evaluation scores, and objective evaluation parameters related to pulmonary nodule segmentation in the testing set were recorded and analyzed.

RESULTS

DCRN obtained a PSNR of 52.16, SSIM of 0.9941, FID of 137.713, and an average diameter difference of 0.0981 mm. DGCN obtained a PSNR of 46.50, SSIM of 0.9990, FID of 166.421, and an average diameter difference of 0.0981 mm on 39 testing cases. There were no significant differences between the SRCT and UHRCT images in subjective evaluation.

CONCLUSION

Our model exhibited a significant enhancement in generating HRCT and SRCT images and outperformed established methods regarding image quality and clinical segmentation accuracy across both internal and external testing datasets.

The predictive value of computed tomography value on high-resolution images in differentiating invasive from indolent lung adenocarcinoma

Background

Invasive adenocarcinoma (IA) has a worse prognosis and different clinical management strategies compared to indolent lung adenocarcinoma including adenocarcinoma in situ (AIS) and minimally IA (MIA). The purpose of this study was to evaluate the predictive value of computed tomography (CT) value in differentiating invasive from indolent lung adenocarcinoma.

Methods

The pathological diagnoses and imaging data of confirmed lung adenocarcinomas manifested as lung nodules with homogeneous internal density which were surgically resected between August 2021 and July 2022 were retrospectively analyzed. Differences in CT values between invasive and indolent lung adenocarcinomas were compared in the primary cohort (n=766), and receiver operating characteristic (ROC) curve analysis was used to determine the optimal cut-off value. The predictive performance of the cut-off value was evaluated in the validation cohort (n=341).

Results

A total of 1,107 lung nodules from 1,014 patients were included in the total cohort. The CT values had a significant difference between invasive and indolent lung adenocarcinomas (P<0.001). Using the primary cohort, we determined the optimal cut-off value of -415 Hounsfield units (HU) of the CT value based on ROC curve, which showed good discrimination between IA and AIS/MIA in both the primary and validation cohorts (sensitivity, 85.98% and 87.42%, specificity, 87.67% and 84.74%, respectively).

Conclusions

The CT value of >-415 HU could be an effective predictor of invasive lung adenocarcinoma, thereby providing an appropriate clinical decision guide.

Predicting the pathological invasiveness in patients with a solitary pulmonary nodule via Shapley additive explanations interpretation of a tree-based machine learning radiomics model: a multicenter study

Background

Radiomics models could help assess the benign and malignant invasiveness and prognosis of pulmonary nodules. However, the lack of interpretability limits application of these models. We thus aimed to construct and validate an interpretable and generalized computed tomography (CT) radiomics model to evaluate the pathological invasiveness in patients with a solitary pulmonary nodule in order to improve the management of these patients.

Methods

We retrospectively enrolled 248 patients with CT-diagnosed solitary pulmonary nodules. Radiomic features were extracted from nodular region and perinodular regions of 3 and 5 mm. After coarse-to-fine feature selection, the radiomics score (radscore) was calculated using the least absolute shrinkage and selection operator logistic method. Univariate and multivariate logistic regression analyses were performed to determine the invasiveness-related clinicoradiological factors. The clinical-radiomics model was then constructed using the logistic and extreme gradient boosting (XGBoost) algorithms. The Shapley additive explanations (SHAP) method was then used to explain the contributions of the features. After removing batch effects with the ComBat algorithm, we assessed the generalization of the explainable clinical-radiomics model in two independent external validation cohorts (n=147 and n=149).

Results

The clinical-radiomic XGBoost model integrating the radscore, CT value, nodule length, and crescent sign demonstrated better predictive performance than did the clinical-radiomics logistic model in assessing pulmonary nodule invasiveness, with an area under the receiver operating characteristic (ROC) curve (AUC) of 0.889 [95% confidence interval (CI), 0.848-0.927] in the training cohort. The SHAP algorithm illustrates the contribution of each feature in the final model. The specific model decision process was visualized using a tree-based decision heatmap. Satisfactory generalization performance was shown with AUCs of 0.889 (95% CI, 0.823-0.942) and 0.915 (95% CI, 0.851-0.963) in the two external validation cohorts.

Conclusions

An interpretable and generalized clinical-radiomics model for predicting pulmonary nodule invasibility was constructed to help clinicians determine the invasiveness of pulmonary nodules and devise assessment strategies in an easily understandable manner.

Invasiveness identification in pure ground-glass nodules: exploring the generalizability of radiomics based on external validation and stress testing

PURPOSE

This study aimed to apply external validation and stress tests to evaluate the generalizability of radiomics models built using various machine-learning methods for identifying the invasiveness of lung adenocarcinomas manifesting as pure ground-glass nodules (pGGNs).

METHODS

This retrospective study enrolled 495 patients (514 pGGNs) confirmed as lung adenocarcinomas by postoperative pathology from three centers. All nodules were included in the primary cohort (randomly divided into training and test cohorts), two external validation cohorts, and two stress test cohorts. Six machine-learning radiomics models were constructed in the training cohort using the optimal features. Performance of radiomics models and clinical models were compared in primary cohort and external validation cohorts. The stress tests included stratified performance evaluation and shifted performance evaluation and contrastive evaluation under three single-condition modification settings. The predictive performance was validated by area under curve (AUC) of receiver operating characteristic (ROC).

RESULTS

Of the six radiomics models, the best logistic regression (LR) model was able to maintain high differential diagnostic capability (AUC: 0.849 ± 0.049) and good stability (relative standard deviation, 5.719%), but it showed poorer performance (AUC = 0.835) than the clinical model (AUC = 0.862) in the external validation cohort E1. The stress tests suggested LR model had no significant difference in performance between subgroups after stratification and had good consistency in the predictions before and after the three transformations (Kappa = 0.960, 0.840, and 0.933, respectively; p < 0.05, all).

CONCLUSION

The rigorous testing procedure facilitates the selection of high-performance radiomics models with good clinical generalizability.

An integrated nomogram combined semantic-radiomic features to predict invasive pulmonary adenocarcinomas in subjects with persistent subsolid nodules

Background

Patients with persistent pulmonary subsolid nodules have a relatively high incidence of lung adenocarcinoma. Preoperative early diagnosis of invasive pulmonary adenocarcinoma spectrum lesions could help avoid extensive advanced cancer management and overdiagnosis in lung cancer screening programs.

Methods

In total, 260 consecutive patients with persistent subsolid nodules ≤30 mm (n=260) confirmed by surgical pathology were retrospectively investigated from February 2016 to August 2020 at the Kaohsiung Veterans General Hospital. All patients underwent surgical resection within 3 months of the chest CT exam. The study subjects were divided into a training cohort (N=195) and a validation cohort (N=65) at a ratio of 3:1. The purpose of our study was to develop and validate a least absolute shrinkage and selection operator-derived nomogram integrating semantic-radiomic features in differentiating preinvasive and invasive pulmonary adenocarcinoma lesions, and compare its predictive value with clinical-semantic, semantic, and radiologist's performance.

Results

In the training cohort of 195 subsolid nodules, 106 invasive lesions and 89 preinvasive lesions were identified. We developed a least absolute shrinkage and selection operator-derived combined nomogram prediction model based on six predictors (nodular size, CTR, roundness, GLCM_Entropy_log10, HISTO_Entropy_log10, and CONVENTIONAL_Humean) to predict the invasive pulmonary adenocarcinoma lesions. Compared with other predictive models, the least absolute shrinkage and selection operator-derived model showed better diagnostic performance with an area under the curve of 0.957 (95% CI: 0.918 to 0.981) for detecting invasive pulmonary adenocarcinoma lesions with balanced sensitivity (92.45%) and specificity (88.64%). The results of Hosmer-Lemeshow test showed P values of 0.394 and 0.787 in the training and validation cohorts, respectively, indicating good calibration power.

Conclusions

We developed a least absolute shrinkage and selection operator-derived model integrating semantic-radiomic features with good calibration. This nomogram may help physicians to identify invasive pulmonary adenocarcinoma lesions for guidance in personalized medicine and make more informed decisions on managing subsolid nodules.

Computed Tomography Findings for Predicting Invasiveness of Lung Adenocarcinomas Manifesting as Pure Ground-Glass Nodules

Purpose: To comprehensively evaluate qualitative and quantitative features for predicting invasiveness of pure ground-glass nodules (pGGNs) using multiplanar computed tomography. Methods: Ninety-three resected pGGNs (16 atypical adenomatous hyperplasia [AAH], 18 adenocarcinoma in situ [AIS], 31 minimally invasive adenocarcinoma [MIA], and 28 invasive adenocarcinoma [IA]) were retrospectively included. Two radiologists analyzed qualitative and quantitative features on three standard planes. Univariable and multivariable logistic regression analyses were performed to identify features to distinguish the pre-invasive (AAH/AIS) from the invasive (MIA/IA) group. Results: Tumor size showed high area under the curve (AUC) for predicting invasiveness (.860, .863, .874, and .893, for axial long diameter [AXLD], multiplanar long diameter, mean diameter, and volume, respectively). The AUC for AXLD (cutoff, 11 mm) was comparable to that of the volume (P = .202). The invasive group had a significantly higher number of qualitative features than the pre-invasive group, regardless of tumor size. Six out of 59 invasive nodules (10.2%) were smaller than 11 mm, and all had at least one qualitative feature. pGGNs smaller than 11 mm without any qualitative features (n = 16) were all pre-invasive. In multivariable analysis, AXLD, vessel change, and the presence or number of qualitative features were independent predictors for invasiveness. The model with AXLD and the number of qualitative features achieved the highest AUC (.902, 95% confidence interval .833-.971). Conclusion: In adenocarcinomas manifesting as pGGNs on computed tomography, AXLD and the number of qualitative features are independent risk factors for invasiveness; small pGGNs (<11 mm) without qualitative features have low probability of invasiveness.

Development and validation of qualitative and quantitative models to predict invasiveness of lung adenocarcinomas manifesting as pure ground-glass nodules based on low-dose computed tomography during lung cancer screening

BACKGROUND

Due to different management strategy and prognosis of different subtypes of lung adenocarcinomas appearing as pure ground-glass nodules (pGGNs), it is important to differentiate invasive adenocarcinoma (IA) from adenocarcinoma in situ/minimally invasive adenocarcinoma (AIS/MIA) during lung cancer screening. The aim of this study was to develop and validate the qualitative and quantitative models to predict the invasiveness of lung adenocarcinoma appearing as pGGNs based on low-dose computed tomography (LDCT) and compare their diagnostic performance with that of intraoperative frozen section (FS).

METHODS

A total of 223 consecutive pathologically confirmed pGGNs from March 2018 to December 2020 were divided into a primary cohort (96 IAs and 64 AIS/MIAs) and validation cohort (39 IAs and 24 AIS/MIAs) according to scans (Brilliance iCT and Somatom Definition Flash) performed at Sichuan Cancer Hospital and Institute. The following LDCT features of pGGNs were analyzed: the qualitative features included nodule location, shape, margin, nodule-lung interface, lobulation, spiculation, pleural indentation, air bronchogram, vacuole, and vessel type, and the quantitative features included the diameter, volume, and mean attenuation. Multivariate logistic regression analysis was used to build a qualitative model, quantitative model, and combined qualitative and quantitative model. The diagnostic performance was assessed according to the following factors: the area under curve (AUC) of the receiver operating characteristic (ROC) curve, sensitivity, specificity, and accuracy.

RESULTS

The AUCs of the qualitative model, quantitative model, combined qualitative and quantitative model, and the FS diagnosis were 0.854, 0.803, 0.873, and 0.870, respectively, in the primary cohort and 0.884, 0.855, 0.875, and 0.946, respectively, in the validation cohort. No significant difference of the AUCs was found among the radiological models and the FS diagnosis in the primary or validation cohort (all corrected P>0.05). Among the radiological models, the combined qualitative and quantitative model consisting of vessel type and volume showed the highest accuracy in both the primary and validation cohorts (0.831 and 0.889, respectively).

CONCLUSIONS

The diagnostic performances of the qualitative and quantitative models based on LDCT to differentiate IA from AIS/MIA in pGGNs are equivalent to that of intraoperative FS diagnosis. The vessel type and volume can be preoperative and non-invasive biomarkers to assess the invasive risk of pGGNs in lung cancer screening.