Artificial intelligence-based vessel suppression for detection of sub-solid nodules in lung cancer screening computed tomography

Construction and validation of a risk score system for diagnosing invasive adenocarcinoma presenting as pulmonary pure ground-glass nodules: a multi-center cohort study in China

Background

Anxiety-driven clinical interventions have been queried due to the nondeterminacy of pure ground-glass nodules (pGGNs). Although radiomics and radiogenomics aid diagnosis, standardization and reproducibility challenges persist. We aimed to assess a risk score system for invasive adenocarcinoma in pGGNs.

Methods

In a retrospective, multi-center study, 772 pGGNs from 707 individuals in The Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital were grouped into training (509 patients with 558 observations) and validation (198 patients with 214 observations) sets consecutively from January 2017 to November 2021. An additional test set of 143 observations in Hainan Cancer Hospital was analyzed in the same period. Computed tomography (CT) signs and clinical features were manually collected, and the quantitative parameters were achieved by artificial intelligence (AI). The positive cutoff score was ≥3. Risk scores system 3 combined carcinoma history, chronic obstructive pulmonary disease (COPD), maximum diameters, nodule volume, mean CT values, type II or III vascular supply signs, and other radiographic characteristics. The evaluation included the area under the curves (AUCs), accuracy, sensitivity, specificity, positive predictive values (PPV), and negative predictive values (NPV) for both the risk score systems 1, 2, 3 and the AI model.

Results

The risk score system 3 [AUC, 0.840; 95% confidence interval (CI): 0.789-0.890] outperformed the AI model (AUC, 0.553; 95% CI: 0.487-0.619), risk score system 1 (AUC, 0.802; 95% CI: 0.754-0.851), and risk score system 2 (AUC, 0.816; 95% CI: 0.766-0.867), with 88.0% (0.850-0.904) accuracy, 95.6% (0.932-0.972) PPV, 0.620 (0.535-0.702) NPV, 89.6% (0.864-0.920) sensitivity, and 80.6% (0.717-0.872) specificity in the training sets. In the validation and test sets, risk score system 3 performed best with AUCs of 0.769 (0.678-0.860) and 0.801 (0.669-0.933).

Conclusions

An AI-based risk scoring system using quantitative image parameters, clinical features, and radiographic characteristics effectively predicts invasive adenocarcinoma in pulmonary pGGNs.

Accuracy of solid portion size measured on multiplanar volume rendering images for assessing invasiveness in lung adenocarcinoma manifesting as subsolid nodules

Background

The solid component of subsolid nodules (SSNs) is closely associated with the invasiveness of lung adenocarcinoma, and its accurate assessment is crucial for selecting treatment method. Therefore, this study aimed to evaluate the accuracy of solid component size within SSNs measured on multiplanar volume rendering (MPVR) and compare it with the dimensions of invasive components on pathology.

Methods

A pilot study was conducted using a chest phantom to determine the optimal MPVR threshold for the solid component within SSN, and then clinical validation was carried out by retrospective inclusion of patients with pathologically confirmed solitary SSN from October 2020 to October 2021. The radiological tumor size on MPVR and solid component size on MPVR (RSSm) and on lung window (RSSl) were measured. The size of the tumor and invasion were measured on the pathological section, and the invasion, fibrosis, and inflammation within SSNs were also recorded. The measurement difference between computed tomography (CT) and pathology, inter-observer and inter-measurement agreement were analyzed. Receiver operating characteristic (ROC) analysis and Bland-Altman plot were performed to evaluate the diagnostic efficiency of MPVR.

Results

A total of 142 patients (mean age, 54±11 years, 39 men) were retrospectively enrolled in the clinical study, with 26 adenocarcinomas in situ, 92 minimally invasive adenocarcinomas (MIAs), and 24 invasive adenocarcinomas (IAs). The RSSl was significantly smaller than pathological invasion size with fair inter-measurement agreement [intraclass correlation coefficient (ICC) =0.562, P<0.001] and moderate interobserver agreement (ICC =0.761, P<0.001). The RSSm was significantly larger than pathological invasion size with the excellent inter-measurement agreement (ICC =0.829, P<0.001) and excellent (ICC =0.952, P<0.001) interobserver agreement. ROC analysis showed that the cutoff value of RSSm for differentiating adenocarcinoma in situ from MIA and MIA from IA was 1.85 and 6.45 mm (sensitivity: 93.8% and 95.5%, specificity: 85.7% and 88.2%, 95% confidence internal: 0.914-0.993 and 0.900-0.983), respectively. The positive predictive value-and negative predictive value of MPVR in predicting invasiveness were 92.8% and 100%, respectively.

Conclusions

Using MPVR to predict the invasive degree of SSN had high accuracy and good inter-observer agreement, which is superior to lung window measurements and helpful for clinical decision-making.

Diagnostic value of artificial intelligence based on computed tomography (CT) density in benign and malignant pulmonary nodules: a retrospective investigation

Objective

To evaluate the diagnostic value of artificial intelligence (AI) in the detection and management of benign and malignant pulmonary nodules (PNs) using computed tomography (CT) density.

Methods

A retrospective analysis was conducted on the clinical data of 130 individuals diagnosed with PNs based on pathological confirmation. The utilization of AI and physicians has been employed in the diagnostic process of distinguishing benign and malignant PNs. The CT images depicting PNs were integrated into AI-based software. The gold standard for evaluating the accuracy of AI diagnosis software and physician interpretation was the pathological diagnosis.

Results

Out of 226 PNs screened from 130 patients diagnosed by AI and physician reading based on CT, 147 were confirmed by pathology. AI had a sensitivity of 94.69% and radiologists had a sensitivity of 85.40% in identifying PNs. The chi-square analysis indicated that the screening capacity of AI was superior to that of physician reading, with statistical significance (p < 0.05). 195 of the 214 PNs suggested by AI were confirmed pathologically as malignant, and 19 were identified as benign; among the 29 PNs suggested by AI as low risk, 13 were confirmed pathologically as malignant, and 16 were identified as benign. From the physician reading, 193 PNs were identified as malignant, 183 were confirmed malignant by pathology, and 10 appeared benign. Physician reading also identified 30 low-risk PNs, 19 of which were pathologically malignant and 11 benign. The physician readings and AI had kappa values of 0.432 and 0.547, respectively. The physician reading and AI area under curves (AUCs) were 0.814 and 0.798, respectively. Both of the diagnostic techniques had worthy diagnostic value, as indicated by their AUCs of >0.7.

Conclusion

It is anticipated that the use of AI-based CT diagnosis in the detection of PNs would increase the precision in early detection of lung carcinoma, as well as yield more precise evidence for clinical management.

An Effective Malignancy Prediction Model for Incidentally Detected Pulmonary Subsolid Nodules Based on Current and Prior CT Scans

INTRODUCTION

It is challenging to diagnose and manage incidentally detected pulmonary subsolid nodules due to their indolent nature and heterogeneity. The objective of this study is to construct a decision tree-based model to predict malignancy of a subsolid nodule based on radiomics features and evolution over time.

MATERIALS AND METHODS

We derived a training set (2947 subsolid nodules), a test set (280 subsolid nodules) from a cohort of outpatient CT scans, and a second test set (5171 subsolid nodules) from the National Lung Cancer Screening Trial (NLST). A Computer-Aided Diagnosis system (CADs) automatically extracted 28 preselected radiomics features, and we calculated the feature change rates as the change of the quantitative measure per time unit between the prior and current CT scans. We built classification models based on XGBoost and employed 5-fold cross validation to optimize the parameters.

RESULTS

The model that combined radiomics features with their change rates performed the best. The Areas Under Curve (AUCs) on the outpatient test set and on the NLST test set were 0.977 (95% CI, 0.958-0.996) and 0.955 (95% CI, 0.930-0.980), respectively. The model performed consistently well on subgroups stratified by nodule diameters, solid components, and CT scan intervals.

CONCLUSION

This decision tree-based model trained with the outpatient dataset gives promising predictive performance on the malignancy of pulmonary subsolid nodules. Additionally, it can assist clinicians to deliver more accurate diagnoses and formulate more in-depth follow-up strategies.

Application value of a computer-aided diagnosis and management system for the detection of lung nodules

Background

Computer-aided diagnosis (CAD) systems can help reduce radiologists' workload. This study assessed the value of a CAD system for the detection of lung nodules on chest computed tomography (CT) images.

Methods

The study retrospectively analyzed the CT images of patients who underwent routine health checkups between August 2019 and November 2019 at 3 hospitals in China. All images were first assessed by 2 radiologists manually in a blinded manner, which was followed by assessment with the CAD system. The location and classification of the lung nodules were determined. The final diagnosis was made by a panel of experts, including 2 associate chief radiologists and 1 chief radiologist at the radiology department. The sensitivity for nodule detection and false-positive nodules per case were calculated.

Results

A total of 1,002 CT images were included in the study, and the process was completed for 999 images. The sensitivity of the CAD system and manual detection was 90.19% and 49.88% (P<0.001), respectively. Similar sensitivity was observed between manual detection and the CAD system in lung nodules >15 mm (P=0.08). The false-positive nodules per case for the CAD system were 0.30±0.84 and those for manual detection were 0.24±0.68 (P=0.12). The sensitivity of the CAD system was higher than that of the radiologists, but the increase in the false-positive rate was only slight.

Conclusions

In addition to reducing the workload for medical professionals, a CAD system developed using a deep-learning model was highly effective and accurate in detecting lung nodules and did not demonstrate a meaningfully higher the false-positive rate.

Artificial Intelligence in Lung Cancer Screening: The Future Is Now

Lung cancer has one of the worst morbidity and fatality rates of any malignant tumour. Most lung cancers are discovered in the middle and late stages of the disease, when treatment choices are limited, and patients' survival rate is low. The aim of lung cancer screening is the identification of lung malignancies in the early stage of the disease, when more options for effective treatments are available, to improve the patients' outcomes. The desire to improve the efficacy and efficiency of clinical care continues to drive multiple innovations into practice for better patient management, and in this context, artificial intelligence (AI) plays a key role. AI may have a role in each process of the lung cancer screening workflow. First, in the acquisition of low-dose computed tomography for screening programs, AI-based reconstruction allows a further dose reduction, while still maintaining an optimal image quality. AI can help the personalization of screening programs through risk stratification based on the collection and analysis of a huge amount of imaging and clinical data. A computer-aided detection (CAD) system provides automatic detection of potential lung nodules with high sensitivity, working as a concurrent or second reader and reducing the time needed for image interpretation. Once a nodule has been detected, it should be characterized as benign or malignant. Two AI-based approaches are available to perform this task: the first one is represented by automatic segmentation with a consequent assessment of the lesion size, volume, and densitometric features; the second consists of segmentation first, followed by radiomic features extraction to characterize the whole abnormalities providing the so-called "virtual biopsy". This narrative review aims to provide an overview of all possible AI applications in lung cancer screening.

Computer-Aided Detection of Subsolid Nodules on Chest Computed Tomography: Assessment of Visualization on Vessel-Suppressed Images

OBJECTIVES

This study aimed to clarify the performance of automatic detection of subsolid nodules by commercially available software on computed tomography (CT) images of various slice thicknesses and compare it with visualization on the accompanying vessel-suppression CT (VS-CT) images.

METHODS

A total of 95 subsolid nodules from 84 CT examinations of 84 patients were included. The reconstructed CT image series of each case with 3-, 2-, and 1-mm slice thicknesses were loaded into a commercially available software application (ClearRead CT) for automatic detection of subsolid nodules and generation of VS-CT images. Automatic nodule detection sensitivity was assessed for 95 nodules on each series of images acquired at 3 slice thicknesses. Four radiologists subjectively evaluated visual assessment of the nodules on VS-CT.

RESULTS

ClearRead CT automatically detected 69.5% (66/95 nodules), 68.4% (65/95 nodules), and 70.5% (67/95 nodules) of all subsolid nodules in 3-, 2-, and 1-mm slices, respectively. The detection rate was higher for part-solid nodules than for pure ground-glass nodules at all slice thicknesses. In the visualization assessment on VS-CT, 3 nodules at each slice thickness (3.2%) were judged as invisible, while 26 of 29 (89.7%), 27 of 30 (90.0%), and 25 of 28 (89.3%) nodules, which were missed by computer-aided detection, were judged as visible in 3-, 2-, and 1-mm slices, respectively.

CONCLUSIONS

The automatic detection rate of subsolid nodules by ClearRead CT was approximately 70% at all slice thicknesses. More than 95% of subsolid nodules were visualized on VS-CT, including nodules undetected by the automated software. Computed tomography acquisition at slices thinner than 3 mm did not confer any benefits.

Artificial intelligence: A critical review of applications for lung nodule and lung cancer

Artificial intelligence (AI) is a broad concept that usually refers to computer programs that can learn from data and perform certain specific tasks. In the recent years, the growth of deep learning, a successful technique for computer vision tasks that does not require explicit programming, coupled with the availability of large imaging databases fostered the development of multiple applications in the medical imaging field, especially for lung nodules and lung cancer, mostly through convolutional neural networks (CNN). Some of the first applications of AI is this field were dedicated to automated detection of lung nodules on X-ray and computed tomography (CT) examinations, with performances now reaching or exceeding those of radiologists. For lung nodule segmentation, CNN-based algorithms applied to CT images show excellent spatial overlap index with manual segmentation, even for irregular and ground glass nodules. A third application of AI is the classification of lung nodules between malignant and benign, which could limit the number of follow-up CT examinations for less suspicious lesions. Several algorithms have demonstrated excellent capabilities for the prediction of the malignancy risk when a nodule is discovered. These different applications of AI for lung nodules are particularly appealing in the context of lung cancer screening. In the field of lung cancer, AI tools applied to lung imaging have been investigated for distinct aims. First, they could play a role for the non-invasive characterization of tumors, especially for histological subtype and somatic mutation predictions, with a potential therapeutic impact. Additionally, they could help predict the patient prognosis, in combination to clinical data. Despite these encouraging perspectives, clinical implementation of AI tools is only beginning because of the lack of generalizability of published studies, of an inner obscure working and because of limited data about the impact of such tools on the radiologists' decision and on the patient outcome. Radiologists must be active participants in the process of evaluating AI tools, as such tools could support their daily work and offer them more time for high added value tasks.

Artificial Intelligence (AI) for Lung Nodules, From the AJR Special Series on AI Applications

Interest in artificial intelligence (AI) applications for lung nodules continues to grow among radiologists, particularly with the expanding eligibility criteria and clinical utilization of lung cancer screening CT. AI has been heavily investigated for detecting and characterizing lung nodules and for guiding prognostic assessment. AI tools have also been used for image postprocessing (e.g., rib suppression on radiography or vessel suppression on CT) and for noninterpretive aspects of reporting and workflow, including management of nodule follow-up. Despite growing interest in and rapid development of AI tools and FDA approval of AI tools for pulmonary nodule evaluation, integration into clinical practice has been limited. Challenges to clinical adoption have included concerns about generalizability, regulatory issues, technical hurdles in implementation, and human skepticism. Further validation of AI tools for clinical use and demonstration of benefit in terms of patient-oriented outcomes also are needed. This article provides an overview of potential applications of AI tools in the imaging evaluation of lung nodules and discusses the challenges faced by practices interested in clinical implementation of such tools.

Deep learning-based growth prediction for sub-solid pulmonary nodules on CT images

Background

Estimating the growth of pulmonary sub-solid nodules (SSNs) is crucial to the successful management of them during follow-up periods. The purpose of this study is to (1) investigate the measurement sensitivity of diameter, volume, and mass of SSNs for identifying growth and (2) seek to establish a deep learning-based model to predict the growth of SSNs.

Methods

A total of 2,523 patients underwent at least 2-year examination records retrospectively collected with sub-solid nodules. A total of 2,358 patients with 3,120 SSNs from the NLST dataset were randomly divided into training and validation sets. Patients from the Yibicom Health Management Center and Guangdong Provincial People’s Hospital were collected as an external test set (165 patients with 213 SSN). Trained models based on LUNA16 and Lndb19 datasets were employed to automatically obtain the diameter, volume, and mass of SSNs. Then, the increase rate in measurements between cancer and non-cancer groups was studied to evaluate the most appropriate way to identify growth-associated lung cancer. Further, according to the selected measurement, all SSNs were classified into two groups: growth and non-growth. Based on the data, the deep learning-based model (SiamModel) and radiomics model were developed and verified.

Results

The double time of diameter, volume, and mass were 711 vs. 963 days (P = 0.20), 552 vs. 621 days (P = 0.04) and 488 vs. 623 days (P< 0.001) in the cancer and non-cancer groups, respectively. Our proposed SiamModel performed better than the radiomics model in both the NLST validation set and external test set, with an AUC of 0.858 (95% CI 0.786–0.921) and 0.760 (95% CI 0.646–0.857) in the validation set and 0.862 (95% CI 0.789–0.927) and 0.681 (95% CI 0.506–0.841) in the external test set, respectively. Furthermore, our SiamModel could use the data from first-time CT to predict the growth of SSNs, with an AUC of 0.855 (95% CI 0.793–0.908) in the NLST validation set and 0.821 (95% CI 0.725–0.904) in the external test set.

Conclusion

Mass increase rate can reflect more sensitively the growth of SSNs associated with lung cancer than diameter and volume increase rates. A deep learning-based model has a great potential to predict the growth of SSNs.

Outstanding negative prediction performance of solid pulmonary nodule volume AI for ultra-LDCT baseline lung cancer screening risk stratification

OBJECTIVE

To evaluate performance of AI as a standalone reader in ultra-low-dose CT lung cancer baseline screening, and compare it to that of experienced radiologists.

METHODS

283 participants who underwent a baseline ultra-LDCT scan in Moscow Lung Cancer Screening, between February 2017-2018, and had at least one solid lung nodule, were included. Volumetric nodule measurements were performed by five experienced blinded radiologists, and independently assessed using an AI lung cancer screening prototype (AVIEW LCS, v1.0.34, Coreline Soft, Co. ltd, Seoul, Korea) to automatically detect, measure, and classify solid nodules. Discrepancies were stratified into two groups: positive-misclassification (PM); nodule classified by the reader as a NELSON-plus /EUPS-indeterminate/positive nodule, which at the reference consensus read was < 100 mm3, and negative-misclassification (NM); nodule classified as a NELSON-plus /EUPS-negative nodule, which at consensus read was ≥ 100 mm3.

RESULTS

1149 nodules with a solid-component were detected, of which 878 were classified as solid nodules. For the largest solid nodule per participant (n = 283); 61 [21.6 %; 53 PM, 8 NM] discrepancies were reported for AI as a standalone reader, compared to 43 [15.1 %; 22 PM, 21 NM], 36 [12.7 %; 25 PM, 11 NM], 29 [10.2 %; 25 PM, 4 NM], 28 [9.9 %; 6 PM, 22 NM], and 50 [17.7 %; 15 PM, 35 NM] discrepancies for readers 1, 2, 3, 4, and 5 respectively.

CONCLUSION

Our results suggest that through the use of AI as an impartial reader in baseline lung cancer screening, negative-misclassification results could exceed that of four out of five experienced radiologists, and radiologists' workload could be drastically diminished by up to 86.7%.

Lung-RADS Category 3 and 4 Nodules on Lung Cancer Screening in Clinical Practice

Background: Lung-RADS category 3 and 4 nodules account for most cancers among screening-detected lung cancers and are considered actionable nodules with management implications. The cancer frequency among such nodules is estimated in the Lung-RADS recommendations and has been investigated primarily using retrospectively assigned Lung-RADS classifications. Objective: To assess the frequency of cancer among lung nodules assigned Lung-RADS category 3 or 4 at lung cancer screening (LCS) in clinical practice and factors that impact the cancer frequency within each category. Methods: This retrospective study was based on review of clinical radiology reports of 9148 consecutive low-dose CT LCS examinations performed in 4798 patients between June 2014 and January 2021 as part of an established LCS program. Unique nodules assigned Lung-RADS category 3 or 4 (4A, 4B, or 4X) that were clinically categorized as benign or malignant by a multidisciplinary conference, considering histologic analysis and follow-up imaging, were selected for further analysis; benign diagnoses based on stability required at least 12 months of follow-up imaging. Indeterminate nodules were excluded. Cancer frequencies were evaluated. Results: Of the 9148 LCS examinations, 857 (9.4%) were assigned Lung-RADS category 3, and 721 (7.9%) were assigned category 4. The final analysis included 1297 unique nodules in 1139 patients (598 men, 541 women; mean age, 66.0±6.3 years). A total of 1108/1297 (85.4%) nodules were deemed benign, and 189/1297 (14.6%) were deemed malignant. Frequencies of malignancy for category 3, 4A, 4B, and 4X nodules were 3.9%, 15.5%, 36.3%, and 76.8%, respectively. A total of 45/46 (97.8%) endobronchial nodules (all category 4A) were deemed benign based on resolution. Cancer frequency was 13.1%, 24.4%, and 13.5% for solid, part-solid, and ground-glass nodules, respectively. Conclusions: When applying Lung-RADS for LCS clinical practice, the frequency of Lung-RADS category 3 and 4 nodules, as well as cancer frequency in these categories, were higher than prevalence and cancer risk estimated for category 3 and 4 nodules in the Lung-RADS recommendations and reported in earlier studies using retrospective category assignments. Nearly all endobronchial category 4A nodules were benign. Clinical Impact: Future Lung-RADS iterations should consider these findings from real-world practice to improve the system's clinical utility.

Bronchial morphological changes are associated with postoperative intractable cough after right upper lobectomy in lung cancer patients

BACKGROUND

To date, postoperative intractable cough (PIC) has not received adequate attention, and the complex perioperative factors when performing pulmonary resection often prevent researchers from addressing this issue. This study aimed to investigate the clinicopathological and radiographic indicators related to PIC in lung cancer patients.

METHODS

In all, 112 patients who had had right upper lobectomy for primary lung cancer from January 2019 to December 2020 were retrospectively reviewed. We collected data via the electronic medical database of our department. Bronchial morphological features were investigated comprehensively via three-dimensional chest computer tomography reconstruction images.

RESULTS

During outpatient follow-up visits, 41 (36.6%) patients complained about persistent dry cough after surgery. Compared with the non-cough group, patients in the refractory cough group showed significant differences in smoking history, right upper lobe stump length, changes of right bronchus intermedius (RBI) diameter, changes of right lower lobe (RLL) basal bronchus diameter, changes of RBI/RLL bronchial angle, and bronchial kink. However, according to multivariable regression analysis, stump length, bronchial kink, and diameter change of the right lower lobe basal bronchus were independently associated with postoperative refractory cough. A nebulization drug was prescribed for the 41 patients diagnosed with PIC, and 33 (80.5%) patients had improved by the next visit.

CONCLUSIONS

After right upper lobectomy, the morphology of the remaining bronchial tree in the residual lung changed significantly. The bronchial morphological alterations were independent risk factors for PIC.

The Added Effect of Artificial Intelligence on Physicians' Performance in Detecting Thoracic Pathologies on CT and Chest X-ray: A Systematic Review

Our systematic review investigated the additional effect of artificial intelligence-based devices on human observers when diagnosing and/or detecting thoracic pathologies using different diagnostic imaging modalities, such as chest X-ray and CT. Peer-reviewed, original research articles from EMBASE, PubMed, Cochrane library, SCOPUS, and Web of Science were retrieved. Included articles were published within the last 20 years and used a device based on artificial intelligence (AI) technology to detect or diagnose pulmonary findings. The AI-based device had to be used in an observer test where the performance of human observers with and without addition of the device was measured as sensitivity, specificity, accuracy, AUC, or time spent on image reading. A total of 38 studies were included for final assessment. The quality assessment tool for diagnostic accuracy studies (QUADAS-2) was used for bias assessment. The average sensitivity increased from 67.8% to 74.6%; specificity from 82.2% to 85.4%; accuracy from 75.4% to 81.7%; and Area Under the ROC Curve (AUC) from 0.75 to 0.80. Generally, a faster reading time was reported when radiologists were aided by AI-based devices. Our systematic review showed that performance generally improved for the physicians when assisted by AI-based devices compared to unaided interpretation.

Consecutive Serial Non-Contrast CT Scan-Based Deep Learning Model Facilitates the Prediction of Tumor Invasiveness of Ground-Glass Nodules

Introduction

Tumors are continuously evolving biological systems which can be monitored by medical imaging. Previous studies only focus on single timepoint images, whether the performance could be further improved by using serial noncontrast CT imaging obtained during nodule follow-up management remains unclear. In this study, we evaluated DL model for predicting tumor invasiveness of GGNs through analyzing time series CT images.

Methods

A total of 168 pathologically confirmed GGN cases (48 noninvasive lesions and 120 invasive lesions) were retrospectively collected and randomly assigned to the development dataset (n = 123) and independent testing dataset (n = 45). All patients underwent consecutive noncontrast CT examinations, and the baseline CT and 3-month follow-up CT images were collected. The gross region of interest (ROI) patches containing only tumor region and the full ROI patches including both tumor and peritumor regions were cropped from CT images. A baseline model was built on the image features and demographic features. Four DL models were proposed: two single-DL model using gross ROI (model 1) or full ROI patches (model 3) from baseline CT images, and two serial-DL models using gross ROI (model 2) or full ROI patches (model 4) from consecutive CT images (baseline scan and 3-month follow-up scan). In addition, a combined model integrating serial full ROI patches and clinical information was also constructed. The performance of these predictive models was assessed with respect to discrimination and clinical usefulness.

Results

The area under the curve (AUC) of the baseline model, models 1, 2, 3, and 4 were 0.562 [(95% confidence interval (C)], 0.406~0.710), 0.693 (95% CI, 0.538-0.822), 0.787 (95% CI, 0.639-0.895), 0.727 (95% CI, 0.573-0.849), and 0.811 (95% CI, 0.667-0.912) in the independent testing dataset, respectively. The results indicated that the peritumor region had potential to contribute to tumor invasiveness prediction, and the model performance was further improved by integrating imaging scans at multiple timepoints. Furthermore, the combined model showed best discrimination ability, with AUC, sensitivity, specificity, and accuracy achieving 0.831 (95% CI, 0.690-0.926), 86.7%, 73.3%, and 82.2%, respectively.

Conclusion

The DL model integrating full ROIs from serial CT images shows improved predictive performance in differentiating noninvasive from invasive GGNs than the model using only baseline CT images, which could benefit the clinical management of GGNs.