Management of incidental pulmonary nodules: current strategies and future perspectives

Dual-layer detector spectral CT-based machine learning models in the differential diagnosis of solitary pulmonary nodules

The benign and malignant status of solitary pulmonary nodules (SPNs) is a key determinant of treatment decisions. The main objective of this study was to validate the efficacy of machine learning (ML) models featured with dual-layer detector spectral computed tomography (DLCT) parameters in identifying the benign and malignant status of SPNs. 250 patients with pathologically confirmed SPN were included in this study. 8 quantitative and 16 derived parameters were obtained based on the regions of interest of the lesions on the patients' DLCT chest enhancement images. 6 ML models were constructed from 10 parameters selected after combining the patients' clinical parameters, including gender, age, and smoking history. The logistic regression model showed the best diagnostic performance with an area under the receiver operating characteristic curve (AUC) of 0.812, accuracy of 0.813, sensitivity of 0.750 and specificity of 0.791 on the test set. The results suggest that the ML models based on DLCT parameters are superior to the traditional CT parameter models in identifying the benign and malignant nature of SPNs, and have greater potential for application.

Comparison of serum from lung cancer patients and from patients with benign lung nodule using FTIR spectroscopy

Lungcancer remains the leading cause of cancer related deaths in worldwide. Earlydiagnosis oflungcancer can significantly improve survival rate. However, due to its close resemblance to the malignant nodules, the possible existence of benign nodules often leads to erroneous decisions. The aim of this study was to explore whether fourier transform infrared (FTIR) spectroscopy could improve the accuracy of early diagnosis of lung cancer by distinguishing lung cancer patients' (LCP') serum from patients with benign lung nodules' (PBLN') serum. In this study, A1243+1081/A1652+1539 ratio in LCP group was increased significantly compared with that in PBLN group, indicating that the ratio could be used to distinguish the serum of LCP from that of PBLN. In addition, the ratios of A2926/A2969, A1744/A2926+2859, A2926+2859/A1652+1539 were also increased significantly in LCP group compared with that in PBLN group. These findings suggest that FTIR spectroscopy might be a potentially effective method for the early diagnosis of lung cancer.

A comprehensive nomogram combining CT-based radiomics with clinical features for differentiation of benign and malignant lung subcentimeter solid nodules

Objective

To establish a nomogram based on non-enhanced computed tomography(CT) imaging radiomics and clinical features for use in predicting the malignancy of sub-centimeter solid nodules (SCSNs).

Materials and methods

Retrospective analysis was performed of records for 198 patients with SCSNs that were surgically resected and examined pathologically at two medical institutions between January 2020 and June 2021. Patients from Center 1 were included in the training cohort (n = 147), and patients from Center 2 were included in the external validation cohort (n = 52). Radiomic features were extracted from chest CT images. The least absolute shrinkage and selection operator (LASSO) regression model was used for radiomic feature extraction and computation of radiomic scores. Clinical features, subjective CT findings, and radiomic scores were used to build multiple predictive models. Model performance was examined by evaluating the area under the receiver operating characteristic curve (AUC). The best model was selected for efficacy evaluation in a validation cohort, and column line plots were created.

Results

Pulmonary malignant nodules were significantly associated with vascular alterations in both the training (p < 0.001) and external validation (p < 0.001) cohorts. Eleven radiomic features were selected after a dimensionality reduction to calculate the radiomic scores. Based on these findings, three prediction models were constructed: subjective model (Model 1), radiomic score model (Model 2), and comprehensive model (Model 3), with AUCs of 0.672, 0.888, and 0.930, respectively. The optimal model with an AUC of 0.905 was applied to the validation cohort, and decision curve analysis indicated that the comprehensive model column line plot was clinically useful.

Conclusion

Predictive models constructed based on CT-based radiomics with clinical features can help clinicians diagnose pulmonary nodules and guide clinical decision making.

Feature selection methods and predictive models in CT lung cancer radiomics

Radiomics is a technique that extracts quantitative features from medical images using data-characterization algorithms. Radiomic features can be used to identify tissue characteristics and radiologic phenotyping that is not observable by clinicians. A typical workflow for a radiomics study includes cohort selection, radiomic feature extraction, feature and predictive model selection, and model training and validation. While there has been increasing attention given to radiomic feature extraction, standardization, and reproducibility, currently, there is a lack of rigorous evaluation of feature selection methods and predictive models. Herein, we review the published radiomics investigations in CT lung cancer and provide an overview of the commonly used radiomic feature selection methods and predictive models. We also compare limitations of various methods in clinical applications and present sources of uncertainty associated with those methods. This review is expected to help raise awareness of the impact of radiomic feature and model selection methods on the integrity of radiomics studies.

An Innovative Method: Predicting the Visibility of Radial Endobronchial Ultrasound for Peripheral Pulmonary Nodules by Virtual Bronchoscopic Navigation

Background: The diagnosis of peripheral pulmonary nodules (PPNs) still is the key and difficult point. Previous studies have demonstrated that the diagnostic yield of radial endobronchial ultrasound (rEBUS) visible nodules is significantly higher than that of invisible nodules. The traditional method of predicting the rEBUS-visibility of nodules is based on the CT-bronchus signs, but its effectiveness may be unsatisfactory. Objective: We innovate a valuable predictive model based on virtual bronchoscopic navigation to identify beforehand which PPNs are likely to be successfully visualized by rEBUS. The innovative predictor is the ratio of the size of lesions (S) to the shortest straight-line distance (D) from the terminal point of the virtual navigation path to the localization point of the nodule. Methods: This is a retrospective study. On the training dataset of 214 patients, a receiver operating characteristic curve was drawn to understand the utility of the predictive model and get the optimal cut-off points. Ninety-two cases were enrolled in the validation dataset to validate the external predictive accuracy of the predictor. Results: The optimal cut-off point of the curve was 1.84 with the Youden index of 0.65, at which point the area under the curve was 0.85 (95% CI: 0.76-0.95). The predictor has a good performance in the validation dataset with sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of 81%, 100%, 100%, 71%, and 87%, respectively. Conclusion: The S/D ratio is a valuable and innovative method to identify beforehand which PPNs are likely to be successfully visualized by rEBUS. If the S/D ratio of the nodule is greater than 1.84, it will be visualized by rEBUS.

Pulmonary Nodule Clinical Trial Data Collection and Intelligent Differential Diagnosis for Medical Internet of Things

In this paper, the medical Internet of things (IoT) is used to pool data from clinical trials of pulmonary nodules, and on this basis, intelligent differential diagnosis techniques are investigated. A filtered orthogonal frequency division multiplexing model based on polarisation coding is proposed, where the input data are fed to a modulator after polarisation cascade coding, and the system performance is analysed under a medical Internet of things modulated additive Gaussian white noise channel. The above polarisation-coded filtered orthogonal frequency division multiplexing system components are applied to electroencephalogram (EEG) signal transmission, to which a threshold compression module and a vector reconstruction module are added to address the system power burden associated with the acquisition and transmission of large amounts of real-time EEG data in the medical IoT. In the threshold compression module, the inherent characteristics of EEG signals are analysed, and the generated EEG data are decomposed into multiple symbolic streams and compressed by applying different thresholds to improve the compression ratio while ensuring the quality of service of the application. A deep neural network-based approach is proposed for the detection and diagnosis of lung nodules. Automatic identification and measurement of simulated lung nodules and the corresponding volumes of nodules in images under different conditions are applied. The sensitivity of each AIADS in identifying lung nodules under different convolution kernel conditions, false positives (FP), false negatives (FN), relative volume errors (RVE), the miss detection rate (MDR) for different types of lung nodules, and the performance of each system in predicting the four types of nodules are calculated. In this paper, an interpretable multibranch feature convolutional neural network model is proposed for the diagnosis of benign and malignant lung nodules. It is demonstrated that the proposed model not only yields interpretable lung nodule classification results but also achieves better lung nodule classification performance with an accuracy rate of 97.8%.

Development, Validation, and Comparison of Image-Based, Clinical Feature-Based and Fusion Artificial Intelligence Diagnostic Models in Differentiating Benign and Malignant Pulmonary Ground-Glass Nodules

Objective

This study aimed to develop effective artificial intelligence (AI) diagnostic models based on CT images of pulmonary nodules only, on descriptional and quantitative clinical or image features, or on a combination of both to differentiate benign and malignant ground-glass nodules (GGNs) to assist in the determination of surgical intervention.

Methods

Our study included a total of 867 nodules (benign nodules: 112; malignant nodules: 755) with postoperative pathological diagnoses from two centers. For the diagnostic models to discriminate between benign and malignant GGNs, we adopted three different artificial intelligence (AI) approaches: a) an image-based deep learning approach to build a deep neural network (DNN); b) a clinical feature-based machine learning approach based on the clinical and image features of nodules; c) a fusion diagnostic model integrating the original images and the clinical and image features. The performance of the models was evaluated on an internal test dataset (the “Changzheng Dataset”) and an independent test dataset collected from an external institute (the “Longyan Dataset”). In addition, the performance of automatic diagnostic models was compared with that of manual evaluations by two radiologists on the ‘Longyan dataset’.

Results

The image-based deep learning model achieved an appealing diagnostic performance, yielding AUC values of 0.75 (95% confidence interval [CI]: 0.62, 0.89) and 0.76 (95% CI: 0.61, 0.90), respectively, on both the Changzheng and Longyan datasets. The clinical feature-based machine learning model performed well on the Changzheng dataset (AUC, 0.80 [95% CI: 0.64, 0.96]), whereas it performed poorly on the Longyan dataset (AUC, 0.62 [95% CI: 0.42, 0.83]). The fusion diagnostic model achieved the best performance on both the Changzheng dataset (AUC, 0.82 [95% CI: 0.71-0.93]) and the Longyan dataset (AUC, 0.83 [95% CI: 0.70-0.96]), and it achieved a better specificity (0.69) than the radiologists (0.33-0.44) on the Longyan dataset.

Conclusion

The deep learning models, including both the image-based deep learning model and the fusion model, have the ability to assist radiologists in differentiating between benign and malignant nodules for the precise management of patients with GGNs.

Malignancy risk stratification for solitary pulmonary nodule: A clinical practice guideline

CLINICAL QUESTION

The detection rate of the solitary pulmonary nodule (SPN) is increasing with the popularization of CT scanning. Malignancy risk stratification for SPN is a major clinical difficulty.

CURRENT PRACTICE

There have been several guidelines for SPN assessment. Inconsistency of these guidelines makes the clinical application difficult and confusing.

RECOMMENDATIONS

In this Rapid Recommendation, solid and subsolid SPNs are recommended to be evaluated respectively. Six factors, namely the combination of age with sex, smoking history, history of malignancy, family history of malignancy, and nodule size, are recommended for malignancy risk stratification for both kinds of SPNs; the border of nodules (spiculation and lobulation) is recommended for evaluating solid SPNs and the density of nodules (pure or mixed ground-glass nodule) is recommended for subsolid nodules. Among them, smoking history and radiologic features (nodule diameter, border, and density) are of relatively higher importance. A scoring system was proposed to assist malignancy risk stratification of SPNs, with a total score ranging from six points to 15 points (if solid) or 17 points (if subsolid). For each SPN, regardless of solid or subsolid in nature, a total score of ≤ 7 points suggested a low risk of being malignant, while 7 to 9 points suggested medium risk, and ≥ 9 points suggested high risk.

HOW THIS GUIDELINE WAS CREATED

This rapid recommendation was developed using the MAGIC (Making GRADE the Irresistible Choice) methodological framework. First, a clinical subcommittee identified the topic of recommendation and requested evidence. Then, an independent evidence synthesis subcommittee performed a comprehensive literature review and evaluated the evidence. Finally, based on findings from the systematic review and use of real-world data, the clinical subcommittee formulated recommendations, including the scoring system, through a consensus procedure.

THE EVIDENCE

A total of 13857 patients with SPNs were included in the meta-analysis and the association between 12 candidate factors and the risk of SPNs being malignant was studied. Eventually, seven factors were recommended for SPNs evaluation, and a scoring system was proposed.

UNDERSTANDING THE RECOMMENDATION

The parameters included are objective. Therefore, this recommendation is feasible in clinical practice. However, there are several uncertainties, such as a lack of further verification. It might be misclassified by the scoring system. Clinicians could choose the most suitable scheme according to the recommendation, along with their own experience in specific situations.

Application Value of PET/CT and MRI in the Diagnosis and Treatment of Patients With Synchronous Multiple Pulmonary Ground-Glass Nodules

Background

Synchronous multiple ground-glass nodules (SMGGNs) in synchronous multiple lung cancers are associated with specific imaging findings. It is difficult to distinguish whether multiple nodules are primary tumors or metastatic lesions in the lungs. The need for PET/CT and contrast-enhanced brain MRI for these patients remains unclear. This study investigated the necessity of these two imaging examinations for SMGGN patients by means of retrospective analysis.

Methods

SMGGN patients who were diagnosed and treated in our hospital from October 2017 to May 2020 and underwent whole-body PET/CT(Cranial excepted) and/or contrast-enhanced brain MRI+DWI were enrolled in this study. We analyzed the imaging and clinical characteristics of these patients to evaluate SMGGN patients’ need to undergo whole-body PET/CT and brain MRI examination.

Results

A total of 87 SMGGN patients were enrolled. 51 patients underwent whole-body PET/CT examinations and did not show signs of primary tumors in other organs, metastatic foci in other organs, or metastasis to surrounding lymph nodes. 87 patients underwent whole-brain MRI, which did not reveal brain metastases but did detect an old cerebral infarction in 23 patients and a new cerebral infarction in one patient. 87 patients underwent surgical treatment in which 219 nodules were removed. All nodules were diagnosed as adenocarcinoma or atypical adenomatous hyperplasia. No lymph node metastasis was noted.

Conclusion

For SMGGN patients, PET/CT and enhanced cranial MRI are unnecessary for SMGGNs patients, but from the perspective of perioperative patient safety, preoperative MRI+DWI examination is recommended for SMGGNs patients.

Developing of risk models for small solid and subsolid pulmonary nodules based on clinical and quantitative radiomics features

Background

Accurate evaluation of pulmonary nodule malignancy is important for lung cancer management. This current study aimed to develop risk models for small solid and subsolid pulmonary nodules based on clinical and quantitative radiomics features.

Methods

This study enrolled 5–20 mm pulmonary nodules detected on thoracic high-resolution computed tomography (HRCT), which were all confirmed pathologically. There were 548 solid nodules (242 malignant vs. 306 benign) and 623 subsolid nodules (SSNs 519 malignant vs. 104 benign). Relevant clinical characteristics were recorded. The CT image prior to the initial treatment was chosen for manual segmentation of the targeted nodule using the ITK-SNAP software. Subsequently, the marked image was processed to quantitatively extract 1218 radiomics features using PyRadiomics. We performed five-fold cross-validation to select potential predictors from clinical and radiomics features using the LASSO method and to evaluate the performance of the established models. In total, four types of models were tried: random forest, XGBOOST, SVM, and logistic models. The established models were compared with the Mayo model.

Results

Lung cancer risk models were developed among four nodule groups: all nodules (410 benign vs. 761 malignant; 1:1.86), nodules ≤10 mm (185 benign vs. 224 malignant; 1:1.21), solid nodules (306 benign vs. 242 malignant; 1.26:1), and SSNs (104 benign vs. 104 malignant; 1:1 matched). Significant clinical and radiomics predictors were selected for each group. The accuracy, area under the ROC curve, sensitivity, and specificity of the best model on validation dataset was 0.86, 0.91, 0.93, 0.73 for all nodules (XGBOOST), 0.82, 0.90, 0.86, 0.76 for nodules ≤10 mm (XGBOOST), 0.80, 0.89, 0.78, 0.82 for solid nodules (XGBOOST) and 0.70, 0.73, 0.73, 0.67 for SSNs (Random Forest). Except for the SSN models, the established clinical-radiomics models were superior to the Mayo model.

Conclusions

Predictive models based on both clinical and radiomics features can be used to assess the malignancy of small solid and subsolid pulmonary nodules, even for nodules that are 10 mm or smaller.

MRI Image Segmentation Model with Support Vector Machine Algorithm in Diagnosis of Solitary Pulmonary Nodule

This study focused on the application value of MRI images processed by a Support Vector Machine (SVM) algorithm-based model in diagnosis of benign and malignant solitary pulmonary nodule (SPN). The SVM algorithm was constrained by a self-paced regularization item and gradient value to establish the MRI image segmentation model (SVM-L) for lung. Its performance was compared factoring into the Dice index (DI), sensitivity (SE), specificity (SP), and Mean Square Error (MSE). 28 SPN patients who underwent the parallel MRI examination were selected as research subjects and were divided into the benign group (11 patients) and malignant group (17 patients) according to different plans for diagnosis and treatment. The apparent diffusion coefficient (ADC) at different b values was analyzed, and the steepest slope (SS) and washout ratio (WR) values in the two groups were calculated. The result showed that the MSE, DI, SE, SP values, and operation time of the SVM-L model were (0.41 ± 0.02), (0.84 ± 0.13), (0.89 ± 0.04), (0.993 ± 0.004), and (30.69 ± 2.60)s, respectively, apparently superior to those of the other algorithms, but there were no statistic differences (P > 0.05) in the WR value between the two groups of patients. The SS values of the time-signal curve in the benign and malignant groups were (2.52 ± 0.69) %/s and (3.34 ± 00.41) %/s, respectively. Obviously, the SS value of the benign group was significantly lower than that of the malignant group (P < 0.01). The ADC value with different b values in the benign group was significantly lower than that of the malignant group (P < 0.01). It suggested that the SVM-L model significantly improved the quality of lung MRI images and increased the accuracy to differentiate benign and malignant SPN, providing reference for the diagnosis and treatment of SPN patients.

Predicting Lung Cancer Risk of Incidental Solid and Subsolid Pulmonary Nodules in Different Sizes

Objective

Malignancy prediction models for pulmonary nodules are most accurate when used within nodules similar to those in which they were developed. This study was to establish models that respectively predict malignancy risk of incidental solid and subsolid pulmonary nodules of different size.

Materials and Methods

This retrospective study enrolled patients with 5-30 mm pulmonary nodules who had a histopathologic diagnosis of benign or malignant. The median time to lung cancer diagnosis was 25 days. Four training/validation datasets were assembled based on nodule texture and size: subsolid nodules (SSNs) ≤15 mm, SSNs between 15 and 30 mm, solid nodules ≤15 mm and those between 15 and 30 mm. Univariate logistic regression was used to identify potential predictors, and multivariate analysis was used to build four models.

Results

The study identified 1008 benign and 1813 malignant nodules from a single hospital, and by random selection 1008 malignant nodules were enrolled for further analysis. There was a much higher malignancy rate among SSNs than solid nodules (rate, 75% vs 39%, P<0.001). Four distinguishing models were respectively developed and the areas under the curve (AUC) in training sets and validation sets were 0.83 (0.78-0.88) and 0.70 (0.61-0.80) for SSNs ≤15 mm, 0.84 (0.74-0.93) and 0.72 (0.57-0.87) for SSNs between 15 and 30 mm, 0.82 (0.77-0.87) and 0.71 (0.61-0.80) for solid nodules ≤15 mm, 0.82 (0.79-0.85) and 0.81 (0.76-0.86) for solid nodules between 15 and 30 mm. Each model showed good calibration and potential clinical applications. Different independent predictors were identified for solid nodules and SSNs of different size.

Conclusion

We developed four models to help characterize subsolid and solid pulmonary nodules of different sizes. The established models may provide decision-making information for thoracic radiologists and clinicians.

Artificial intelligence: radiologists' expectations and opinions gleaned from a nationwide online survey

PURPOSE

To report the results of a nationwide online survey on artificial intelligence (AI) among radiologist members of the Italian Society of Medical and Interventional Radiology (SIRM).

METHODS AND MATERIALS

All members were invited to the survey as an initiative by the Imaging Informatics Chapter of SIRM. The survey consisted of 13 questions about the participants' demographic information, perceived advantages and issues related to AI implementation in radiological practice, and their overall opinion about AI.

RESULTS

In total, 1032 radiologists (equaling 9.5% of active SIRM members for the year 2019) joined the survey. Perceived AI advantages included a lower diagnostic error rate (750/1027, 73.0%) and optimization of radiologists' work (697/1027, 67.9%). The risk of a poorer professional reputation of radiologists compared with non-radiologists (617/1024, 60.3%), and increased costs and workload due to AI system maintenance and data analysis (399/1024, 39.0%) were seen as potential issues. Most radiologists stated that specific policies should regulate the use of AI (933/1032, 90.4%) and were not afraid of losing their job due to it (917/1032, 88.9%). Overall, 77.0% of respondents (794/1032) were favorable to the adoption of AI, whereas 18.0% (186/1032) were uncertain and 5.0% (52/1032) were unfavorable.

CONCLUSIONS

Radiologists had a mostly positive attitude toward the implementation of AI in their working practice. They were not concerned that AI will replace them, but rather that it might diminish their professional reputation.