Solitary solid pulmonary nodules: a CT-based deep learning nomogram helps differentiate tuberculosis granulomas from lung adenocarcinomas

Prediction of Malignancy and Pathological Types of Solid Lung Nodules on CT Scans Using a Volumetric SWIN Transformer

Lung adenocarcinoma and squamous cell carcinoma are the two most common pathological lung cancer subtypes. Accurate diagnosis and pathological subtyping are crucial for lung cancer treatment. Solitary solid lung nodules with lobulation and spiculation signs are often indicative of lung cancer; however, in some cases, postoperative pathology finds benign solid lung nodules. It is critical to accurately identify solid lung nodules with lobulation and spiculation signs before surgery; however, traditional diagnostic imaging is prone to misdiagnosis, and studies on artificial intelligence-assisted diagnosis are few. Therefore, we introduce a volumetric SWIN Transformer-based method. It is a multi-scale, multi-task, and highly interpretable model for distinguishing between benign solid lung nodules with lobulation and spiculation signs, lung adenocarcinomas, and lung squamous cell carcinoma. The technique's effectiveness was improved by using 3-dimensional (3D) computed tomography (CT) images instead of conventional 2-dimensional (2D) images to combine as much information as possible. The model was trained using 352 of the 441 CT image sequences and validated using the rest. The experimental results showed that our model could accurately differentiate between benign lung nodules with lobulation and spiculation signs, lung adenocarcinoma, and squamous cell carcinoma. On the test set, our model achieves an accuracy of 0.9888, precision of 0.9892, recall of 0.9888, and an F1-score of 0.9888, along with a class activation mapping (CAM) visualization of the 3D model. Consequently, our method could be used as a preoperative tool to assist in diagnosing solitary solid lung nodules with lobulation and spiculation signs accurately and provide a theoretical basis for developing appropriate clinical diagnosis and treatment plans for the patients.

Hypermetabolic pulmonary lesions detection and diagnosis based on PET/CT imaging and deep learning models

PURPOSE

This study aims to develop and evaluate deep learning models for the detection and classification of hypermetabolic lung lesions into four categories: benign, lung cancer, pulmonary lymphoma, and metastasis. These categories are defined by their pathological origin, clinical relevance, and therapeutic implications.

METHODS

A lesion localisation model was first developed using manually annotated PET/CT images. For classification, a multi-dimensional joint network was employed, incorporating both image patches and two-dimensional projections. Classification performance was quantified by metrics like accuracy, and compared to that of a radiomics model. Additionally, false-positive segmentations were manually reviewed and analysed for clinical evaluation.

RESULTS

The study retrospectively included 647 cases (409 males/238 females) over more than 8 years from five centres, divided into an internal dataset (426 cases from Shanghai Ruijin Hospital), an external test set I (151 cases from four other institutions), and an external test set II (70 cases from a new imaging device). The localisation model achieved detection rates of 81.19%, 75.48%, and 77.59% on the internal, external test set I, and external test set II, respectively. The classification model outperformed the radiomics approach, with area-under-curves of 88.4%, 80.7%, and 66.6%, respectively. Most false-positive segmentations were clinically acceptable, corresponding to suspicious lesions in adjacent regions, particularly lymph nodes.

CONCLUSION

Deep learning models based on PET/CT imaging can effectively detect, segment, and classify hypermetabolic lung lesions, and identify suspicious adjacent lesions. These results highlight the potential of artificial intelligence in clinical decision-making and lung disease diagnosis.

Diagnostic Performance of Artificial Intelligence-Based Methods for Tuberculosis Detection: Systematic Review

BACKGROUND

Tuberculosis (TB) remains a significant health concern, contributing to the highest mortality among infectious diseases worldwide. However, none of the various TB diagnostic tools introduced is deemed sufficient on its own for the diagnostic pathway, so various artificial intelligence (AI)-based methods have been developed to address this issue.

OBJECTIVE

We aimed to provide a comprehensive evaluation of AI-based algorithms for TB detection across various data modalities.

METHODS

Following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analysis) 2020 guidelines, we conducted a systematic review to synthesize current knowledge on this topic. Our search across 3 major databases (Scopus, PubMed, Association for Computing Machinery [ACM] Digital Library) yielded 1146 records, of which we included 152 (13.3%) studies in our analysis. QUADAS-2 (Quality Assessment of Diagnostic Accuracy Studies version 2) was performed for the risk-of-bias assessment of all included studies.

RESULTS

Radiographic biomarkers (n=129, 84.9%) and deep learning (DL; n=122, 80.3%) approaches were predominantly used, with convolutional neural networks (CNNs) using Visual Geometry Group (VGG)-16 (n=37, 24.3%), ResNet-50 (n=33, 21.7%), and DenseNet-121 (n=19, 12.5%) architectures being the most common DL approach. The majority of studies focused on model development (n=143, 94.1%) and used a single modality approach (n=141, 92.8%). AI methods demonstrated good performance in all studies: mean accuracy=91.93% (SD 8.10%, 95% CI 90.52%-93.33%; median 93.59%, IQR 88.33%-98.32%), mean area under the curve (AUC)=93.48% (SD 7.51%, 95% CI 91.90%-95.06%; median 95.28%, IQR 91%-99%), mean sensitivity=92.77% (SD 7.48%, 95% CI 91.38%-94.15%; median 94.05% IQR 89%-98.87%), and mean specificity=92.39% (SD 9.4%, 95% CI 90.30%-94.49%; median 95.38%, IQR 89.42%-99.19%). AI performance across different biomarker types showed mean accuracies of 92.45% (SD 7.83%), 89.03% (SD 8.49%), and 84.21% (SD 0%); mean AUCs of 94.47% (SD 7.32%), 88.45% (SD 8.33%), and 88.61% (SD 5.9%); mean sensitivities of 93.8% (SD 6.27%), 88.41% (SD 10.24%), and 93% (SD 0%); and mean specificities of 94.2% (SD 6.63%), 85.89% (SD 14.66%), and 95% (SD 0%) for radiographic, molecular/biochemical, and physiological types, respectively. AI performance across various reference standards showed mean accuracies of 91.44% (SD 7.3%), 93.16% (SD 6.44%), and 88.98% (SD 9.77%); mean AUCs of 90.95% (SD 7.58%), 94.89% (SD 5.18%), and 92.61% (SD 6.01%); mean sensitivities of 91.76% (SD 7.02%), 93.73% (SD 6.67%), and 91.34% (SD 7.71%); and mean specificities of 86.56% (SD 12.8%), 93.69% (SD 8.45%), and 92.7% (SD 6.54%) for bacteriological, human reader, and combined reference standards, respectively. The transfer learning (TL) approach showed increasing popularity (n=89, 58.6%). Notably, only 1 (0.7%) study conducted domain-shift analysis for TB detection.

CONCLUSIONS

Findings from this review underscore the considerable promise of AI-based methods in the realm of TB detection. Future research endeavors should prioritize conducting domain-shift analyses to better simulate real-world scenarios in TB detection.

TRIAL REGISTRATION

PROSPERO CRD42023453611; https://www.crd.york.ac.uk/PROSPERO/view/CRD42023453611.

General lightweight framework for vision foundation model supporting multi-task and multi-center medical image analysis

The foundation model, trained on extensive and diverse datasets, has shown strong performance across numerous downstream tasks. Nevertheless, its application in the medical domain is significantly hindered by issues such as data volume, heterogeneity, and privacy concerns. Therefore, we propose the Vision Foundation Model General Lightweight (VFMGL) framework, which facilitates the decentralized construction of expert clinical models for various medical tasks. The VFMGL framework transfers general knowledge from large-parameter vision foundation models to construct lightweight, robust expert clinical models tailored to specific medical tasks. Through extensive experiments and analyses across a range of medical tasks and scenarios, we demonstrate that VFMGL achieves superior performance in both medical image classification and segmentation tasks, effectively managing the challenges posed by data heterogeneity. These results underscore the potential of VFMGL in advancing the efficacy and reliability of AI-driven medical diagnostics.

Revolutionizing diagnosis of pulmonary Mycobacterium tuberculosis based on CT: a systematic review of imaging analysis through deep learning

Introduction

The mortality rate associated with Mycobacterium tuberculosis (MTB) has seen a significant rise in regions heavily affected by the disease over the past few decades. The traditional methods for diagnosing and differentiating tuberculosis (TB) remain thorny issues, particularly in areas with a high TB epidemic and inadequate resources. Processing numerous images can be time-consuming and tedious. Therefore, there is a need for automatic segmentation and classification technologies based on lung computed tomography (CT) scans to expedite and enhance the diagnosis of TB, enabling the rapid and secure identification of the condition. Deep learning (DL) offers a promising solution for automatically segmenting and classifying lung CT scans, expediting and enhancing TB diagnosis.

Methods

This review evaluates the diagnostic accuracy of DL modalities for diagnosing pulmonary tuberculosis (PTB) after searching the PubMed and Web of Science databases using the preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines.

Results

Seven articles were found and included in the review. While DL has been widely used and achieved great success in CT-based PTB diagnosis, there are still challenges to be addressed and opportunities to be explored, including data scarcity, model generalization, interpretability, and ethical concerns. Addressing these challenges requires data augmentation, interpretable models, moral frameworks, and clinical validation.

Conclusion

Further research should focus on developing robust and generalizable DL models, enhancing model interpretability, establishing ethical guidelines, and conducting clinical validation studies. DL holds great promise for transforming PTB diagnosis and improving patient outcomes.

Value of CT-Based Deep Learning Model in Differentiating Benign and Malignant Solid Pulmonary Nodules ≤ 8 mm

RATIONALE AND OBJECTIVES

We examined the effectiveness of computed tomography (CT)-based deep learning (DL) models in differentiating benign and malignant solid pulmonary nodules (SPNs) ≤ 8 mm.

MATERIALS AND METHODS

The study patients (n = 719) were divided into internal training, internal validation, and external validation cohorts; all had small SPNs and had undergone preoperative chest CTs and surgical resection. We developed five DL models incorporating features of the nodule and five different peri-nodular regions with the Multiscale Dual Attention Network (MDANet) to differentiate benign and malignant SPNs. We selected the best-performing model, which was then compared to four conventional algorithms (VGG19, ResNet50, ResNeXt50, and DenseNet121). Furthermore, another five DL models were constructed using MDANet to distinguish benign tumors from inflammatory nodules and the one performed best was selected out.

RESULTS

Model 4, which incorporated the nodule and 15 mm peri-nodular region, best differentiated benign and malignant SPNs. The model had an area under the curve (AUC), accuracy, recall, precision, and F1-score of 0.730, 0.724, 0.711, 0.705, and 0.707 in the external validation cohort. Model 4 also performed better than the other four conventional algorithms. Model 8, which incorporated the nodule and 10 mm peri-nodular region, was the best model for distinguishing benign tumors from inflammatory nodules. The model had an AUC, accuracy, recall, precision, and F1-score of 0.871, 0.938, 0.863, 0.904, and 0.882 in the external validation cohort.

CONCLUSION

The study concludes that CT-based DL models built with MDANet can accurately discriminate among small benign and malignant SPNs, benign tumors and inflammatory nodules.

A CT-based radiomics analyses for differentiating drug‑resistant and drug-sensitive pulmonary tuberculosis

BACKGROUND

To explore the value of computed tomography based radiomics in the differential diagnosis of drug-sensitive and drug-resistant pulmonary tuberculosis.

METHODS

The clinical and computed tomography image data of 177 patients who were diagnosed with pulmonary tuberculosis through sputum culture and completed drug-susceptibility testing from April 2018 to December 2020 at the Second Hospital of Nanjing were retrospectively analyzed. Patients with drug-resistant pulmonary tuberculosis (n = 78) and drug-sensitive pulmonary tuberculosis (n = 99) were randomly divided into a training set (n = 124) and a validation set (n = 53) at a ratio of 7:3. Regions of interest were drawn to delineate the lesions and radiomics features were extracted from non-contrast computed tomography images. A radiomics signature based on the valuable radiomics features was constructed and a radiomics score was calculated. Demographic data, clinical symptoms, laboratory results and computed tomography imaging characteristics were evaluated to establish a clinical model. Combined with the Rad-score and clinical factors, a radiomics-clinical model nomogram was constructed.

RESULTS

Thirteen features were used to construct the radiomics signature. The radiomics signature showed good discrimination in the training set (area under the curve (AUC), 0.891; 95% confidence interval (CI), 0.832-0.951) and the validation set (AUC, 0.803; 95% CI, 0.674-0.932). In the clinical model, the AUC of the training set was 0.780(95% CI, 0.700-0.859), while the AUC of the validation set was 0.692 (95% CI, 0.546-0.839). The radiomics-clinical model showed good calibration and discrimination in the training set (AUC, 0.932;95% CI, 0.888-0.977) and the validation set (AUC, 0.841; 95% CI, 0.719-0.962).

CONCLUSIONS

Simple radiomics signature is of great value in differentiating drug-sensitive and drug-resistant pulmonary tuberculosis patients. The radiomics-clinical model nomogram showed good predictive, which may help clinicians formulate precise treatments.

A CT-based radiomics predictive nomogram to identify pulmonary tuberculosis from community-acquired pneumonia: a multicenter cohort study

Purpose

To develop a predictive nomogram based on computed tomography (CT) radiomics to distinguish pulmonary tuberculosis (PTB) from community-acquired pneumonia (CAP).

Methods

A total of 195 PTB patients and 163 CAP patients were enrolled from three hospitals. It is divided into a training cohort, a testing cohort and validation cohort. Clinical models were established by using significantly correlated clinical features. Radiomics features were screened by the least absolute shrinkage and selection operator (LASSO) algorithm. Radiomics scores (Radscore) were calculated from the formula of radiomics features. Clinical radiomics conjoint nomogram was established according to Radscore and clinical features, and the diagnostic performance of the model was evaluated by receiver operating characteristic (ROC) curve analysis.

Results

Two clinical features and 12 radiomic features were selected as optimal predictors for the establishment of clinical radiomics conjoint nomogram. The results showed that the predictive nomogram had an outstanding ability to discriminate between the two diseases, and the AUC of the training cohort was 0.947 (95% CI, 0.916-0.979), testing cohort was 0.888 (95% CI, 0.814-0.961) and that of the validation cohort was 0.850 (95% CI, 0.778-0.922). Decision curve analysis (DCA) indicated that the nomogram has outstanding clinical value.

Conclusions

This study developed a clinical radiomics model that uses radiomics features to identify PTB from CAP. This model provides valuable guidance to clinicians in identifying PTB.

AI-Driven Thoracic X-ray Diagnostics: Transformative Transfer Learning for Clinical Validation in Pulmonary Radiography

Our research evaluates advanced artificial (AI) methodologies to enhance diagnostic accuracy in pulmonary radiography. Utilizing DenseNet121 and ResNet50, we analyzed 108,948 chest X-ray images from 32,717 patients and DenseNet121 achieved an area under the curve (AUC) of 94% in identifying the conditions of pneumothorax and oedema. The model's performance surpassed that of expert radiologists, though further improvements are necessary for diagnosing complex conditions such as emphysema, effusion, and hernia. Clinical validation integrating Latent Dirichlet Allocation (LDA) and Named Entity Recognition (NER) demonstrated the potential of natural language processing (NLP) in clinical workflows. The NER system achieved a precision of 92% and a recall of 88%. Sentiment analysis using DistilBERT provided a nuanced understanding of clinical notes, which is essential for refining diagnostic decisions. XGBoost and SHapley Additive exPlanations (SHAP) enhanced feature extraction and model interpretability. Local Interpretable Model-agnostic Explanations (LIME) and occlusion sensitivity analysis further enriched transparency, enabling healthcare providers to trust AI predictions. These AI techniques reduced processing times by 60% and annotation errors by 75%, setting a new benchmark for efficiency in thoracic diagnostics. The research explored the transformative potential of AI in medical imaging, advancing traditional diagnostics and accelerating medical evaluations in clinical settings.

An exploratory deep learning approach to investigate tuberculosis pathogenesis in nonhuman primate model: Combining automated radiological analysis with clinical and biomarkers data

BACKGROUND

Tuberculosis (TB) kills approximately 1.6 million people yearly despite the fact anti-TB drugs are generally curative. Therefore, TB-case detection and monitoring of therapy, need a comprehensive approach. Automated radiological analysis, combined with clinical, microbiological, and immunological data, by machine learning (ML), can help achieve it.

METHODS

Six rhesus macaques were experimentally inoculated with pathogenic Mycobacterium tuberculosis in the lung. Data, including Computed Tomography (CT), were collected at 0, 2, 4, 8, 12, 16, and 20 weeks.

RESULTS

Our ML-based CT analysis (TB-Net) efficiently and accurately analyzed disease progression, performing better than standard deep learning model (LLM OpenAI's CLIP Vi4). TB-Net based results were more consistent than, and confirmed independently by, blinded manual disease scoring by two radiologists and exhibited strong correlations with blood biomarkers, TB-lesion volumes, and disease-signs during disease pathogenesis.

CONCLUSION

The proposed approach is valuable in early disease detection, monitoring efficacy of therapy, and clinical decision making.

Differentiation of granulomatous nodules with lobulation and spiculation signs from solid lung adenocarcinomas using a CT deep learning model

BACKGROUND

The diagnosis of solitary pulmonary nodules has always been a difficult and important point in clinical research, especially granulomatous nodules (GNs) with lobulation and spiculation signs, which are easily misdiagnosed as malignant tumors. Therefore, in this study, we utilised a CT deep learning (DL) model to distinguish GNs with lobulation and spiculation signs from solid lung adenocarcinomas (LADCs), to improve the diagnostic accuracy of preoperative diagnosis.

METHODS

420 patients with pathologically confirmed GNs and LADCs from three medical institutions were retrospectively enrolled. The regions of interest in non-enhanced CT (NECT) and venous contrast-enhanced CT (VECT) were identified and labeled, and self-supervised labels were constructed. Cases from institution 1 were randomly divided into a training set (TS) and an internal validation set (IVS), and cases from institutions 2 and 3 were treated as an external validation set (EVS). Training and validation were performed using self-supervised transfer learning, and the results were compared with the radiologists' diagnoses.

RESULTS

The DL model achieved good performance in distinguishing GNs and LADCs, with area under curve (AUC) values of 0.917, 0.876, and 0.896 in the IVS and 0.889, 0.879, and 0.881 in the EVS for NECT, VECT, and non-enhanced with venous contrast-enhanced CT (NEVECT) images, respectively. The AUCs of radiologists 1, 2, 3, and 4 were, respectively, 0.739, 0.783, 0.883, and 0.901 in the (IVS) and 0.760, 0.760, 0.841, and 0.844 in the EVS.

CONCLUSIONS

A CT DL model showed great value for preoperative differentiation of GNs with lobulation and spiculation signs from solid LADCs, and its predictive performance was higher than that of radiologists.

The diagnostic value of CT-based radiomics nomogram for solitary indeterminate smoothly marginated solid pulmonary nodules

Objectives

This study aimed to explore the value of radiomics nomogram based on computed tomography (CT) on the diagnosis of benign and malignant solitary indeterminate smoothly marginated solid pulmonary nodules (SMSPNs).

Methods

This study retrospectively reviewed 205 cases with solitary indeterminate SMSPNs on CT, including 112 cases of benign nodules and 93 cases of malignant nodules. They were divided into training (n=143) and validation (n=62) cohorts based on different CT scanners. Radiomics features of the nodules were extracted from the lung window CT images. The variance threshold method, SelectKBest, and least absolute shrinkage and selection operator were used to select the key radiomics features to construct the rad-score. Through multivariate logistic regression analysis, a nomogram was built by combining rad-score, clinical factors, and CT features. The nomogram performance was evaluated by the area under the receiver operating characteristic curve (AUC).

Results

A total of 19 radiomics features were selected to construct the rad-score, and the nomogram was constructed by the rad-score, one clinical factor (history of malignant tumor), and three CT features (including calcification, pleural retraction, and lobulation). The nomogram performed better than the radiomics model, clinical model, and experienced radiologists who specialized in thoracic radiology for nodule diagnosis. The AUC values of the nomogram were 0.942 in the training cohort and 0.933 in the validation cohort. The calibration curve and decision curve showed that the nomogram demonstrated good consistency and clinical applicability.

Conclusion

The CT-based radiomics nomogram achieved high efficiency in the preoperative diagnosis of solitary indeterminate SMSPNs, and it is of great significance in guiding clinical decision-making.

CRDet: A circle representation detector for lung granulomas based on multi-scale attention features with center point calibration

Lung granuloma is a very common lung disease, and its specific diagnosis is important for determining the exact cause of the disease as well as the prognosis of the patient. And, an effective lung granuloma detection model based on computer-aided diagnostics (CAD) can help pathologists to localize granulomas, thereby improving the efficiency of the specific diagnosis. However, for lung granuloma detection models based on CAD, the significant size differences between granulomas and how to better utilize the morphological features of granulomas are both critical challenges to be addressed. In this paper, we propose an automatic method CRDet to localize granulomas in histopathological images and deal with these challenges. We first introduce the multi-scale feature extraction network with self-attention to extract features at different scales at the same time. Then, the features will be converted to circle representations of granulomas by circle representation detection heads to achieve the alignment of features and ground truth. In this way, we can also more effectively use the circular morphological features of granulomas. Finally, we propose a center point calibration method at the inference stage to further optimize the circle representation. For model evaluation, we built a lung granuloma circle representation dataset named LGCR, including 288 images from 50 subjects. Our method yielded 0.316 mAP and 0.571 mAR, outperforming the state-of-the-art object detection methods on our proposed LGCR.

Deep learning model based on multi-lesion and time series CT images for predicting the benefits from anti-HER2 targeted therapy in stage IV gastric cancer

OBJECTIVE

To develop and validate a deep learning model based on multi-lesion and time series CT images in predicting overall survival (OS) in patients with stage IV gastric cancer (GC) receiving anti-HER2 targeted therapy.

METHODS

A total of 207 patients were enrolled in this multicenter study, with 137 patients for retrospective training and internal validation, 33 patients for prospective validation, and 37 patients for external validation. All patients received anti-HER2 targeted therapy and underwent pre- and post-treatment CT scans (baseline and at least one follow-up). The proposed deep learning model evaluated the multiple lesions in time series CT images to predict risk probabilities. We further evaluated and validated the risk score of the nomogram combining a two-follow-up lesion-based deep learning model (LDLM-2F), tumor markers, and clinical information for predicting the benefits from treatment (Nomo-LDLM-2F).

RESULTS

In the internal validation and prospective cohorts, the one-year AUCs for Nomo-LDLM-2F using the time series medical images and tumor markers were 0.894 (0.728-1.000) and 0.809 (0.561-1.000), respectively. In the external validation cohort, the one-year AUC of Nomo-LDLM-2F without tumor markers was 0.771 (0.510-1.000). Patients with a low Nomo-LDLM-2F score derived survival benefits from anti-HER2 targeted therapy significantly compared to those with a high Nomo-LDLM-2F score (all p < 0.05).

CONCLUSION

The Nomo-LDLM-2F score derived from multi-lesion and time series CT images holds promise for the effective readout of OS probability in patients with HER2-positive stage IV GC receiving anti-HER2 therapy.

CRITICAL RELEVANCE STATEMENT

The deep learning model using baseline and early follow-up CT images aims to predict OS in patients with stage IV gastric cancer receiving anti-HER2 targeted therapy. This model highlights the spatiotemporal heterogeneity of stage IV GC, assisting clinicians in the early evaluation of the efficacy of anti-HER2 therapy.

KEY POINTS

• Multi-lesion and time series model revealed the spatiotemporal heterogeneity in anti-HER2 therapy. • The Nomo-LDLM-2F score was a valuable prognostic marker for anti-HER2 therapy. • CT-based deep learning model incorporating time-series tumor markers improved performance.

Deep learning vs. robust federal learning for distinguishing adrenal metastases from benign lesions with multi-phase CT images

Background

Differentiating adrenal adenomas from metastases poses a significant challenge, particularly in patients with a history of extra-adrenal malignancy. This study investigates the performance of three-phase computed tomography (CT) based robust federal learning algorithm and traditional deep learning for distinguishing metastases from benign adrenal lesions.

Material and methods

This retrospective analysis includes 1187 instances who underwent three-phase CT scans between January 2008 and March 2021, comprising 720 benign lesions and 467 metastases. Utilizing the three-phase CT images, both a Robust Federal Learning Signature (RFLS) and a traditional Deep Learning Signature (DLS) were constructed using the Least Absolute Shrinkage and Selection Operator (LASSO) logistic regression. Their diagnostic capabilities were subsequently validated and compared using metrics such as the Areas Under the Receiver Operating Curve (AUCs), Net Reclassification Improvement (NRI), and Decision Curve Analysis (DCA).

Results

Compared with DLS, the RFLS showed better capability in distinguishing metastases from benign adrenal lesions (average AUC: 0.816 vs.0.798, NRI = 0.126, P < 0.072; 0.889 vs.0.838, NRI = 0.209, P < 0.001; 0.903 vs.0.825, NRI = 0.643, p < 0.001) in the four-testing cohort, respectively. DCA showed that the RFLS added more net benefit than DLS for clinical utility. Moreover, Comparison with state-of-the-art federal learning methods, the results once again confirmed that the RFLS significantly improved the diagnostic performance based on three-phase CT (AUC: AP, 0.727 vs. 0.757 vs. 0.739 vs. 0.796; PCP, 0.781 vs. 0.851 vs. 0.790 vs. 0.882; VP, 0.789 vs. 0.814 vs. 0.779 vs. 0.886).

Conclusion

RFLS was superior to DLS for preoperative distinguishing metastases from benign adrenal lesions with multi-phase CT Images.

[The contribution of artificial intelligence (AI) subsequent to the processing of thoracic imaging]

The contribution of artificial intelligence (AI) to medical imaging is currently the object of widespread experimentation. The development of deep learning (DL) methods, particularly convolution neural networks (CNNs), has led to performance gains often superior to those achieved by conventional methods such as machine learning. Radiomics is an approach aimed at extracting quantitative data not accessible to the human eye from images expressing a disease. The data subsequently feed machine learning models and produce diagnostic or prognostic probabilities. As for the multiple applications of AI methods in thoracic imaging, they are undergoing evaluation. Chest radiography is a practically ideal field for the development of DL algorithms able to automatically interpret X-rays. Current algorithms can detect up to 14 different abnormalities present either in isolation or in combination. Chest CT is another area offering numerous AI applications. Various algorithms have been specifically formed and validated for the detection and characterization of pulmonary nodules and pulmonary embolism, as well as segmentation and quantitative analysis of the extent of diffuse lung diseases (emphysema, infectious pneumonias, interstitial lung disease). In addition, the analysis of medical images can be associated with clinical, biological, and functional data (multi-omics analysis), the objective being to construct predictive approaches regarding disease prognosis and response to treatment.

A CT-based deep learning model: visceral pleural invasion and survival prediction in clinical stage IA lung adenocarcinoma

Pathologic visceral pleural invasion (VPI) in patients with early-stage lung cancer can result in the upstaging of T1 to T2, in addition to having implications for surgical resection and prognostic outcomes. This study was designed with the goal of establishing and validating a CT-based deep learning (DL) model capable of predicting VPI status and stratifying patients based on their prognostic outcomes. In total, 2077 patients from three centers with pathologically confirmed clinical stage IA lung adenocarcinoma were enrolled. DL signatures were extracted with a 3D residual neural network. DL model was able to effectively predict VPI status. VPI predicted by the DL models, as well as pathologic VPI, was associated with shorter disease-free survival. The established deep learning signature provides a tool capable of aiding the accurate prediction of VPI in patients with clinical stage IA lung adenocarcinoma, thus enabling prognostic stratification.

The classification of benign and malignant lung nodules based on CT radiomics: a systematic review, quality score assessment, and meta-analysis

Radiomics methods are increasingly used to identify benign and malignant lung nodules, and early monitoring is essential in prognosis and treatment strategy formulation. To evaluate the diagnostic performance of computed tomography (CT)-based radiomics for distinguishing between benign and malignant lung nodules by performing a meta-analysis. Between January 2000 and December 2021, we searched the PubMed and Embase electronic databases for studies in English. Studies were included if they demonstrated the sensitivity and specificity of CT-based radiomics for diagnosing benign and malignant lung nodules. The studies were evaluated using the QUADAS-2 and radiomics quality scores (RQS). The inhomogeneity of the data and publishing bias were also evaluated. Some subgroup analyses were performed to investigate the impact of diagnostic efficiency. The Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) Guidelines were followed for this meta-analysis. A total of 20 studies involving 3793 patients were included. The combined sensitivity, specificity, diagnostic odds ratio, and area under the summary receiver operating characteristic curve based on CT radiomics diagnosis of benign and malignant lung nodules were 0.81, 0.86, 27.00, and 0.91, respectively. Deek's funnel plot asymmetry test confirmed no significant publication bias in all studies. Fagan nomograms showed a 40% increase in post-test probability among pretest-positive patients. Current evidence shows that CT-based radiomics has high accuracy in the diagnosis of benign and malignant lung nodules.

A novel fusion algorithm for benign-malignant lung nodule classification on CT images

The accurate recognition of malignant lung nodules on CT images is critical in lung cancer screening, which can offer patients the best chance of cure and significant reductions in mortality from lung cancer. Convolutional Neural Network (CNN) has been proven as a powerful method in medical image analysis. Radiomics which is believed to be of interest based on expert opinion can describe high-throughput extraction from CT images. Graph Convolutional Network explores the global context and makes the inference on both graph node features and relational structures. In this paper, we propose a novel fusion algorithm, RGD, for benign-malignant lung nodule classification by incorporating Radiomics study and Graph learning into the multiple Deep CNNs to form a more complete and distinctive feature representation, and ensemble the predictions for robust decision-making. The proposed method was conducted on the publicly available LIDC-IDRI dataset in a 10-fold cross-validation experiment and it obtained an average accuracy of 93.25%, a sensitivity of 89.22%, a specificity of 95.82%, precision of 92.46%, F1 Score of 0.9114 and AUC of 0.9629. Experimental results illustrate that the RGD model achieves superior performance compared with the state-of-the-art methods. Moreover, the effectiveness of the fusion strategy has been confirmed by extensive ablation studies. In the future, the proposed model which performs well on the pulmonary nodule classification on CT images will be applied to increase confidence in the clinical diagnosis of lung cancer.

Preoperative predicting invasiveness of lung adenocarcinoma manifesting as ground-glass nodules based on multimodal images of dual-layer spectral detector CT radiomics models

OBJECTIVE

To construct and validate conventional and radiomics models based on dual-layer spectral CT radiomics for preoperative prediction of lung ground glass nodules (GGNs) invasiveness.

MATERIALS AND METHODS

A retrospective study was conducted on 176 GGNs patients who underwent chest non-contrast enhancement scan on dual-layer spectral detector CT at our hospital within 2 weeks before surgery. Patients were randomized into the training cohort and testing cohort. Clinical features, imaging features and spectral quantitative parameters were collected to establish a conventional model. Radiomics models were established by extracting 1781 radiomics features form regions of interest of each spectral image [120 kVp poly energetic images (PI), 60 keV images and electron density maps], respectively. After selecting the optimal radiomic features and integrating multiple machine learning models, the conventional model, PI model, 60 keV model, electron density (ED) model and combined model based on multimodal spectral images were finally established. The performance of these models was assessed through the evaluation of discrimination, calibration, and clinical application.

RESULTS

In the conventional model, age, vacuole sign, 60 keV and ED were independent risk factors of invasiveness. The combined model using logistic regression-least absolute shrinkage and selection operator classifiers was the optimal model with a higher area under the curve of the training (0.961, 95% confidence interval, CI: 0.932-0.991) and testing set (0.944, 0.890-0.999).

CONCLUSION

The combined models are helpful to predict the invasiveness of GGNs before surgery and guide the individualized treatment of patients.

Deep Features from Pretrained Networks Do Not Outperform Hand-Crafted Features in Radiomics

In radiomics, utilizing features extracted from pretrained deep networks could result in models with a higher predictive performance than those relying on hand-crafted features. This study compared the predictive performance of models trained with either deep features, hand-crafted features, or a combination of these features in terms of the area under the receiver-operating characteristic curve (AUC) and other metrics. We trained models on ten radiological datasets using five feature selection methods and three classifiers. Our results indicate that models based on deep features did not show an improved AUC compared to those utilizing hand-crafted features (deep: AUC 0.775, hand-crafted: AUC 0.789; p = 0.28). Including morphological features alongside deep features led to overall improvements in prediction performance for all models (+0.02 gain in AUC; p < 0.001); however, the best model did not benefit from this (+0.003 gain in AUC; p = 0.57). Using all hand-crafted features in addition to the deep features resulted in a further overall improvement (+0.034 in AUC; p < 0.001), but only a minor improvement could be observed for the best model (deep: AUC 0.798, hand-crafted: AUC 0.789; p = 0.92). Furthermore, our results show that models based on deep features extracted from networks pretrained on medical data have no advantage in predictive performance over models relying on features extracted from networks pretrained on ImageNet data. Our study contributes a benchmarking analysis of models trained on hand-crafted and deep features from pretrained networks across multiple datasets. It also provides a comprehensive understanding of their applicability and limitations in radiomics. Our study shows, in conclusion, that models based on features extracted from pretrained deep networks do not outperform models trained on hand-crafted ones.

Artificial Intelligence in Paediatric Tuberculosis

Tuberculosis (TB) continues to be a leading cause of death in children despite global efforts focused on early diagnosis and interventions to limit the spread of the disease. This challenge has been made more complex in the context of the coronavirus pandemic, which has disrupted the "End TB Strategy" and framework set out by the World Health Organization (WHO). Since the inception of artificial intelligence (AI) more than 60 years ago, the interest in AI has risen and more recently we have seen the emergence of multiple real-world applications, many of which relate to medical imaging. Nonetheless, real-world AI applications and clinical studies are limited in the niche area of paediatric imaging. This review article will focus on how AI, or more specifically deep learning, can be applied to TB diagnosis and management in children. We describe how deep learning can be utilised in chest imaging to provide computer-assisted diagnosis to augment workflow and screening efforts. We also review examples of recent AI applications for TB screening in resource constrained environments and we explore some of the challenges and the future directions of AI in paediatric TB.

Deep Learning Radiomics of Preoperative Breast MRI for Prediction of Axillary Lymph Node Metastasis in Breast Cancer

The objective of this study is to develop a radiomic signature constructed from deep learning features and a nomogram for prediction of axillary lymph node metastasis (ALNM) in breast cancer patients. Preoperative magnetic resonance imaging data from 479 breast cancer patients with 488 lesions were studied. The included patients were divided into two cohorts by time (training/testing cohort, n = 366/122). Deep learning features were extracted from diffusion-weighted imaging-quantitatively measured apparent diffusion coefficient (DWI-ADC) imaging and dynamic contrast-enhanced MRI (DCE-MRI) by a pretrained neural network of DenseNet121. After the selection of both radiomic and clinicopathological features, deep learning signature and a nomogram were built for independent validation. Twenty-three deep learning features were automatically selected in the training cohort to establish the deep learning signature of ALNM. Three clinicopathological factors, including LN palpability (odds ratio (OR) = 6.04; 95% confidence interval (CI) = 3.06-12.54, P = 0.004), tumor size in MRI (OR = 1.45, 95% CI = 1.18-1.80, P = 0.104), and Ki-67 (OR = 1.01; 95% CI = 1.00-1.02, P = 0.099), were selected and combined with radiomic signature to build a combined nomogram. The nomogram showed excellent predictive ability for ALNM (AUC 0.80 and 0.71 in training and testing cohorts, respectively). The sensitivity, specificity, and accuracy were 65%, 80%, and 75%, respectively, in the testing cohort. MRI-based deep learning radiomics in patients with breast cancer could be used to predict ALNM, providing a noninvasive approach to structuring the treatment strategy.

Spectral CT-based radiomics signature for distinguishing malignant pulmonary nodules from benign

OBJECTIVES

To evaluate the discriminatory capability of spectral CT-based radiomics to distinguish benign from malignant solitary pulmonary solid nodules (SPSNs).

MATERIALS AND METHODS

A retrospective study was performed including 242 patients with SPSNs who underwent contrast-enhanced dual-layer Spectral Detector CT (SDCT) examination within one month before surgery in our hospital, which were randomly divided into training and testing datasets with a ratio of 7:3. Regions of interest (ROIs) based on 40-65 keV images of arterial phase (AP), venous phases (VP), and 120kVp of SDCT were delineated, and radiomics features were extracted. Then the optimal radiomics-based score in identifying SPSNs was calculated and selected for building radiomics-based model. The conventional model was developed based on significant clinical characteristics and spectral quantitative parameters, subsequently, the integrated model combining radiomics-based model and conventional model was established. The performance of three models was evaluated with discrimination, calibration, and clinical application.

RESULTS

The 65 keV radiomics-based scores of AP and VP had the optimal performance in distinguishing benign from malignant SPSNs (AUC_65keV-AP = 0.92, AUC_65keV-VP = 0.88). The diagnostic efficiency of radiomics-based model (AUC = 0.96) based on 65 keV images of AP and VP outperformed conventional model (AUC = 0.86) in the identification of SPSNs, and that of integrated model (AUC = 0.97) was slightly further improved. Evaluation of three models showed the potential for generalizability.

CONCLUSIONS

Among the 40-65 keV radiomics-based scores based on SDCT, 65 keV radiomics-based score had the optimal performance in distinguishing benign from malignant SPSNs. The integrated model combining radiomics-based model based on 65 keV images of AP and VP with Z_eff-AP was significantly superior to conventional model in the discrimination of SPSNs.

Diagnostic Accuracy of the Artificial Intelligence Methods in Medical Imaging for Pulmonary Tuberculosis: A Systematic Review and Meta-Analysis

Tuberculosis (TB) remains one of the leading causes of death among infectious diseases worldwide. Early screening and diagnosis of pulmonary tuberculosis (PTB) is crucial in TB control, and tend to benefit from artificial intelligence. Here, we aimed to evaluate the diagnostic efficacy of a variety of artificial intelligence methods in medical imaging for PTB. We searched MEDLINE and Embase with the OVID platform to identify trials published update to November 2022 that evaluated the effectiveness of artificial-intelligence-based software in medical imaging of patients with PTB. After data extraction, the quality of studies was assessed using quality assessment of diagnostic accuracy studies 2 (QUADAS-2). Pooled sensitivity and specificity were estimated using a bivariate random-effects model. In total, 3987 references were initially identified and 61 studies were finally included, covering a wide range of 124,959 individuals. The pooled sensitivity and the specificity were 91% (95% confidence interval (CI), 89-93%) and 65% (54-75%), respectively, in clinical trials, and 94% (89-96%) and 95% (91-97%), respectively, in model-development studies. These findings have demonstrated that artificial-intelligence-based software could serve as an accurate tool to diagnose PTB in medical imaging. However, standardized reporting guidance regarding AI-specific trials and multicenter clinical trials is urgently needed to truly transform this cutting-edge technology into clinical practice.

The diagnostic and prognostic value of radiomics and deep learning technologies for patients with solid pulmonary nodules in chest CT images

BACKGROUND

Solid pulmonary nodules are different from subsolid nodules and the diagnosis is much more challenging. We intended to evaluate the diagnostic and prognostic value of radiomics and deep learning technologies for solid pulmonary nodules.

METHODS

Retrospectively enroll patients with pathologically-confirmed solid pulmonary nodules and collect clinical data. Obtain pre-treatment high-resolution thoracic CT and manually delineate the nodule in 3D. Then, all patients were randomly divided into training and testing sets at a ratio of 7:3, and convolutional neural networks (CNN) models and random forest (RF) models were established. Survival analyses were performed for patients with solid adenocarcinomas.

RESULTS

Totally 720 solid pulmonary nodules were enrolled, 348 benign and 372 malignant. The CNN model with clinical features achieved the highest AUC [0.819, 95% confidence interval (CI): 0.760-0.877] with a sensitivity of 0.778, specificity of 0.788 and accuracy of 0.783. No significant differences were observed between the CNN and radiomics models. There were 295 solid adenocarcinomas in survival analysis. Different disease-free survival was observed between the low-risk and high-risk groups divided according to the radiomics Rad-score. However, the groups based on deep learning signatures showed similar survival. Cox regression analysis indicated that the radiomics Rad-score (hazard ratio: 5.08, 95% CI: 2.61-9.90) was an independent predictor of recurrence.

CONCLUSIONS

The radiomics and deep learning models can well predict the malignancy of solid pulmonary nodules. Radiomics signatures also demonstrate prognostic value in solid adenocarcinomas.

Classification of tumor from computed tomography images: A brain-inspired multisource transfer learning under probability distribution adaptation

Preoperative diagnosis of gastric cancer and primary gastric lymphoma is challenging and has important clinical significance. Inspired by the inductive reasoning learning of the human brain, transfer learning can improve diagnosis performance of target task by utilizing the knowledge learned from the other domains (source domain). However, most studies focus on single-source transfer learning and may lead to model performance degradation when a large domain shift exists between the single-source domain and target domain. By simulating the multi-modal information learning and transfer mechanism of human brain, this study designed a multisource transfer learning feature extraction and classification framework, which can enhance the prediction performance of the target model by using multisource medical data (domain). First, this manuscript designs a feature extraction network that takes the maximum mean difference based on the Wasserstein distance as an adaptive measure of probability distribution and extracts the domain-specific invariant representations between source and target domain data. Then, aiming at the random generation of parameters bringing uncertainties to prediction accuracy and generalization ability of extreme learning machine network, the 1-norm regularization is used to implement sparse constraints of the output weight matrix and improve the robustness of the model. Finally, some experiments are carried out on the data of two medical centers. The experimental results show that the area under curves (AUCs) of the method are 0.958 and 0.929 in the two validation cohorts, respectively. The method in this manuscript can provide doctors with a better diagnostic reference, which has certain practical significance.

Radiomics combined with clinical features in distinguishing non-calcifying tuberculosis granuloma and lung adenocarcinoma in small pulmonary nodules

Aim

To evaluate the performance of radiomics models with the combination of clinical features in distinguishing non-calcified tuberculosis granuloma (TBG) and lung adenocarcinoma (LAC) in small pulmonary nodules.

Methodology

We conducted a retrospective analysis of 280 patients with pulmonary nodules confirmed by surgical biopsy from January 2017 to December 2020. Samples were divided into LAC group (n = 143) and TBG group (n = 137). We assigned them to a training dataset (n = 196) and a testing dataset (n = 84). Clinical features including gender, age, smoking, CT appearance (size, location, spiculated sign, lobulated shape, vessel convergence, and pleural indentation) were extracted and included in the radiomics models. 3D slicer and FAE software were used to delineate the Region of Interest (ROI) and extract clinical features. The performance of the model was evaluated by the Area Under the Receiver Operating Characteristic (ROC) Curve (AUC).

Results

Based on the model selection, clinical features gender, and age in the LAC group and TBG group showed a significant difference in both datasets (P < 0.05). CT appearance lobulated shape was also significantly different in the LAC group and TBG group (Training dataset, P = 0.034; Testing dataset, P = 0.030). AUC were 0.8344 (95% CI [0.7712-0.8872]) and 0.751 (95% CI [0.6382-0.8531]) in training and testing dataset, respectively.

Conclusion

With the capacity to detect differences between TBG and LAC based on their clinical features, radiomics models with a combined of clinical features may function as the potential non-invasive tool for distinguishing TBG and LAC in small pulmonary nodules.

Clinical and Computed Tomography Characteristics of Solitary Pulmonary Nodules Caused by Fungi: A Comparative Study

Purpose

To clarify the clinical and computed tomography (CT) indicators in distinguishing pulmonary nodules caused by fungal infection from lung cancers.

Methods

From January 2013 to April 2022, 68 patients with solitary fungal nodules (64 were solid and 4 were mixed ground-glass nodules) and 140 cases with solid cancerous nodules with similar size were enrolled. Their clinical characteristics and CT manifestations of the solid nodules were summarized and compared, respectively.

Results

Compared with patients with lung cancers, cases were younger (51.2 ± 11.5 vs 61.3 ± 10.2 years) and non-smokers (72.1% vs 57.9%) and immunocompromised (44.1% vs 17.9%) individuals were more common in patients with fungal nodules (each P < 0.05). The air crescent sign (ACS) (34.4% vs 0%), halo sign (HS) (23.4% vs 4.3%), and satellite lesions (45.3% vs 2.9%) were more frequently detected in fungal nodules than in cancerous ones (each P < 0.05). Air bronchogram similarly occurred in fungal and cancerous nodules, whereas the natural ones were more common in the former (100% vs 16.7%, P = 0.000). However, the fungal nodules had a lower enhancement degree (29.0 ± 19.2 HU vs 40.3 ± 28.3 HU, P = 0.038) and frequency of hilar and/or mediastinal lymph node enlargement (2.9% vs 14.3%, P = 0.013) compared with the cancerous nodules.

Conclusion

In the younger, non-smoking and immunocompromised patients, a solitary pulmonary solid nodule with ACS, HS, satellite lesions and/or natural air bronchogram but without significant enhancement, fungal infection is a probable diagnosis.

Quantitative parameters of enhanced dual-energy computed tomography for differentiating lung cancers from benign lesions in solid pulmonary nodules

Objectives

This study aimed to investigate the ability of quantitative parameters of dual-energy computed tomography (DECT) and nodule size for differentiation between lung cancers and benign lesions in solid pulmonary nodules.

Materials and Methods

A total of 151 pathologically confirmed solid pulmonary nodules including 78 lung cancers and 73 benign lesions from 147 patients were consecutively and retrospectively enrolled who underwent dual-phase contrast-enhanced DECT. The following features were analyzed: diameter, volume, Lung CT Screening Reporting and Data System (Lung-RADS) categorization, and DECT-derived quantitative parameters including effective atomic number (Zeff), iodine concentration (IC), and normalized iodine concentration (NIC) in arterial and venous phases. Multivariable logistic regression analysis was used to build a combined model. The diagnostic performance was assessed by area under curve (AUC) of receiver operating characteristic curve, sensitivity, and specificity.

Results

The independent factors for differentiating lung cancers from benign solid pulmonary nodules included diameter, Lung-RADS categorization of diameter, volume, Zeff in arterial phase (Zeff_A), IC in arterial phase (IC_A), NIC in arterial phase (NIC_A), Zeff in venous phase (Zeff_V), IC in venous phase (IC_V), and NIC in venous phase (NIC_V) (all P < 0.05). The IC_V, NIC_V, and combined model consisting of diameter and NIC_V showed good diagnostic performance with AUCs of 0.891, 0.888, and 0.893, which were superior to the diameter, Lung-RADS categorization of diameter, volume, Zeff_A, and Zeff_V (all P < 0.001). The sensitivities of IC_V, NIC_V, and combined model were higher than those of IC_A and NIC_A (all P < 0.001). The combined model did not increase the AUCs compared with IC_V (P = 0.869) or NIC_V (P = 0.633).

Conclusion

The DECT-derived IC_V and NIC_V may be useful in differentiating lung cancers from benign lesions in solid pulmonary nodules.

Radiomics based on enhanced CT for differentiating between pulmonary tuberculosis and pulmonary adenocarcinoma presenting as solid nodules or masses

PURPOSE

To investigate the incremental value of enhanced CT-based radiomics in discriminating between pulmonary tuberculosis (PTB) and pulmonary adenocarcinoma (PAC) presenting as solid nodules or masses and to develop an optimal radiomics model.

METHODS

A total of 128 lesions (from 123 patients) from three hospitals were retrospectively analyzed and were randomly divided into training and test datasets at a ratio of 7:3. Independent predictors in subjective image features were used to develop the subjective image model (SIM). The plain CT-based and enhanced CT-based radiomics features were screened by the correlation coefficient method, univariate analysis, and the least absolute shrinkage and selection operator, then used to build the plain CT radiomics model (PRM) and enhanced CT radiomics model (ERM), respectively. Finally, the combined model (CM) combining PRM and ERM was established. In addition, the performance of three radiologists and one respiratory physician was evaluated. The areas under the receiver operating characteristic curve (AUCs) were used to assess the performance of each model.

RESULTS

The differential diagnostic capability of the ERM (training: AUC = 0.933; test: AUC = 0.881) was better than that of the PRM (training: AUC = 0.861; test: AUC = 0.756) and the SIM (training: AUC = 0.760; test: AUC = 0.611). The CM was optimal (training: AUC = 0.948; test: AUC = 0.917) and outperformed the respiratory physician and most radiologists.

CONCLUSIONS

The ERM was more helpful than the PRM for identifying PTB and PAC that present as solid nodules or masses, and the CM was the best.

The Application of Artificial Intelligence in the Diagnosis and Drug Resistance Prediction of Pulmonary Tuberculosis

With the increasing incidence and mortality of pulmonary tuberculosis, in addition to tough and controversial disease management, time-wasting and resource-limited conventional approaches to the diagnosis and differential diagnosis of tuberculosis are still awkward issues, especially in countries with high tuberculosis burden and backwardness. In the meantime, the climbing proportion of drug-resistant tuberculosis poses a significant hazard to public health. Thus, auxiliary diagnostic tools with higher efficiency and accuracy are urgently required. Artificial intelligence (AI), which is not new but has recently grown in popularity, provides researchers with opportunities and technical underpinnings to develop novel, precise, rapid, and automated implements for pulmonary tuberculosis care, including but not limited to tuberculosis detection. In this review, we aimed to introduce representative AI methods, focusing on deep learning and radiomics, followed by definite descriptions of the state-of-the-art AI models developed using medical images and genetic data to detect pulmonary tuberculosis, distinguish the infection from other pulmonary diseases, and identify drug resistance of tuberculosis, with the purpose of assisting physicians in deciding the appropriate therapeutic schedule in the early stage of the disease. We also enumerated the challenges in maximizing the impact of AI in this field such as generalization and clinical utility of the deep learning models.

Discriminating TB lung nodules from early lung cancers using deep learning

Background

In developing countries where both high rates of smoking and endemic tuberculosis (TB) are often present, identification of early lung cancer can be significantly confounded by the presence of nodules such as those due to latent TB (LTB). It is very challenging to distinguish lung cancer and LTB without invasive procedures, which have their own risks of morbidity and even mortality.

Methods

Our method uses a customized VGG16-based 15-layer 2-dimensional deep convolutional neural network (DNN) architecture with transfer learning. The DNN was trained and tested on sets of CT images set extracted from the National Lung Screening Trial and the National Institute of Allergy and Infectious Disease TB Portals. Performance of the DNN was evaluated under locked and step-wise unlocked pretrained weight conditions.

Results

The DNN with unlocked pretrained weights achieved an accuracy of 90.4% with an F score of 90.1%.

Conclusions

Our findings support the potential for a DNN to serve as a noninvasive screening tool capable of reliably detecting and distinguishing between lung cancer and LTB.

Factors associated with concurrent malignancy risk among patients with incidental solitary pulmonary nodule: A systematic review taskforce for developing rapid recommendations

OBJECTIVE

To assess the association between prespecified factors and the malignancy risk of solitary pulmonary nodules (SPNs) to support the development of rapid recommendations for daily use in the Chinese setting.

METHODS

The expert panel for the rapid recommendations voted for 12 candidate factors based on published guidelines, selected publications, and clinical experiences. We then searched Medline, Embase, and Web of Science up to October 17, 2021, for studies investigating the association between these factors and the diagnosis of malignant SPNs in patients with CT-identified SPNs through multivariable regression analysis. The risk of bias was assessed using the Agency for Healthcare Research and Quality (AHRQ) Checklist. We pooled adjusted odds ratios (aOR) between candidate factors and the diagnosis of the malignant SPNs.

RESULTS

A total of 32 cross-sectional studies were included. Nine factors were statistically associated with malignant SPNs: age (aOR 1.06, 95% confidence interval [CI]: 1.05-1.07), smoking history (2.83, 1.84-4.36), history of extrathoracic malignancy (5.66, 2.80-11.46), history of malignancy (4.64, 3.37-6.39), family history of malignancy (3.11, 1.66-5.83), nodule diameter (1.23, 1.17-1.31), spiculation (3.41, 2.64-4.41), lobulation (3.85, 2.47-6.01), and mixed ground-glass opacity (mGGO) density of the nodule (5.56, 2.47-12.52). No statistical association was found between family history of lung cancer, emphysema, nodule border, and malignant SPNs.

CONCLUSION

Nine prespecified factors were associated with the concurrent malignancy risk among patients with SPNs. Risk stratification for SPNs is warranted in clinical practice.

The Role of Chest CT Radiomics in Diagnosis of Lung Cancer or Tuberculosis: A Pilot Study

In many low-income countries, the poor availability of lung biopsy leads to delayed diagnosis of lung cancer (LC), which can appear radiologically similar to tuberculosis (TB). To assess the ability of CT Radiomics in differentiating between TB and LC, and to evaluate the potential predictive role of clinical parameters, from March 2020 to September 2021, patients with histological diagnosis of TB or LC underwent chest CT evaluation and were retrospectively enrolled. Exclusion criteria were: availability of only enhanced CT scans, previous lung surgery and significant CT motion artefacts. After manual 3D segmentation of enhanced CT, two radiologists, in consensus, extracted and compared radiomics features (T-test or Mann−Whitney), and they tested their performance, in differentiating LC from TB, via Receiver operating characteristic (ROC) curves. Forty patients (28 LC and 12 TB) were finally enrolled, and 31 were male, with a mean age of 59 ± 13 years. Significant differences were found in normal WBC count (p < 0.019) and age (p < 0.001), in favor of the LC group (89% vs. 58%) and with an older population in LC group, respectively. Significant differences were found in 16/107 radiomic features (all p < 0.05). LargeDependenceEmphasis and LargeAreaLowGrayLevelEmphasis showed the best performance in discriminating LC from TB, (AUC: 0.92, sensitivity: 85.7%, specificity: 91.7%, p < 0.0001; AUC: 0.92, sensitivity: 75%, specificity: 100%, p < 0.0001, respectively). Radiomics may be a non-invasive imaging tool in many poor nations, for differentiating LC from TB, with a pivotal role in improving oncological patients’ management; however, future prospective studies will be necessary to validate these initial findings.

Development and Validation of a Preoperative CT-Based Nomogram to Differentiate Invasive from Non-Invasive Pulmonary Adenocarcinoma in Solitary Pulmonary Nodules

Purpose

Patients and Methods

Results

Conclusion

Development and validation of a preoperative CT-based radiomic nomogram to predict pathology invasiveness in patients with a solitary pulmonary nodule: a machine learning approach, multicenter, diagnostic study

OBJECTIVES

To develop and validate a preoperative CT-based nomogram combined with radiomic and clinical-radiological signatures to distinguish preinvasive lesions from pulmonary invasive lesions.

METHODS

This was a retrospective, diagnostic study conducted from August 1, 2018, to May 1, 2020, at three centers. Patients with a solitary pulmonary nodule were enrolled in the GDPH center and were divided into two groups (7:3) randomly: development (n = 149) and internal validation (n = 54). The SYSMH center and the ZSLC Center formed an external validation cohort of 170 patients. The least absolute shrinkage and selection operator (LASSO) algorithm and logistic regression analysis were used to feature signatures and transform them into models.

RESULTS

The study comprised 373 individuals from three independent centers (female: 225/373, 60.3%; median [IQR] age, 57.0 [48.0-65.0] years). The AUCs for the combined radiomic signature selected from the nodular area and the perinodular area were 0.93, 0.91, and 0.90 in the three cohorts. The nomogram combining the clinical and combined radiomic signatures could accurately predict interstitial invasion in patients with a solitary pulmonary nodule (AUC, 0.94, 0.90, 0.92) in the three cohorts, respectively. The radiomic nomogram outperformed any clinical or radiomic signature in terms of clinical predictive abilities, according to a decision curve analysis and the Akaike information criteria.

CONCLUSIONS

This study demonstrated that a nomogram constructed by identified clinical-radiological signatures and combined radiomic signatures has the potential to precisely predict pathology invasiveness.

KEY POINTS

• The radiomic signature from the perinodular area has the potential to predict pathology invasiveness of the solitary pulmonary nodule. • The new radiomic nomogram was useful in clinical decision-making associated with personalized surgical intervention and therapeutic regimen selection in patients with early-stage non-small-cell lung cancer.

Content-based image retrieval with a Convolutional Siamese Neural Network: Distinguishing lung cancer and tuberculosis in CT images

BACKGROUND

CT findings of lung cancer and tuberculosis are sometimes similar, potentially leading to misdiagnosis. This study aims to combine deep learning and content-based image retrieval (CBIR) to distinguish lung cancer (LC) from nodular/mass atypical tuberculosis (NMTB) in CT images.

METHODS

This study proposes CBIR with a convolutional Siamese neural network (CBIR-CSNN). First, the lesion patches are cropped out to compose LC and NMTB datasets and the pairs of two arbitrary patches form a patch-pair dataset. Second, this patch-pair dataset is utilized to train a CSNN. Third, a test patch is treated as a query. The distance between this query and 20 patches in both datasets is calculated using the trained CSNN. The patches closest to the query are used to give the final prediction by majority voting. One dataset of 719 patients is used to train and test the CBIR-CSNN. Another external dataset with 30 patients is employed to verify CBIR-CSNN.

RESULTS

The CBIR-CSNN achieves excellent performance at the patch level with an mAP (Mean Average Precision) of 0.953, an accuracy of 0.947, and an area under the curve (AUC) of 0.970. At the patient level, the CBIR-CSNN correctly predicted all labels. In the external dataset, the CBIR-CSNN has an accuracy of 0.802 and AUC of 0.858 at the patch level, and 0.833 and 0.902 at the patient level.

CONCLUSIONS

This CBIR-CSNN can accurately and automatically distinguish LC from NMTB using CT images. CBIR-CSNN has excellent representation capability, compatibility with few-shot learning, and visual explainability.

Combined model of radiomics and clinical features for differentiating pneumonic-type mucinous adenocarcinoma from lobar pneumonia: An exploratory study

PURPOSE

The purpose of this study was to distinguish pneumonic-type mucinous adenocarcinoma (PTMA) from lobar pneumonia (LP) by pre-treatment CT radiological and clinical or radiological parameters.

METHODS

A total of 199 patients (patients diagnosed with LP = 138, patients diagnosed with PTMA = 61) were retrospectively evaluated and assigned to either the training cohort (n = 140) or the validation cohort (n = 59). Radiomics features were extracted from chest CT plain images. Multivariate logistic regression analysis was conducted to develop a radiomics model and a nomogram model, and their clinical utility was assessed. The performance of the constructed models was assessed with the receiver operating characteristic (ROC) curve and the area under the curve (AUC). The clinical application value of the models was comprehensively evaluated using decision curve analysis (DCA).

RESULTS

The radiomics signature, consisting of 14 selected radiomics features, showed excellent performance in distinguishing between PTMA and LP, with an AUC of 0.90 (95% CI, 0.83-0.96) in the training cohort and 0.88 (95% CI, 0.79-0.97) in the validation cohort. A nomogram model was developed based on the radiomics signature and clinical features. It had a powerful discriminative ability, with the highest AUC values of 0.94 (95% CI, 0.90-0.98) and 0.91 (95% CI, 0.84-0.99) in the training cohort and validation cohort, respectively, which were significantly superior to the clinical model alone. There were no significant differences in calibration curves from Hosmer-Lemeshow tests between training and validation cohorts (p = 0.183 and p = 0.218), which indicated the good performance of the nomogram model. DCA indicated that the nomogram model exhibited better performance than the clinical model.

CONCLUSIONS

The nomogram model based on radiomics signatures of CT images and clinical risk factors could help to differentiate PTMA from LP, which can provide appropriate therapy decision support for clinicians, especially in situations where differential diagnosis is difficult.

Clinical and CT Radiomics Nomogram for Preoperative Differentiation of Pulmonary Adenocarcinoma From Tuberculoma in Solitary Solid Nodule

Aim

To investigate clinical and computed tomography (CT) radiomics nomogram for preoperative differentiation of lung adenocarcinoma (LAC) from lung tuberculoma (LTB) in patients with pulmonary solitary solid nodule (PSSN).

Materials and Methods

A total of 313 patients were recruited in this retrospective study, including 96 pathologically confirmed LAC and 217 clinically confirmed LTB. Patients were assigned at random to training set (n = 220) and validation set (n = 93) according to 7:3 ratio. A total of 2,589 radiomics features were extracted from each three-dimensional (3D) lung nodule on thin-slice CT images and radiomics signatures were built using the least absolute shrinkage and selection operator (LASSO) logistic regression. The predictive nomogram was established based on radiomics and clinical features. Decision curve analysis was performed with training and validation sets to assess the clinical usefulness of the prediction model.

Results

A total of six clinical features were selected as independent predictors, including spiculated sign, vacuole, minimum diameter of nodule, mediastinal lymphadenectasis, sex, and age. The radiomics nomogram of lung nodules, consisting of 15 selected radiomics parameters and six clinical features showed good prediction in the training set [area under the curve (AUC), 1.00; 95% confidence interval (CI), 0.99-1.00] and validation set (AUC, 0.99; 95% CI, 0.98-1.00). The nomogram model that combined radiomics and clinical features was better than both single models (p < 0.05). Decision curve analysis showed that radiomics features were beneficial to clinical settings.

Conclusion

The radiomics nomogram, derived from unenhanced thin-slice chest CT images, showed favorable prediction efficacy for differentiating LAC from LTB in patients with PSSN.

Preoperative CT-Based Deep Learning Model for Predicting Risk Stratification in Patients With Gastrointestinal Stromal Tumors

Objective

To develop and evaluate a deep learning model (DLM) for predicting the risk stratification of gastrointestinal stromal tumors (GISTs).

Methods

Preoperative contrast-enhanced CT images of 733 patients with GISTs were retrospectively obtained from two centers between January 2011 and June 2020. The datasets were split into training (n = 241), testing (n = 104), and external validation cohorts (n = 388). A DLM for predicting the risk stratification of GISTs was developed using a convolutional neural network and evaluated in the testing and external validation cohorts. The performance of the DLM was compared with that of radiomics model by using the area under the receiver operating characteristic curves (AUROCs) and the Obuchowski index. The attention area of the DLM was visualized as a heatmap by gradient-weighted class activation mapping.

Results

In the testing cohort, the DLM had AUROCs of 0.90 (95% confidence interval [CI]: 0.84, 0.96), 0.80 (95% CI: 0.72, 0.88), and 0.89 (95% CI: 0.83, 0.95) for low-malignant, intermediate-malignant, and high-malignant GISTs, respectively. In the external validation cohort, the AUROCs of the DLM were 0.87 (95% CI: 0.83, 0.91), 0.64 (95% CI: 0.60, 0.68), and 0.85 (95% CI: 0.81, 0.89) for low-malignant, intermediate-malignant, and high-malignant GISTs, respectively. The DLM (Obuchowski index: training, 0.84; external validation, 0.79) outperformed the radiomics model (Obuchowski index: training, 0.77; external validation, 0.77) for predicting risk stratification of GISTs. The relevant subregions were successfully highlighted with attention heatmap on the CT images for further clinical review.

Conclusion

The DLM showed good performance for predicting the risk stratification of GISTs using CT images and achieved better performance than that of radiomics model.

Frequent EGFR Mutations and Better Prognosis in Positron Emission Tomography-Negative, Solid-Type Lung Cancer

BACKGROUND

The differential diagnosis of a solitary solid-type lung nodule is diverse. 18F-fluorodeoxyglucose positron emission tomography (PET) has a high sensitivity in the diagnosis of solid-type lung cancers; however, PET-negative, solid-type lung cancers are rarely observed. In this study, we analyzed the clinical/genetic features and prognosis of PET-negative, solid-type lung cancers.

PATIENTS AND METHODS

Between January 2007 and February 2020, 709 patients with solid-type lung cancers (tumor size ≥2.0 cm) underwent pulmonary resection. Clinical, genetic, and prognostic features were evaluated in 27 patients (3.8%) with tumors showing negative PET results defined as SUVmax <2.0.

RESULTS

All 27 patients had lung adenocarcinoma; 23 had invasive adenocarcinomas and 4 had invasive mucinous adenocarcinomas. The PET-negative group showed high frequencies of females and never-smokers. Recurrence-free survival was significantly better in the PET-negative group compared with PET-positive counterparts extracted using propensity score matching from patients who underwent pulmonary resection during the same period (P = .0052). Furthermore, 83% of PET-negative, solid-type invasive lung adenocarcinoma patients harbored EGFR mutation, which was significantly higher than that of PET-positive, solid-type invasive lung adenocarcinoma patients (38%, n = 225) who received EGFR mutation testing in our cohort (P < .0001). PET-negative, solid-type lung adenocarcinoma patients with EGFR mutations had significantly better recurrence-free survival compared with PET-positive, solid-type lung adenocarcinoma patients with EGFR mutations extracted using propensity score matching (P = .0030).

CONCLUSION

PET-negative, solid-type lung cancers are characterized with a high incidence of EGFR mutation and a better prognosis compared with PET-positive, solid-type lung cancer.

Assessing the predictive accuracy of lung cancer, metastases, and benign lesions using an artificial intelligence-driven computer aided diagnosis system

Background

Artificial intelligence (AI) products have been widely used for the clinical detection of primary lung tumors. However, their performance and accuracy in risk prediction for metastases or benign lesions remain underexplored. This study evaluated the accuracy of an AI-driven commercial computer-aided detection (CAD) product (InferRead CT Lung Research, ICLR) in malignancy risk prediction using a real-world database.

Methods

This retrospective study assessed 486 consecutive resected lung lesions, including 320 adenocarcinomas, 40 other malignancies, 55 metastases, and 71 benign lesions, from September 2015 to November 2018. The malignancy risk probability of each lesion was obtained using the ICLR software based on a 3D convolutional neural network (CNN) with DenseNet architecture as a backbone (without clinical data). Two resident doctors independently graded each lesion using patient clinical history. One doctor (R1) has 3 years of chest radiology experience, and the other doctor (R2) has 3 years of general radiology experience. Cochran's Q test was used to assess the performances of the AI compared to the radiologists.

Results

The accuracy of malignancy-risk prediction using the ICLR for adenocarcinomas, other malignancies, metastases, and benign lesions was 93.4% (299/320), 95.0% (38/40), 50.9% (28/55), and 40.8% (29/71), respectively. The accuracy was significantly higher in adenocarcinomas and other malignancies compared to metastases and benign lesions (all P<0.05). The overall accuracy of risk prediction for R1 was 93.6% (455/486) and 87.4% for R2 (425/486), both of which were higher than the 81.1% accuracy obtained with the ICLR (394/486) (R1 vs. ICLR: P<0.001; R2 vs. ICLR: P=0.001), especially in assessing the risk of metastases (P<0.05). R1 performed better than R2 at risk prediction (P=0.001).

Conclusions

The accuracy of the ICLR for risk prediction is very high for primary lung cancers but poor for metastases and benign lesions.

Discriminating Small-Sized (2 cm or Less), Noncalcified, Solitary Pulmonary Tuberculoma and Solid Lung Adenocarcinoma in Tuberculosis-Endemic Areas

Background. Pulmonary tuberculoma can mimic lung malignancy and thereby pose a diagnostic dilemma to clinicians. The purpose of this study was to establish an accurate, convenient, and clinically practical model for distinguishing small-sized, noncalcified, solitary pulmonary tuberculoma from solid lung adenocarcinoma. Methods. Thirty-one patients with noncalcified, solitary tuberculoma and 30 patients with solid adenocarcinoma were enrolled. Clinical characteristics and CT morphological features of lesions were compared between the two groups. Multivariate logistic regression analyses were applied to identify independent predictors of pulmonary tuberculoma and lung adenocarcinoma. Receiver operating characteristic (ROC) analysis was performed to investigate the discriminating efficacy. Results. The mean age of patients with tuberculoma and adenocarcinoma was 46.8 ± 12.3 years (range, 28–64) and 61.1 ± 9.9 years (range, 41–77), respectively. No significant differences were observed concerning smoking history and smoking index, underlying disease, or tumor markers between the two groups. Univariate and multivariate analyses showed age and lobulation combined with pleural indentation demonstrated excellent discrimination. The sensitivity, specificity, accuracy, and the area under the ROC curve were 87.1%, 93.3%, 90.2%, and 0.956 (95% confidence interval (CI), 0.901–1.000), respectively. Conclusion. The combination of clinical characteristics and CT morphological features can be used to distinguish noncalcified, solitary tuberculoma from solid adenocarcinoma with high diagnostic performance and has a clinical application value.

Lung Cancer and Granuloma Identification Using a Deep Learning Model to Extract 3-Dimensional Radiomics Features in CT Imaging

BACKGROUND

We aimed to evaluate a deep learning (DL) model combining perinodular and intranodular radiomics features and clinical features for preoperative differentiation of solitary granuloma nodules (GNs) from solid lung cancer nodules in patients with spiculation, lobulation, or pleural indentation on CT.

PATIENTS AND METHODS

We retrospectively recruited 915 patients with solitary solid pulmonary nodules and suspicious signs of malignancy. Data including clinical characteristics and subjective CT findings were obtained. A 3-dimensional U-Net-based DL model was used for tumor segmentation and extraction of 3-dimensional radiomics features. We used the Maximum Relevance and Minimum Redundancy (mRMR) algorithm and the eXtreme Gradient Boosting (XGBoost) algorithm to select the intranodular, perinodular, and gross nodular radiomics features. We propose a medical image DL (IDL) model, a clinical image DL (CIDL) model, a radiomics DL (RDL) model, and a clinical image radiomics DL (CIRDL) model to preoperatively differentiate GNs from solid lung cancer. Five-fold cross-validation was used to select and evaluate the models. The prediction performance of the models was evaluated using receiver operating characteristic and calibration curves.

RESULTS

The CIRDL model achieved the best performance in differentiating between GNs and solid lung cancer (area under the curve [AUC] = 0.9069), which was significantly higher compared with the IDL (AUC = 0.8322), CIDL (AUC = 0.8652), intra-RDL (AUC = 0.8583), peri-RDL (AUC = 0.8259), and gross-RDL (AUC = 0.8705) models.

CONCLUSION

The proposed CIRDL model is a noninvasive diagnostic tool to differentiate between granuloma nodules and solid lung cancer nodules and reduce the need for invasive diagnostic and surgical procedures.