RadioGraphics Update: Lung-RADS 2022

Evaluation of the safety of short-term follow-up CT for the management of consolidation in lung cancer screening

OBJECTIVES

Focal consolidation on CT may be inflammatory or malignant, and PET-CT imaging is rarely discriminatory. Furthermore, consolidation may demonstrate spontaneous resolution obviating the need for PET-CT imaging. This retrospective study sought to assess the safety and cost-effectiveness of short-interval 6-week follow-up CT for consolidation in a lung cancer screening programme.

METHODS

Between January 2019 and January 2024, participants in a regional lung cancer screening programme with focal indeterminate consolidation underwent a 6-week repeat CT rather than immediate PET-CT and invasive investigation. The proportion of participants with non-resolving consolidation, the risk of malignancy in consolidation at a 6-week follow-up, and the risk of upstaging over a 6-week delay were determined. Cost savings were estimated from National Health Service reference costs.

RESULTS

In 10,247 CT studies, focal indeterminate consolidation was detected in 113 participants (1.1%) (mean age 68 years, range 55-76, 65 males). Consolidation spontaneously resolved at 6 weeks in 63/110 (57%) who attended follow-up; 14/110 (12.7%) participants had malignancy; no patients upstaged during follow-up. An estimated cost saving of £47,600/10,000 screening CTs performed might be obtained through a conservative approach of short-term interval CT, rather than immediate PET-CT and further investigation.

CONCLUSION

Early repeat CT avoids PET-CT in more than half of patients with consolidation and can be utilised to reduce over-investigation of screen-detected consolidation, which may demonstrate spontaneous resolution.

KEY POINTS

Question Is short-term interval follow-up CT in lung cancer screening a safe and cost-effective approach to managing indeterminate (inflammatory or malignant) consolidation? Findings Short-interval CT imaging demonstrates spontaneous resolution of consolidation in over 50% participants in this study, whilst persistent consolidation has a high likelihood of malignancy. Clinical relevance Short-interval CT did not result in upstaging of malignancy and therefore can be considered a safe strategy to prevent the over-investigation of screen-detected consolidation, supporting recent European and American screening recommendations.

The Azygos Esophageal Recess Is Not to Be Missed in Screening Lung Cancer With LDCT

PURPOSE

Lesion overlooking and late diagnostic workup can compromise the efficacy of low-dose CT (LDCT) screening of lung cancer (LC), implying more advanced and less curable disease stages. We hypothesized that the azygos esophageal recess (AER) of the right lower lobe (RLL) might be an area prone to lesion overlooking in LC screening.

MATERIALS AND METHODS

Two radiologists reviewed the LDCT examinations of all the screen-detected incident LCs observed in the active arm of 2 randomized clinical trials: ITALUNG and national lung screening trial. Those in the AER were compared with those in the remainder of the RLL for possible differences in diagnostic lag according to the Lung-RADS 1.1 recommendations, size, stage, and mortality.

RESULTS

Six (11.7%) of 51 screen-detected incident LCs of the RLL were located in the AER. The diagnostic lag time was significantly longer ( P =0.046) in the AER LC (mean 14±9 mo) than in the LC in the remaining RLL (mean 7.3±1 mo). Size and stage at diagnosis were not significantly different. All 6 subjects with LC in the AER and 16 (35.5%) of 45 subjects with LC in the remaining RLL ( P =0.004) died of LC after a median follow-up of 12 years.

CONCLUSION

Our retrospective study indicates that AER might represent a lung region of the RLL prone to have early LC overlooked due to detection or interpretation errors with possible detrimental consequences for the subject undergoing LC screening.

External validation of the performance of commercially available deep-learning-based lung nodule detection on low-dose CT images for lung cancer screening in Japan

PURPOSE

Artificial intelligence (AI) algorithms for lung nodule detection have been developed to assist radiologists. However, external validation of its performance on low-dose CT (LDCT) images is insufficient. We examined the performance of the commercially available deep-learning-based lung nodule detection (DL-LND) using LDCT images at Japanese lung cancer screening (LCS).

MATERIALS AND METHODS

Included were 43 patients with suspected lung cancer on LDCT images and pathologically confirmed lung cancer. The reference standard for nodules whose diameter exceeded 4 mm was set by a radiologist who referred to the reports of two other radiologists reading the LDCT images. After we applied commercially available DL-LND to the LDCT images, the radiologist reviewed all nodules detected by DL-LND. When he failed to identify an existing nodule, it was also included in the reference standard. To validate the performance of DL-LND, the sensitivity for lung nodules and lung cancer, the positive-predictive value (PPV) for lung nodules, and the mean number of false-positive (FP) nodules per CT scan were recorded.

RESULTS

The radiologist detected 97 nodules including 43 lung cancers and missed 3 solid nodules detected by DL-LND. A total of 100 nodules was included in the reference standard. DL-LND detected 396 nodules including 40 lung cancers. The sensitivity for the 100 nodules was 96.0%; the PPV was 24.2% (96/396). The mean number of FP nodules per CT scan was 7.0; sensitivity for lung cancer was 93.0% (40/43). DL-LND missed three lung cancers; 2 of these were atypical pulmonary cysts.

CONCLUSION

We externally verified that the sensitivity for lung nodules and lung cancer by DL-LND was very high. However, its low PPV and the increased FP nodules remains a serious drawback of DL-LND.

Incidentally Detected Adrenal Nodules on Lung Cancer Screening CT

OBJECTIVE

To assess adherence to ACR recommendations for managing incidental adrenal lesions detected on lung cancer screening (LCS) CT examinations.

METHODS

We performed a retrospective analysis of all LCS CT examinations within our health care system from January 2015 to August 2023. We included CTs that were reported with Lung-RADS "S" modifier for a focal adrenal lesion. We recorded whether follow-up imaging and biochemical testing were recommended and whether they were performed. Follow-up recommendations in reports were assessed for adherence to ACR recommendations.

RESULTS

During the study period, 191 patients had a focal adrenal nodule reported. Per ACR recommendations, 36 of 191 (19%) warranted follow-up, but only 23 of 36 (64%) of these received follow-up recommendations. Of those 191, 155 (81%) did not require follow-up per ACR, and 25 of those 155 (16%) received follow-up recommendations. Of those who were advised follow-up, 34 of 48 (71%) received dedicated follow-up, 9 of 48 (19%) received follow-up imaging for another reason, and 5 of 48 (10%) did not receive any follow-up. Among those in whom follow-up was not recommended, 21 of 143 (15%) received dedicated follow-up, 101 of 143 (71%) received follow-up imaging for another reason, and 21 of 143 (15%) did not receive any follow-up. No malignant lesions were diagnosed. Per ACR recommendations, 183 of 191 (96%) of patients should have received biochemical testing; however, it was recommended in only 4 patients (2%).

DISCUSSION

There was suboptimal adherence to ACR recommendations for managing incidental adrenal lesions on LCS CTs, with both unnecessary and missing follow-up recommendations. Recommendations for biochemical testing were nearly nonexistent, despite being part of the ACR algorithm.

Minimally invasive biomarkers for triaging lung nodules-challenges and future perspectives

CT chest scans are commonly performed worldwide, either in routine clinical practice for a wide range of indications or as part of lung cancer screening programs. Many of these scans detect lung nodules, which are small, rounded opacities measuring 8-30 mm. While the concern about nodules is that they may represent early lung cancer, in screening programs, only 1% of such nodules turn out to be cancer. This leads to a series of complex decisions and, at times, unnecessary biopsies for nodules that are ultimately determined to be benign. Additionally, patients may be anxious about the status of detected lung nodules. The high rate of false positive lung nodule detections has driven advancements in biomarker-based research aimed at triaging lung nodules (benign versus malignant) to identify truly malignant nodules better. Biomarkers found in biofluids and breath hold promise owing to their minimally invasive sampling methods, ease of use, and cost-effectiveness. Although several biomarkers have demonstrated clinical utility, their sensitivity and specificity are still relatively low. Combining multiple biomarkers could enhance the characterisation of small pulmonary nodules by addressing the limitations of individual biomarkers. This approach may help reduce unnecessary invasive procedures and accelerate diagnosis in the future. This review offers a thorough overview of emerging minimally invasive biomarkers for triaging lung nodules, emphasising key challenges and proposing potential solutions for biomarker-based nodule differentiation. It focuses on diagnosis rather than screening, analysing research published primarily in the past five years with some exceptions. The incorporation of biomarkers into clinical practice will facilitate the early detection of malignant nodules, leading to timely interventions and improved outcomes. Further efforts are needed to increase the cost-effectiveness and practicality of many of these applications in clinical settings. However, the range of technologies is advancing rapidly, and they may soon be implemented in clinics in the near future.

Malignancy risk stratification for pulmonary nodules: comparing a deep learning approach to multiparametric statistical models in different disease groups

OBJECTIVES

Incidentally detected pulmonary nodules present a challenge in clinical routine with demand for reliable support systems for risk classification. We aimed to evaluate the performance of the lung-cancer-prediction-convolutional-neural-network (LCP-CNN), a deep learning-based approach, in comparison to multiparametric statistical methods (Brock model and Lung-RADS®) for risk classification of nodules in cohorts with different risk profiles and underlying pulmonary diseases.

MATERIALS AND METHODS

Retrospective analysis was conducted on non-contrast and contrast-enhanced CT scans containing pulmonary nodules measuring 5-30 mm. Ground truth was defined by histology or follow-up stability. The final analysis was performed on 297 patients with 422 eligible nodules, of which 105 nodules were malignant. Classification performance of the LCP-CNN, Brock model, and Lung-RADS® was evaluated in terms of diagnostic accuracy measurements including ROC-analysis for different subcohorts (total, screening, emphysema, and interstitial lung disease).

RESULTS

LCP-CNN demonstrated superior performance compared to the Brock model in total and screening cohorts (AUC 0.92 (95% CI: 0.89-0.94) and 0.93 (95% CI: 0.89-0.96)). Superior sensitivity of LCP-CNN was demonstrated compared to the Brock model and Lung-RADS® in total, screening, and emphysema cohorts for a risk threshold of 5%. Superior sensitivity of LCP-CNN was also shown across all disease groups compared to the Brock model at a threshold of 65%, compared to Lung-RADS® sensitivity was better or equal. No significant differences in the performance of LCP-CNN were found between subcohorts.

CONCLUSION

This study offers further evidence of the potential to integrate deep learning-based decision support systems into pulmonary nodule classification workflows, irrespective of the individual patient risk profile and underlying pulmonary disease.

KEY POINTS

Question Is a deep-learning approach (LCP-CNN) superior to multiparametric models (Brock model, Lung-RADS®) in classifying pulmonary nodule risk across varied patient profiles? Findings LCP-CNN shows superior performance in risk classification of pulmonary nodules compared to multiparametric models with no significant impact on risk profiles and structural pulmonary diseases. Clinical relevance LCP-CNN offers efficiency and accuracy, addressing limitations of traditional models, such as variations in manual measurements or lack of patient data, while producing robust results. Such approaches may therefore impact clinical work by complementing or even replacing current approaches.

Improved Consistency of Lung Nodule Categorization in CT Scans with Heterogeneous Slice Thickness by Deep Learning-Based 3D Super-Resolution

Background/Objectives: Accurate volumetric assessment of lung nodules is an essential element of low-dose lung cancer screening programs. Current guidance recommends applying specific thresholds to measured nodule volume to make the following clinical decisions. In reality, however, CT scans often have heterogeneous slice thickness which is known to adversely impact the accuracy of nodule volume assessment. Methods: In this study, a deep learning (DL)-based 3D super-resolution method is proposed for generating thin-slice CT images from heterogeneous thick-slice CT images in lung cancer screening. We evaluated the performance in a qualitative way by radiologist's perceptual assessment as well as in a quantitative way by accuracy of nodule volume measurements and agreement of volume-based Lung-RADS nodule category. Results: A 5-point Likert scale tabulated by two radiologists showed that the quality of DL-generated thin-slice images from thick-slice CT images were on a par with the image quality of ground truth thin-slice CT images. Furthermore, thick- and thin-slice CT images had a nodule volume difference of 52.2 percent on average which was reduced to a 15.7 percent difference with DL-generated thin-slice CT. In addition, the proposed method increased the agreement of lung nodule categorization using Lung-RADS by 74 percent. Conclusions: The proposed DL approach for slice thickness normalization has a potential for improving the accuracy of lung nodule volumetry and facilitating more reliable early lung nodule detection.

Evaluating Large Language Models for Automated Reporting and Data Systems Categorization: Cross-Sectional Study

BACKGROUND

Large language models show promise for improving radiology workflows, but their performance on structured radiological tasks such as Reporting and Data Systems (RADS) categorization remains unexplored.

OBJECTIVE

This study aims to evaluate 3 large language model chatbots-Claude-2, GPT-3.5, and GPT-4-on assigning RADS categories to radiology reports and assess the impact of different prompting strategies.

METHODS

This cross-sectional study compared 3 chatbots using 30 radiology reports (10 per RADS criteria), using a 3-level prompting strategy: zero-shot, few-shot, and guideline PDF-informed prompts. The cases were grounded in Liver Imaging Reporting & Data System (LI-RADS) version 2018, Lung CT (computed tomography) Screening Reporting & Data System (Lung-RADS) version 2022, and Ovarian-Adnexal Reporting & Data System (O-RADS) magnetic resonance imaging, meticulously prepared by board-certified radiologists. Each report underwent 6 assessments. Two blinded reviewers assessed the chatbots' response at patient-level RADS categorization and overall ratings. The agreement across repetitions was assessed using Fleiss κ.

RESULTS

Claude-2 achieved the highest accuracy in overall ratings with few-shot prompts and guideline PDFs (prompt-2), attaining 57% (17/30) average accuracy over 6 runs and 50% (15/30) accuracy with k-pass voting. Without prompt engineering, all chatbots performed poorly. The introduction of a structured exemplar prompt (prompt-1) increased the accuracy of overall ratings for all chatbots. Providing prompt-2 further improved Claude-2's performance, an enhancement not replicated by GPT-4. The interrun agreement was substantial for Claude-2 (k=0.66 for overall rating and k=0.69 for RADS categorization), fair for GPT-4 (k=0.39 for both), and fair for GPT-3.5 (k=0.21 for overall rating and k=0.39 for RADS categorization). All chatbots showed significantly higher accuracy with LI-RADS version 2018 than with Lung-RADS version 2022 and O-RADS (P<.05); with prompt-2, Claude-2 achieved the highest overall rating accuracy of 75% (45/60) in LI-RADS version 2018.

CONCLUSIONS

When equipped with structured prompts and guideline PDFs, Claude-2 demonstrated potential in assigning RADS categories to radiology cases according to established criteria such as LI-RADS version 2018. However, the current generation of chatbots lags in accurately categorizing cases based on more recent RADS criteria.

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.

Coronary Artery Calcification on Low-Dose Lung Cancer Screening CT in South Korea: Visual and Artificial Intelligence-Based Assessment and Association With Cardiovascular Events

BACKGROUND. Coronary artery calcification (CAC) on lung cancer screening low-dose chest CT (LDCT) is a cardiovascular risk marker. South Korea was the first Asian country to initiate a national LDCT lung cancer screening program, although CAC-related outcomes are poorly explored. OBJECTIVE. The purpose of this article is to evaluate CAC prevalence and severity using visual analysis and artificial intelligence (AI) methods and to characterize CAC's association with major adverse cardiovascular events (MACEs) in patients undergoing LDCT in Korea's national lung cancer screening program. METHODS. This retrospective study included 1002 patients (mean age, 62.4 ± 5.4 [SD] years; 994 men, eight women) who underwent LDCT at two Korean medical centers between April 2017 and May 2023 as part of Korea's national lung cancer screening program. Two radiologists independently assessed CAC presence and severity using visual analysis, consulting a third radiologist to resolve differences. Two AI software applications were also used to assess CAC presence and severity. MACE occurrences were identified by EMR review. RESULTS. Interreader agreement for CAC presence and severity, expressed as kappa, was 0.793 and 0.671, respectively. CAC prevalence was 53.4% by consensus visual assessment, 60.1% by AI software I, and 56.6% by AI software II. CAC severity was mild, moderate, and severe by consensus visual analysis in 28.0%, 10.3%, and 15.1%; by AI software I in 39.9%, 14.0%, and 6.2%; and by AI software II in 34.9%, 14.3%, and 7.3%. MACEs occurred in 36 of 625 (5.6%) patients with follow-up after LDCT (median, 1108 days). MACE incidence in patients with no, mild, moderate, and severe CAC for consensus visual analysis was 1.1%, 5.0%, 2.9%, and 8.6%, respectively (p < .001); for AI software I, it was 1.3%, 3.0%, 7.9%, and 11.3% (p < .001); and for AI software II, it was 1.2%, 3.4%, 7.7%, and 9.6% (p < .001). CONCLUSION. For Korea's national lung cancer screening program, MACE occurrence increased significantly with increasing CAC severity, whether assessed by visual analysis or AI software. The study is limited by the large sex imbalance for Korea's national lung cancer screening program. CLINICAL IMPACT. The findings provide reference data for health care practitioners engaged in developing and overseeing national lung cancer screening programs, highlighting the importance of routine CAC evaluation.