Automated evaluation of diaphragm configuration based on chest CT in COPD patients.

BACKGROUND: Severe chronic obstructive pulmonary disease (COPD) often results in hyperinflation and flattening of the diaphragm. An automated computed tomography (CT)-based tool for quantifying diaphragm configuration, a biomarker for COPD, was developed in-house and tested in a large cohort of COPD patients. METHODS: We used the LungQ platform to extract the lung-diaphragm intersection, as direct diaphragm segmentation is challenging. The tool computed the diaphragm index (surface area/projected surface area) as a measure of diaphragm configuration on inspiratory scans in a COPDGene subcohort. Visual inspection of 250 randomly selected segmentations served as a quality check. Associations between the diaphragm index, Global Initiative for Chronic Obstructive Lung Disease (GOLD) stages, forced expiratory volume in 1 s (FEV1) % predicted, and CT-derived emphysema scores were explored using analysis of variance and Pearson correlation. RESULTS: The tool yielded incomplete segmentation in 9.2% (2.4% major defect, 6.8% minor defect) of 250 randomly selected cases. In 8431 COPDGene subjects (4240 healthy; 4191 COPD), the diaphragm index was increasingly lower with higher GOLD stages (never-smoked 1.83 ± 0.16; GOLD-0 1.79 ± 0.18; GOLD-1 1.71 ± 0.15; GOLD-2: 1.67 ± 0.16; GOLD-3 1.58 ± 0.14; GOLD-4 1.54 ± 0.11) (p < 0.001). Associations were found between the diaphragm index and both FEV1% predicted (r = 0.44, p < 0.001) and emphysema score (r = -0.36, p < 0.001). CONCLUSION: We developed an automated tool to quantify the diaphragm configuration in chest CT. The diaphragm index was associated with COPD severity, FEV1%predicted, and emphysema score. RELEVANCE STATEMENT: Due to the hypothesized relationship between diaphragm dysfunction and diaphragm configuration in COPD patients, automatic quantification of diaphragm configuration may prove useful in evaluating treatment efficacy in terms of lung volume reduction. KEY POINTS: Severe COPD changes diaphragm configuration to a flattened state, impeding function. An automated tool quantified diaphragm configuration on chest-CT providing a diaphragm index. The diaphragm index was correlated to COPD severity and may aid treatment assessment.

CT Myocardial Perfusion and CT-FFR versus Invasive FFR for Hemodynamic Relevance of Coronary Artery Disease.

Background CT-derived fractional flow reserve (CT-FFR) and dynamic CT myocardial perfusion imaging enhance the specificity of coronary CT angiography (CCTA) for ruling out coronary artery disease (CAD). However, evidence on comparative diagnostic value remains scarce. Purpose To compare the diagnostic accuracy of CCTA plus CT-FFR, CCTA plus CT perfusion, and sequential CCTA plus CT-FFR and CT perfusion for detecting hemodynamically relevant CAD with that of invasive angiography. Materials and Methods This secondary analysis of a prospective study included patients with chest pain referred for invasive coronary angiography at nine centers from July 2016 to September 2019. CCTA and CT perfusion were performed with third-generation dual-source CT scanners. CT-FFR was assessed on-site. Independent core laboratories analyzed CCTA alone, CCTA plus CT perfusion, CCTA plus CT-FFR, and a sequential approach involving CCTA plus CT-FFR and CT perfusion for the presence of hemodynamically relevant stenosis. Invasive coronary angiography with invasive fractional flow reserve was the reference standard. Diagnostic accuracy metrics and the area under the receiver operating characteristic curve (AUC) were compared with the Sign test and DeLong test. Results Of the 105 participants (mean age, 64 years ± 8 [SD]; 68 male), 49 (47%) had hemodynamically relevant stenoses at invasive coronary angiography. CCTA plus CT-FFR and CCTA plus CT perfusion showed no evidence of a difference for participant-based sensitivities (90% vs 90%, P > .99), specificities (77% vs 79%, P > .99) and vessel-based AUCs (0.84 [95% CI: 0.77, 0.91] vs 0.83 [95% CI: 0.75, 0.91], P = .90). Both had higher participant-based specificity than CCTA alone (54%, both P < .001) without evidence of a difference in sensitivity between CCTA (94%) and CCTA plus CT perfusion (P = .50) or CCTA plus CT-FFR (P = .63). The sequential approach combining CCTA plus CT-FFR with CT perfusion achieved higher participant-based specificity than CCTA plus CT-FFR (88% vs 77%, P = .03) without evidence of a difference in participant-based sensitivity (88% vs 90%, P > .99) and vessel-based AUC (0.85 [95% CI: 0.77, 0.93], P = .78). Compared with CCTA plus CT perfusion, the sequential approach showed no evidence of a difference in participant-based sensitivity (P > .99), specificity (P = .06), or vessel-based AUC (P = .54). Conclusion There was no evidence of a difference in diagnostic accuracy between CCTA plus CT-FFR and CCTA plus CT perfusion for detecting hemodynamically relevant CAD. A sequential approach combining CCTA plus CT-FFR with CT perfusion led to improved participant-based specificity with no evidence of a difference in sensitivity compared with CCTA plus CT-FFR. ClinicalTrials.gov registration no.: NCT02810795 © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Sinitsyn in this issue.

Distribution of Solid Lung Nodules Presence and Size by Age and Sex in a Northern European Nonsmoking Population.

Background Most of the data regarding prevalence and size distribution of solid lung nodules originates from lung cancer screening studies that target high-risk populations or from Asian general cohorts. In recent years, the identification of lung nodules in non-high-risk populations, scanned for clinical indications, has increased. However, little is known about the presence of solid lung nodules in the Northern European nonsmoking population. Purpose To study the prevalence and size distribution of solid lung nodules by age and sex in a nonsmoking population. Materials and Methods Participants included nonsmokers (never or former smokers) from the population-based Imaging in Lifelines study conducted in the Northern Netherlands. Participants (age ≥ 45 years) with completed lung function tests underwent chest low-dose CT scans. Seven trained readers registered the presence and size of solid lung nodules measuring 30 mm3 or greater using semiautomated software. The prevalence and size of lung nodules (≥30 mm3), clinically relevant lung nodules (≥100 mm3), and actionable nodules (≥300 mm3) are presented by 5-year categories and by sex. Results A total of 10 431 participants (median age, 60.4 years [IQR, 53.8-70.8 years]; 56.6% [n = 5908] female participants; 46.1% [n = 4812] never smokers and 53.9% [n = 5619] former smokers) were included. Of these, 42.0% (n = 4377) had at least one lung nodule (male participants, 47.5% [2149 of 4523]; female participants, 37.7% [2228 of 5908]). The prevalence of lung nodules increased from age 45-49.9 years (male participants, 39.4% [219 of 556]; female participants, 27.7% [236 of 851]) to age 80 years or older (male participants, 60.7% [246 of 405]; female participants, 50.9% [163 of 320]). Clinically relevant lung nodules were present in 11.1% (1155 of 10 431) of participants, with prevalence increasing with age (male participants, 8.5%-24.4%; female participants, 3.7%-15.6%), whereas actionable nodules were present in 1.1%-6.4% of male participants and 0.6%-4.9% of female participants. Conclusion Lung nodules were present in a substantial proportion of all age groups in the Northern European nonsmoking population, with slightly higher prevalence for male participants than female participants. © RSNA, 2024 Supplemental material is available for this article.

Low-Dose CT-derived Bronchial Parameters in Individuals with Healthy Lungs.

Background CT-derived bronchial parameters have been linked to chronic obstructive pulmonary disease and asthma severity, but little is known about these parameters in healthy individuals. Purpose To investigate the distribution of bronchial parameters at low-dose CT in individuals with healthy lungs from a Dutch general population. Materials and Methods In this prospective study, low-dose chest CT performed between May 2017 and October 2022 were obtained from participants who had completed the second-round assessment of the prospective, longitudinal Imaging in Lifelines study. Participants were aged at least 45 years, and those with abnormal spirometry, self-reported respiratory disease, or signs of lung disease at CT were excluded. Airway lumens and walls were segmented automatically. The square root of the bronchial wall area of a hypothetical airway with an internal perimeter of 10 mm (Pi10), luminal area (LA), wall thickness (WT), and wall area percentage were calculated. Associations between sex, age, height, weight, smoking status, and bronchial parameters were assessed using univariable and multivariable analyses. Results The study sample was composed of 8869 participants with healthy lungs (mean age, 60.9 years ± 10.4 [SD]; 4841 [54.6%] female participants), including 3672 (41.4%) never-smokers and 1197 (13.5%) individuals who currently smoke. Bronchial parameters for male participants were higher than those for female participants (Pi10, slope [ß] range = 3.49-3.66 mm; LA, ß range = 25.40-29.76 mm2; WT, ß range = 0.98-1.03 mm; all P < .001). Increasing age correlated with higher Pi10, LA, and WT (r2 range = 0.06-0.09, 0.02-0.01, and 0.02-0.07, respectively; all P < .001). Never-smoking individuals had the lowest Pi10 followed by formerly smoking and currently smoking individuals (3.62 mm ± 0.13, 3.68 mm ± 0.14, and 3.70 mm ± 0.14, respectively; all P < .001). In multivariable regression models, age, sex, height, weight, and smoking history explained up to 46% of the variation in bronchial parameters. Conclusion In healthy individuals, bronchial parameters differed by sex, height, weight, and smoking history; male sex and increasing age were associated with wider lumens and thicker walls. © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Emrich and Varga-Szemes in this issue.

Feasibility of virtual non-iodine coronary calcium scoring on dual source photon-counting coronary CT angiography: a dynamic phantom study.

BACKGROUND: The aim of our current systematic dynamic phantom study was first, to optimize reconstruction parameters of coronary CTA (CCTA) acquired on photon counting CT (PCCT) for coronary artery calcium (CAC) scoring, and second, to assess the feasibility of calculating CAC scores from CCTA, in comparison to reference calcium scoring CT (CSCT) scans. METHODS: In this phantom study, an artificial coronary artery was translated at velocities corresponding to 0, < 60, and 60-75 beats per minute (bpm) within an anthropomorphic phantom. The density of calcifications was 100 (very low), 200 (low), 400 (medium), and 800 (high) mgHA/cm3, respectively. CCTA was reconstructed with the following parameters: virtual non-iodine (VNI), with and without iterative reconstruction (QIR level 2, QIR off, respectively); kernels Qr36 and Qr44f; slice thickness/increment 3.0/1.5 mm and 0.4/0.2 mm. The agreement in risk group classification between CACCCTA and CACCSCT scoring was measured using Cohen weighted linear κ with 95% CI. RESULTS: For CCTA reconstructed with 0.4 mm slice thickness, calcium detectability was perfect (100%). At < 60 bpm, CACCCTA of low, and medium density calcification was underestimated by 53%, and 15%, respectively. However, CACCCTA was not significantly different from CACCSCT of very low, and high-density calcifications. The best risk agreement was achieved when CCTA was reconstructed with QIR off, Qr44f, and 0.4 mm slice thickness (κ = 0.762, 95% CI 0.671-0.853). CONCLUSION: In this dynamic phantom study, the detection of calcifications with different densities was excellent with CCTA on PCCT using thin-slice VNI reconstruction. Agatston scores were underestimated compared to CSCT but agreement in risk classification was substantial. CLINICAL RELEVANCE STATEMENT: Photon counting CT may enable the implementation of coronary artery calcium scoring from coronary CTA in daily clinical practice. KEY POINTS: Photon-counting CTA allows for excellent detectability of low-density calcifications at all heart rates. Coronary artery calcium scoring from coronary CTA acquired on photon counting CT is feasible, although improvement is needed. Adoption of the standard acquisition and reconstruction protocol for calcium scoring is needed for improved quantification of coronary artery calcium to fully employ the potential of photon counting CT.

Phase angle and donor type are determinants of coronary artery calcification in stable kidney transplant recipients at twelve months after transplantation.

BACKGROUND AND AIM: Coronary artery calcification (CAC) partially explains the excess cardiovascular morbidity and mortality after kidney transplantation. This study aimed to investigate determinants of CAC in stable kidney transplant recipients at 12 months post-transplantation. METHODS AND RESULTS: CAC-score was quantified by the Agatston method using non-contrast enhanced computed tomography, and age- and sex-standardized CAC-percentiles were calculated. Univariable and multivariable multinomial logistic regression was performed to study potential determinants of CAC. The independent determinants were included in multivariable multinomial logistic regression adjusting for potential confounders. 203 KTRs (age 54.0 ± 14.7 years, 61.1% male) were included. Participants were categorized into four groups according to CAC percentiles (p = 0 [CAC-score = 0], n = 68; p ≥ 1%-p ≤ 50% [CAC score = 29.0 (4.0-166.0)], n = 31; p > 50 ≤ 75% [CAC score = 101.0 (23.8-348.3)], n = 26; and p>75% [CAC score = 581.0 (148.0-1652)], n = 83). Upon multivariable multinomial logistic regression, patients with a narrower phase angle and patients who had received a graft from a deceased donor had a higher risk of being in the >75th CAC-percentile. CONCLUSIONS: This study identifies not only metabolic and transplant-related factors, but also phase angle, a composite marker of cell integrity, as an independent determinant of CAC at 12 months after kidney transplantation. This study offers new perspectives for future research into the value of bioelectrical impedance analysis in relation to vascular calcification in kidney transplant recipients.

Who is at risk of lung nodules on low-dose CT in a Western country? A population-based approach.

BACKGROUND: This population-based study aimed to identify the risk factors for lung nodules in a Western European general population. METHODS: We quantified the presence or absence of lung nodules among 12 055 participants of the Dutch population-based ImaLife (Imaging in Lifelines) study (age ≥45 years) who underwent low-dose chest computed tomography. Outcomes included the presence of 1) at least one solid lung nodule (volume ≥30 mm3) and 2) a clinically relevant lung nodule (volume ≥100 mm3). Fully adjusted multivariable logistic regression models were applied overall and stratified by smoking status to identify independent risk factors for the presence of nodules. RESULTS: Among the 12 055 participants (44.1% male; median age 60 years; 39.9% never-smokers; 98.7% White), we found lung nodules in 41.8% (5045 out of 12 055) and clinically relevant nodules in 11.4% (1377 out of 12 055); the corresponding figures among never-smokers were 38.8% and 9.5%, respectively. Factors independently associated with increased odds of having any lung nodule included male sex, older age, low educational level, former smoking, asbestos exposure and COPD. Among never-smokers, a family history of lung cancer increased the odds of both lung nodules and clinically relevant nodules. Among former and current smokers, low educational level was positively associated with lung nodules, whereas being overweight was negatively associated. Among current smokers, asbestos exposure and low physical activity were associated with clinically relevant nodules. CONCLUSIONS: The study provides a large-scale evaluation of lung nodules and associated risk factors in a Western European general population: lung nodules and clinically relevant nodules were prevalent, and never-smokers with a family history of lung cancer were a non-negligible group.

Ct-defined emphysema prevalence in a Chinese and Dutch general population.

PURPOSE: We determine and compare the prevalence, subtypes, severity, and risk factors for emphysema assessed by low-dose CT(LDCT) in Chinese and Dutch general populations. METHODS: This cross-sectional study included LDCT scans of 1143 participants between May and October 2017 from a Chinese Cohort study and 1200 participants with same age range and different smoking status between May and October 2019 from a Dutch population-based study. An experienced radiologist visually assessed the scans for emphysema presence (≥trace), subtype, and severity. Logistic regression analyses, overall and stratified by smoking status, were performed and adjusted for fume exposure, demographic and smoking data. RESULTS: The Chinese population had a comparable proportion of women to the Dutch population (54.9 % vs 58.9 %), was older (61.7 ± 6.3 vs 59.8 ± 8.1), included more never smokers (66.4 % vs 38.3 %), had a higher emphysema prevalence ([58.8 % vs 39.7 %], adjusted odds ratio, aOR = 2.06, 95 %CI = 1.68-2.53), and more often had centrilobular emphysema (54.8 % vs 32.8 %, p < 0.001), but no differences in emphysema severity. After stratification, only in never smokers an increased odds of emphysema was observed in the Chinese compared to the Dutch (aOR = 2.55, 95 %CI = 1.95-3.35). Never smokers in both populations shared older age (aOR = 1.59, 95 %CI = 1.25-2.02 vs 1.26, 95 %CI = 0.97-1.64) and male sex (aOR = 1.50, 95 %CI = 1.02-2.22 vs 1.93, 95 %CI = 1.26-2.96) as risk factors for emphysema. CONCLUSIONS: Only never smokers had a higher prevalence of mainly centrilobular emphysema in the Chinese general population compared to the Dutch after adjusting for confounders, indicating that factors other than smoking, age and sex contribute to presence of CT-defined emphysema.

Effect of emphysema on AI software and human reader performance in lung nodule detection from low-dose chest CT.

BACKGROUND: Emphysema influences the appearance of lung tissue in computed tomography (CT). We evaluated whether this affects lung nodule detection by artificial intelligence (AI) and human readers (HR). METHODS: Individuals were selected from the "Lifelines" cohort who had undergone low-dose chest CT. Nodules in individuals without emphysema were matched to similar-sized nodules in individuals with at least moderate emphysema. AI results for nodular findings of 30-100 mm3 and 101-300 mm3 were compared to those of HR; two expert radiologists blindly reviewed discrepancies. Sensitivity and false positives (FPs)/scan were compared for emphysema and non-emphysema groups. RESULTS: Thirty-nine participants with and 82 without emphysema were included (n = 121, aged 61 ± 8 years (mean ± standard deviation), 58/121 males (47.9%)). AI and HR detected 196 and 206 nodular findings, respectively, yielding 109 concordant nodules and 184 discrepancies, including 118 true nodules. For AI, sensitivity was 0.68 (95% confidence interval 0.57-0.77) in emphysema versus 0.71 (0.62-0.78) in non-emphysema, with FPs/scan 0.51 and 0.22, respectively (p = 0.028). For HR, sensitivity was 0.76 (0.65-0.84) and 0.80 (0.72-0.86), with FPs/scan of 0.15 and 0.27 (p = 0.230). Overall sensitivity was slightly higher for HR than for AI, but this difference disappeared after the exclusion of benign lymph nodes. FPs/scan were higher for AI in emphysema than in non-emphysema (p = 0.028), while FPs/scan for HR were higher than AI for 30-100 mm3 nodules in non-emphysema (p = 0.009). CONCLUSIONS: AI resulted in more FPs/scan in emphysema compared to non-emphysema, a difference not observed for HR. RELEVANCE STATEMENT: In the creation of a benchmark dataset to validate AI software for lung nodule detection, the inclusion of emphysema cases is important due to the additional number of FPs. KEY POINTS: • The sensitivity of nodule detection by AI was similar in emphysema and non-emphysema. • AI had more FPs/scan in emphysema compared to non-emphysema. • Sensitivity and FPs/scan by the human reader were comparable for emphysema and non-emphysema. • Emphysema and non-emphysema representation in benchmark dataset is important for validating AI.

ESR Essentials: imaging in stable chest pain - practice recommendations by ESCR.

Stable chest pain is a common symptom with multiple potential causes. Non-invasive imaging has an important role in diagnosis and guiding management through the assessment of coronary stenoses, atherosclerotic plaque, myocardial ischaemia or infarction, and cardiac function. Computed tomography (CT) provides the anatomical evaluation of coronary artery disease (CAD) with the assessment of stenosis, plaque type and plaque burden, with additional functional information available from CT fractional flow reserve (FFR) or CT myocardial perfusion imaging. Stress magnetic resonance imaging, nuclear stress myocardial perfusion imaging, and stress echocardiography can assess myocardial ischaemia and other cardiac functional parameters. Coronary CT angiography can be used as a first-line test for many patients with stable chest pain, particularly those with low to intermediate pre-test probability. Functional testing may be considered for patients with known CAD, where the clinical significance is uncertain based on anatomical testing, or in patients with high pre-test probability. This practice recommendations document can be used to guide the selection of non-invasive imaging for patients with stable chest pain and provides brief recommendations on how to perform and report these diagnostic tests. KEY POINTS: The selection of non-invasive imaging tests for patients with stable chest pain should be based on symptoms, pre-test probability, and previous history. Coronary CT angiography can be used as a first-line test for many patients with stable chest pain, particularly those with low to intermediate pre-test probability. Functional testing can be considered for patients with known CAD, where the clinical significance of CAD is uncertain based on anatomical testing, or in patients with high pre-test probability. KEY RECOMMENDATIONS: Non-invasive imaging is an important part of the assessment of patients with stable chest pain. The selection of non-invasive imaging test should be based on symptoms, pre-test probability, and previous history. (Level of evidence: High). Coronary CT angiography can be used as a first line test for many patients with stable chest pain, particularly those with low to intermediate pre-test probability. CT provides information on stenoses, plaque type, plaque volume, and if required functional information with CT fractional flow reserve or CT perfusion. (Level of evidence: High). Functional testing can be considered for patients with known CAD, where the clinical significance of CAD is uncertain based on anatomical testing, or in patients with high pre-test probability. Stress MRI, SPECT, PET, and echocardiography can provide information on myocardial ischemia, along with cardiac functional and other information. (Level of evidence: Medium).

The relationship of fat and muscle measurements with emphysema and bronchial wall thickening in smokers.

Introduction: Differences in body composition in patients with COPD may have important prognostic value and may provide opportunities for patient-specific management. We investigated the relation of thoracic fat and muscle with computed tomography (CT)-measured emphysema and bronchial wall thickening. Methods: Low-dose baseline chest CT scans from 1031 male lung cancer screening participants from one site were quantified for emphysema, bronchial wall thickening, subcutaneous fat, visceral fat and skeletal muscle. Body composition measurements were performed by segmenting the first slice above the aortic arch using Hounsfield unit thresholds with region growing and manual corrections. COPD presence and severity were evaluated with pre-bronchodilator spirometry testing. Results: Participants had a median age of 61.5 years (58.6-65.6, 25th-75th percentile) and median number of 38.0 pack-years (28.0-49.5); 549 (53.2%) were current smokers. Overall, 396 (38.4%) had COPD (256 Global Initiative for Chronic Obstructive Lung Disease (GOLD) 1, 140 GOLD 2-3). Participants with COPD had less subcutaneous fat, visceral fat and skeletal muscle (p<0.001 for all). With increasing GOLD stages, subcutaneous (p=0.005) and visceral fat values (p=0.004) were higher, and skeletal muscle was lower (p=0.004). With increasing severity of CT-derived emphysema, subcutaneous fat, visceral fat and skeletal muscle values were lower (p<0.001 for all). With increasing CT-derived bronchial wall thickness, subcutaneous and visceral fat values were higher (p<0.001 for both), without difference in skeletal muscle. All statistical relationships remained when adjusted for age, pack-years and smoking status. Conclusion: COPD presence and emphysema severity are associated with smaller amounts of thoracic fat and muscle, whereas bronchial wall thickening is associated with fat accumulation.

Enhancing a deep learning model for pulmonary nodule malignancy risk estimation in chest CT with uncertainty estimation.

OBJECTIVE: To investigate the effect of uncertainty estimation on the performance of a Deep Learning (DL) algorithm for estimating malignancy risk of pulmonary nodules. METHODS AND MATERIALS: In this retrospective study, we integrated an uncertainty estimation method into a previously developed DL algorithm for nodule malignancy risk estimation. Uncertainty thresholds were developed using CT data from the Danish Lung Cancer Screening Trial (DLCST), containing 883 nodules (65 malignant) collected between 2004 and 2010. We used thresholds on the 90th and 95th percentiles of the uncertainty score distribution to categorize nodules into certain and uncertain groups. External validation was performed on clinical CT data from a tertiary academic center containing 374 nodules (207 malignant) collected between 2004 and 2012. DL performance was measured using area under the ROC curve (AUC) for the full set of nodules, for the certain cases and for the uncertain cases. Additionally, nodule characteristics were compared to identify trends for inducing uncertainty. RESULTS: The DL algorithm performed significantly worse in the uncertain group compared to the certain group of DLCST (AUC 0.62 (95% CI: 0.49, 0.76) vs 0.93 (95% CI: 0.88, 0.97); p < .001) and the clinical dataset (AUC 0.62 (95% CI: 0.50, 0.73) vs 0.90 (95% CI: 0.86, 0.94); p < .001). The uncertain group included larger benign nodules as well as more part-solid and non-solid nodules than the certain group. CONCLUSION: The integrated uncertainty estimation showed excellent performance for identifying uncertain cases in which the DL-based nodule malignancy risk estimation algorithm had significantly worse performance. CLINICAL RELEVANCE STATEMENT: Deep Learning algorithms often lack the ability to gauge and communicate uncertainty. For safe clinical implementation, uncertainty estimation is of pivotal importance to identify cases where the deep learning algorithm harbors doubt in its prediction. KEY POINTS: • Deep learning (DL) algorithms often lack uncertainty estimation, which potentially reduce the risk of errors and improve safety during clinical adoption of the DL algorithm. • Uncertainty estimation identifies pulmonary nodules in which the discriminative performance of the DL algorithm is significantly worse. • Uncertainty estimation can further enhance the benefits of the DL algorithm and improve its safety and trustworthiness.

Competence of radiologists in cardiac CT and MR imaging in Europe: insights from the ESCR Registry.

RATIONALE: To provide an overview of the current status of cardiac multimodality imaging practices in Europe and radiologist involvement using data from the European Society of Cardiovascular Radiology (ESCR) MRCT-registry. MATERIALS AND METHODS: Numbers on cardiac CT and MRI examinations were extracted from the MRCT-registry of the ESCR, entered between January 2011 and October 2023 (n = 432,265). Data collection included the total/annual numbers of examinations, indications, complications, and reporting habits. RESULTS: Thirty-two countries contributed to the MRCT-registry, including 29 European countries. Between 2011 and 2022, there was a 4.5-fold increase in annually submitted CT examinations, from 3368 to 15,267, and a 3.8-fold increase in MRI examinations, from 3445 to 13,183. The main indications for cardiac CT were suspected coronary artery disease (CAD) (59%) and transcatheter aortic valve replacement planning (21%). The number of patients with intermediate pretest probability who underwent CT for suspected CAD showed an increase from 61% in 2012 to 82% in 2022. The main MRI indications were suspected myocarditis (26%), CAD (21%), and suspected cardiomyopathy (19%). Adverse event rates were very low for CT (0.3%) and MRI (0.7%) examinations. Reporting of CT and MRI examinations was performed mainly by radiologists (respectively 76% and 71%) and, to a lesser degree, in consensus with non-radiologists (19% and 27%, respectively). The remaining examinations (4.9% CT and 1.7% MRI) were reported by non-radiological specialties or in separate readings of radiologists and non-radiologists. CONCLUSIONS: Real-life data on cardiac imaging in Europe using the largest available MRCT-registry demonstrate a considerable increase in examinations over the past years, the vast majority of which are read by radiologists. These findings indicate that radiologists contribute to meeting the increasing demands of competent and effective care in cardiac imaging to a relevant extent. CLINICAL RELEVANCE STATEMENT: The number of cardiac CT and MRI examinations has risen over the past years, and radiologists read the vast majority of these studies as recorded in the MRCT-registry. KEY POINTS: • The number of cardiac imaging examinations is constantly increasing. • Radiologists play a central role in providing cardiac CT and MR imaging services to a large volume of patients. • Cardiac CT and MR imaging examinations performed and read by radiologists show a good safety profile.

A framework to integrate artificial intelligence training into radiology residency programs: preparing the future radiologist.

OBJECTIVES: To present a framework to develop and implement a fast-track artificial intelligence (AI) curriculum into an existing radiology residency program, with the potential to prepare a new generation of AI conscious radiologists. METHODS: The AI-curriculum framework comprises five sequential steps: (1) forming a team of AI experts, (2) assessing the residents' knowledge level and needs, (3) defining learning objectives, (4) matching these objectives with effective teaching strategies, and finally (5) implementing and evaluating the pilot. Following these steps, a multidisciplinary team of AI engineers, radiologists, and radiology residents designed a 3-day program, including didactic lectures, hands-on laboratory sessions, and group discussions with experts to enhance AI understanding. Pre- and post-curriculum surveys were conducted to assess participants' expectations and progress and were analyzed using a Wilcoxon rank-sum test. RESULTS: There was 100% response rate to the pre- and post-curriculum survey (17 and 12 respondents, respectively). Participants' confidence in their knowledge and understanding of AI in radiology significantly increased after completing the program (pre-curriculum means 3.25 ± 1.48 (SD), post-curriculum means 6.5 ± 0.90 (SD), p-value = 0.002). A total of 75% confirmed that the course addressed topics that were applicable to their work in radiology. Lectures on the fundamentals of AI and group discussions with experts were deemed most useful. CONCLUSION: Designing an AI curriculum for radiology residents and implementing it into a radiology residency program is feasible using the framework presented. The 3-day AI curriculum effectively increased participants' perception of knowledge and skills about AI in radiology and can serve as a starting point for further customization. CRITICAL RELEVANCE STATEMENT: The framework provides guidance for developing and implementing an AI curriculum in radiology residency programs, educating residents on the application of AI in radiology and ultimately contributing to future high-quality, safe, and effective patient care. KEY POINTS: • AI education is necessary to prepare a new generation of AI-conscious radiologists. • The AI curriculum increased participants' perception of AI knowledge and skills in radiology. • This five-step framework can assist integrating AI education into radiology residency programs.

Performance of Lung-RADS in different target populations: a systematic review and meta-analysis.

OBJECTIVES: Multiple lung cancer screening studies reported the performance of Lung CT Screening Reporting and Data System (Lung-RADS), but none systematically evaluated its performance across different populations. This systematic review and meta-analysis aimed to evaluate the performance of Lung-RADS (versions 1.0 and 1.1) for detecting lung cancer in different populations. METHODS: We performed literature searches in PubMed, Web of Science, Cochrane Library, and Embase databases on October 21, 2022, for studies that evaluated the accuracy of Lung-RADS in lung cancer screening. A bivariate random-effects model was used to estimate pooled sensitivity and specificity, and heterogeneity was explored in stratified and meta-regression analyses. RESULTS: A total of 31 studies with 104,224 participants were included. For version 1.0 (27 studies, 95,413 individuals), pooled sensitivity was 0.96 (95% confidence interval [CI]: 0.90-0.99) and pooled specificity was 0.90 (95% CI: 0.87-0.92). Studies in high-risk populations showed higher sensitivity (0.98 [95% CI: 0.92-0.99] vs. 0.84 [95% CI: 0.50-0.96]) and lower specificity (0.87 [95% CI: 0.85-0.88] vs. 0.95 (95% CI: 0.92-0.97]) than studies in general populations. Non-Asian studies tended toward higher sensitivity (0.97 [95% CI: 0.91-0.99] vs. 0.91 [95% CI: 0.67-0.98]) and lower specificity (0.88 [95% CI: 0.85-0.90] vs. 0.93 [95% CI: 0.88-0.96]) than Asian studies. For version 1.1 (4 studies, 8811 individuals), pooled sensitivity was 0.91 (95% CI: 0.83-0.96) and specificity was 0.81 (95% CI: 0.67-0.90). CONCLUSION: Among studies using Lung-RADS version 1.0, considerable heterogeneity in sensitivity and specificity was noted, explained by population type (high risk vs. general), population area (Asia vs. non-Asia), and cancer prevalence. CLINICAL RELEVANCE STATEMENT: Meta-regression of lung cancer screening studies using Lung-RADS version 1.0 showed considerable heterogeneity in sensitivity and specificity, explained by the different target populations, including high-risk versus general populations, Asian versus non-Asian populations, and populations with different lung cancer prevalence. KEY POINTS: • High-risk population studies showed higher sensitivity and lower specificity compared with studies performed in general populations by using Lung-RADS version 1.0. • In non-Asian studies, the diagnostic performance of Lung-RADS version 1.0 tended to be better than in Asian studies. • There are limited studies on the performance of Lung-RADS version 1.1, and evidence is lacking for Asian populations.

Explainable machine learning model based on clinical factors for predicting the disappearance of indeterminate pulmonary nodules.

BACKGROUND: During lung cancer screening, indeterminate pulmonary nodules (IPNs) are a frequent finding. We aim to predict whether IPNs are resolving or non-resolving to reduce follow-up examinations, using machine learning (ML) models. We incorporated dedicated techniques to enhance prediction explainability. METHODS: In total, 724 IPNs (size 50-500 mm3, 575 participants) from the Dutch-Belgian Randomized Lung Cancer Screening Trial were used. We implemented six ML models and 14 factors to predict nodule disappearance. Random search was applied to determine the optimal hyperparameters on the training set (579 nodules). ML models were trained using 5-fold cross-validation and tested on the test set (145 nodules). Model predictions were evaluated by utilizing the recall, precision, F1 score, and the area under the receiver operating characteristic curve (AUC). The best-performing model was used for three feature importance techniques: mean decrease in impurity (MDI), permutation feature importance (PFI), and SHAPley Additive exPlanations (SHAP). RESULTS: The random forest model outperformed the other ML models with an AUC of 0.865. This model achieved a recall of 0.646, a precision of 0.816, and an F1 score of 0.721. The evaluation of feature importance achieved consistent ranking across all three methods for the most crucial factors. The MDI, PFI, and SHAP methods highlighted volume, maximum diameter, and minimum diameter as the top three factors. However, the remaining factors revealed discrepant ranking across methods. CONCLUSION: ML models effectively predict IPN disappearance using participant demographics and nodule characteristics. Explainable techniques can assist clinicians in developing understandable preliminary assessments.

Willingness to participate in combination screening for lung cancer, chronic obstructive pulmonary disease and cardiovascular disease in four European countries.

OBJECTIVES: Lung cancer screening (LCS), using low-dose computed tomography (LDCT), can be more efficient by simultaneously screening for chronic obstructive pulmonary disease (COPD) and cardiovascular disease (CVD), the Big-3 diseases. This study aimed to determine the willingness to participate in (combinations of) Big-3 screening in four European countries and the relative importance of amendable participation barriers. METHODS: An online cross-sectional survey aimed at (former) smokers aged 50-75 years elicited the willingness of individuals to participate in Big-3 screening and used analytical hierarchy processing (AHP) to determine the importance of participation barriers. RESULTS: Respondents were from France (n = 391), Germany (n = 338), Italy (n = 399), and the Netherlands (n = 342), and consisted of 51.2% men. The willingness to participate in screening was marginally influenced by the diseases screened for (maximum difference of 3.1%, for Big-3 screening (73.4%) vs. lung cancer and COPD screening (70.3%)) and by country (maximum difference of 3.7%, between France (68.5%) and the Netherlands (72.3%)). The largest effect on willingness to participate was personal perceived risk of lung cancer. The most important barriers were the missed cases during screening (weight 0.19) and frequency of screening (weight 0.14), while diseases screened for (weight 0.11) ranked low. CONCLUSIONS: The difference in willingness to participate in LCS showed marginal increase with inclusion of more diseases and limited variation between countries. A marginal increase in participation might result in a marginal additional benefit of Big-3 screening. The amendable participation barriers are similar to previous studies, and the new criterion, diseases screened for, is relatively unimportant. CLINICAL RELEVANCE STATEMENT: Adding diseases to combination screening modestly improves participation, driven by personal perceived risk. These findings guide program design and campaigns for lung cancer and Big-3 screening. Benefits of Big-3 screening lie in long-term health and economic impact, not participation increase. KEY POINTS: • It is unknown whether or how combination screening might affect participation. • The addition of chronic obstructive pulmonary disease and cardiovascular disease to lung cancer screening resulted in a marginal increase in willingness to participate. • The primary determinant influencing individuals' engagement in such programs is their personal perceived risk of the disease.

Artificial CT images can enhance variation of case images in diagnostic radiology skills training.

OBJECTIVES: We sought to investigate if artificial medical images can blend with original ones and whether they adhere to the variable anatomical constraints provided. METHODS: Artificial images were generated with a generative model trained on publicly available standard and low-dose chest CT images (805 scans; 39,803 2D images), of which 17% contained evidence of pathological formations (lung nodules). The test set (90 scans; 5121 2D images) was used to assess if artificial images (512 × 512 primary and control image sets) blended in with original images, using both quantitative metrics and expert opinion. We further assessed if pathology characteristics in the artificial images can be manipulated. RESULTS: Primary and control artificial images attained an average objective similarity of 0.78 ± 0.04 (ranging from 0 [entirely dissimilar] to 1[identical]) and 0.76 ± 0.06, respectively. Five radiologists with experience in chest and thoracic imaging provided a subjective measure of image quality; they rated artificial images as 3.13 ± 0.46 (range of 1 [unrealistic] to 4 [almost indistinguishable to the original image]), close to their rating of the original images (3.73 ± 0.31). Radiologists clearly distinguished images in the control sets (2.32 ± 0.48 and 1.07 ± 0.19). In almost a quarter of the scenarios, they were not able to distinguish primary artificial images from the original ones. CONCLUSION: Artificial images can be generated in a way such that they blend in with original images and adhere to anatomical constraints, which can be manipulated to augment the variability of cases. CRITICAL RELEVANCE STATEMENT: Artificial medical images can be used to enhance the availability and variety of medical training images by creating new but comparable images that can blend in with original images. KEY POINTS: • Artificial images, similar to original ones, can be created using generative networks. • Pathological features of artificial images can be adjusted through guiding the network. • Artificial images proved viable to augment the depth and broadening of diagnostic training.