Performance evaluation of a 4D similarity filter for dynamic CT angiography imaging of the liver.

BACKGROUND: Dynamic computed tomography (CT) angiography of the abdomen provides perfusion information and characteristics of the tissues present in the abdomen. This information could potentially help characterize liver metastases. However, radiation dose has to be relatively low for the patient, causing the images to have very high noise content. Denoising methods are needed to increase image quality. PURPOSE: The purpose of this study was to investigate the performance, limitations, and behavior of a new 4D filtering method, called the 4D Similarity Filter (4DSF), to reduce image noise in temporal CT data. METHODS: The 4DSF averages voxels with similar time-intensity curves (TICs). Each phase is filtered individually using the information of all phases except for the one being filtered. This approach minimizes the bias toward the noise initially present in this phase. Since the 4DSF does not base similarity on spatial proximity, loss of spatial resolution is avoided. The 4DSF was evaluated on a 12-phase liver dynamic CT angiography acquisition of 52 digital anthropomorphic phantoms, each containing one hypervascular 1 cm lesion with a small necrotic core. The metrics used for evaluation were noise reduction, lesion contrast-to-noise ratio (CNR), CT number accuracy using peak-time and peak-intensity of the TICs, and resolution loss. The results were compared to those obtained by the time-intensity profile similarity (TIPS) filter, which uses the whole TIC for determining similarity, and the combination 4DSF followed by TIPS filter (4DSF + TIPS). RESULTS: The 4DSF alone resulted in a median noise reduction by a factor of 6.8, which is lower than that obtained by the TIPS filter at 8.1, and 4DSF + TIPS at 12.2. The 4DSF increased the median CNR from 0. 44 to 1.85, which is less than the TIPS filter at 2.59 and 4DSF + TIPS at 3.12. However, the peak-intensity accuracy in the TICs was superior for the 4DSF, with a median intensity decrease of -34 HU compared to -88 and -50 HU for the hepatic artery when using the TIPS filter and 4DSF + TIPS, respectively. The median peak-time accuracy was inferior for the 4DSF filter and 4DSF + TIPS, with a time shift of -1 phases for the portal vein TIC compared to no shift in time when using the TIPS. The analysis of the full-width-at-half-maximum (FWHM) of a small artery showed significantly less spatial resolution loss for the 4DSF at 3.2 pixels, compared to the TIPS filter at 4.3 pixels, and 3.4 pixels for the 4DSF + TIPS. CONCLUSION: The 4DSF can reduce noise with almost no resolution loss, making the filter very suitable for denoising 4D CT data for detection tasks, even in very low dose, i.e., very high noise level, situations. In combination with the TIPS filter, the noise reduction can be increased even further.

Cost-effectiveness analysis of a lung cancer screening program in the netherlands: a simulation based on NELSON and NLST study outcomes.

BACKGROUND: In the Netherlands, lung cancer is the leading cause of cancer-related death, accounting for more than 10,000 annual deaths. Lung cancer screening (LCS) studies using low-dose computed tomography (LDCT) have demonstrated that early detection reduces lung cancer mortality. However, no LCS program has been implemented yet in the Netherlands. A national LCS program has the potential to enhance the health outcomes for lung cancer patients in the Netherlands. OBJECTIVE AND METHODS: This study evaluates the cost-effectiveness of LCS compared to no-screening in the Netherlands, by simulating the screening outcomes based on data from NEderlands-Leuvens Longkanker Screenings ONderzoek (NELSON) and National Lung Screening Trial (NLST). We simulated annual screening up to 74 years of age, using inclusion criteria from the respective studies. A decision tree and Markov model was used to predict the incremental costs, quality-adjusted life-years (QALYs), and incremental cost-effectiveness ratio (ICERs) for the screening population. The analysis used a lifetime horizon and a societal perspective. RESULTS: Compared to no-screening, LCS resulted in an ICER of €5,169 per QALY for the NELSON simulation, and an ICER of €17,119 per QALY for the NLST simulation. The screening costs were highly impactful for the cost-effectiveness. The most influential parameter was the CT scan cost. Cost reduction for CT from €201 to €101 per scan would reduce the ICER to €2,335 using NELSON criteria. Additionally, LCS could prevent 15,115 and 12,611 premature lung cancer deaths, accompanied by 1.66 and 1.31 QALYs gained per lung cancer case for the NELSON and NLST simulations, respectively. CONCLUSION: LCS was estimated to be cost-effective in the Netherlands for both simulations at a willingness-to-pay threshold of €20,000 per QALY. Using the NELSON criteria, less than €5,500 per QALY had to be spent. Lowering the cost per CT exam would lead to a further reduction of this amount.

Feasibility of capturing vessel expansion with 4D-CTA: Phantom study to determine reproducibility, spatial and temporal resolution.

BACKGROUND: Dynamic Computed Tomography Angiography (4D CTA) has the potential of providing insight into the biomechanical properties of the vessel wall, by capturing motion of the vessel wall. For vascular pathologies, like intracranial aneurysms, this could potentially refine diagnosis, prognosis, and treatment decision-making. PURPOSE: The objective of this research is to determine the feasibility of a 4D CTA scanner for accurately measuring harmonic diameter changes in an in-vitro simulated vessel. METHODS: A silicon tube was exposed to a simulated heartbeat. Simulated heart rates between 40 and 100 beats-per-minute (bpm) were tested and the flow amplitude was varied, resulting in various changes of tube diameter. A 320-detector row CT system with ECG-gating captured three consecutive cycles of expansion. Image registration was used to calculate the diameter change. A vascular echography set-up was used as a reference, using a 9 MHz linear array transducer. The reproducibility of 4D CTA was represented by the Pearson correlation (r) between the three consecutive diameter change patterns, captured by 4D CTA. The peak value similarity (pvs) was calculated between the 4D CTA and US measurements for increasing frequencies and was chosen as a measure of temporal resolution. Spatial resolution was represented by the Sum of the Relative Percentual Difference (SRPD) between 4D CTA and US diameter change patterns for increasing amplitudes. RESULTS: The reproducibility of 4D CTA measurements was good (r ≥ 0.9) if the diameter change was larger than 0.3 mm, moderate (0.7 ≤ r < 0.9) if the diameter change was between 0.1 and 0.3 mm, and low (r < 0.7) if the diameter change was smaller than 0.1 mm. Regarding the temporal resolution, the amplitude of 4D CTA was similar to the US measurements (pvs ≥ 90%) for the frequencies of 40 and 50 bpm. Frequencies between 60 and 80 bpm result in a moderate similarity (70% ≤ pvs < 90%). A low similarity (pvs < 70%) is observed for 90 and 100 bpm. Regarding the spatial resolution, diameter changes above 0.30 mm result in SRPDs consistently below 50%. CONCLUSION: In a phantom setting, 4D CTA can be used to reliably capture reproducible tube diameter changes exceeding 0.30 mm. Low pulsation frequencies (40 or 50 bpm) provide an accurate measurement of the maximum tube diameter change.

Clinical applicability of signal heterogeneity and tumor border assessment on T2-weighted MR images to distinguish astrocytic from oligodendroglial origin of gliomas.

BACKGROUND AND PURPOSE: Radiological features on magnetic resonance imaging (MRI) were attributed to oligodendroglioma, although the diagnostic accuracy in a real-world clinical setting remains partially elusive. This study investigated the accuracy and robustness of tumor heterogeneity and tumor border delineation on T2-weighted MRI to distinguish oligodendroglioma from astrocytoma. MATERIALS AND METHODS: Eight readers from three different specialties (radiology, neurology, neurosurgery) with varying levels of experience blindly rated 79 T2-weighted MR images of patients with either oligodendroglioma or astrocytoma. After the first reading session, all readers were re-invited for a second reading session within three weeks. Diagnostic accuracy, including area under the receiver operator characteristics curve (AUC), and intra-observer variability and inter-observer variability were used as outcome measures. RESULTS: Pooled sensitivity and specificity to distinguish oligodendroglioma from astrocytoma for the use of tumor heterogeneity were 59.9 % respectively 74.5 %, and 85.7 % respectively 40.1 % for tumor border. A second reading session did not result in a significant change in sensitivity or specificity for tumor heterogeneity (P = 0.752 and P = 0.733, respectively) or tumor border (P = 0.309 and P = 0.271, respectively). An AUC of 0.825 was achieved with regard to predicting oligodendroglial origin of gliomas. Intra-observer agreement ranged from moderate to very good for tumor heterogeneity (kappa-value 0.43-0.87) and tumor border (0.40-0.84). A moderate inter-oberserver agreement was achieved for tumor heterogeneity and tumor border (kappa-value of 0.50 and 0.45, respectively). CONCLUSION: This study demonstrates that tumor heterogeneity and tumor borders on T2-weighted MRI could be used with moderate Finter-observer agreement to non-invasively distinguish oligodendroglioma from astrocytoma.

Ferumoxtran-10-enhanced MRI for pre-operative metastatic lymph node detection in pancreatic, duodenal, or periampullary adenocarcinoma.

OBJECTIVES: To assess 3-Tesla (3-T) ultra-small superparamagnetic iron oxide (USPIO)-enhanced MRI in detecting lymph node (LN) metastases for resectable adenocarcinomas of the pancreas, duodenum, or periampullary region in a node-to-node validation against histopathology. METHODS: Twenty-seven consecutive patients with a resectable pancreatic, duodenal, or periampullary adenocarcinoma were enrolled in this prospective single expert centre study. Ferumoxtran-10-enhanced 3-T MRI was performed pre-surgery. LNs found on MRI were scored for suspicion of metastasis by two expert radiologists using a dedicated scoring system. Node-to-node matching from in vivo MRI to histopathology was performed using a post-operative ex vivo 7-T MRI of the resection specimen. Sensitivity and specificity were calculated using crosstabs. RESULTS: Eighteen out of 27 patients (median age 65 years, 11 men) were included in the final analysis (pre-surgery withdrawal n = 4, not resected because of unexpected metastases peroperatively n = 2, and excluded because of inadequate contrast-agent uptake n = 3). On MRI 453 LNs with a median size of 4.0 mm were detected, of which 58 (13%) were classified as suspicious. At histopathology 385 LNs with a median size of 5.0 mm were found, of which 45 (12%) were metastatic. For 55 LNs node-to-node matching was possible. Analysis of these 55 matched LNs, resulted in a sensitivity and specificity of 83% (95% CI: 36-100%) and 92% (95% CI: 80-98%), respectively. CONCLUSION: USPIO-enhanced MRI is a promising technique to preoperatively detect and localise LN metastases in patients with pancreatic, duodenal, or periampullary adenocarcinoma. CLINICAL RELEVANCE STATEMENT: Detection of (distant) LN metastases with USPIO-enhanced MRI could be used to determine a personalised treatment strategy that could involve neoadjuvant or palliative chemotherapy, guided resection of distant LNs, or targeted radiotherapy. REGISTRATION: The study was registered on clinicaltrials.gov NCT04311047. https://clinicaltrials.gov/ct2/show/NCT04311047?term=lymph+node&cond=Pancreatic+Cancer&cntry=NL&draw=2&rank=1 . KEY POINTS: LN metastases of pancreatic, duodenal, or periampullary adenocarcinoma cannot be reliably detected with current imaging. This technique detected LN metastases with a sensitivity and specificity of 83% and 92%, respectively. MRI with ferumoxtran-10 is a promising technique to improve preoperative staging in these cancers.

Performance of AI to exclude normal chest radiographs to reduce radiologists' workload.

INTRODUCTION: This study investigates the performance of a commercially available artificial intelligence (AI) system to identify normal chest radiographs and its potential to reduce radiologist workload. METHODS: Retrospective analysis included consecutive chest radiographs from two medical centers between Oct 1, 2016 and Oct 14, 2016. Exclusions comprised follow-up exams within the inclusion period, bedside radiographs, incomplete images, imported radiographs, and pediatric radiographs. Three chest radiologists categorized findings into normal, clinically irrelevant, clinically relevant, urgent, and critical. A commercial AI system processed all radiographs, scoring 10 chest abnormalities on a 0-100 confidence scale. AI system performance was evaluated using the area under the ROC curve (AUC), assessing the detection of normal radiographs. Sensitivity was calculated for the default and a conservative operating point. the detection of negative predictive value (NPV) for urgent and critical findings, as well as the potential workload reduction, was calculated. RESULTS: A total of 2603 radiographs were acquired in 2141 unique patients. Post-exclusion, 1670 radiographs were analyzed. Categories included 479 normal, 332 clinically irrelevant, 339 clinically relevant, 501 urgent, and 19 critical findings. The AI system achieved an AUC of 0.92. Sensitivity for normal radiographs was 92% at default and 53% at the conservative operating point. At the conservative operating point, NPV was 98% for urgent and critical findings, and could result in a 15% workload reduction. CONCLUSION: A commercially available AI system effectively identifies normal chest radiographs and holds the potential to lessen radiologists' workload by omitting half of the normal exams from reporting. CLINICAL RELEVANCE STATEMENT: The AI system is able to detect half of all normal chest radiographs at a clinically acceptable operating point, thereby potentially reducing the workload for the radiologists by 15%. KEY POINTS: The AI system reached an AUC of 0.92 for the detection of normal chest radiographs. Fifty-three percent of normal chest radiographs were identified with a NPV of 98% for urgent findings. AI can reduce the workload of chest radiography reporting by 15%.

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.

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.

Review of strategies to reduce the contamination of the water environment by gadolinium-based contrast agents.

Gadolinium-based contrast agents (GBCA) are essential for diagnostic MRI examinations. GBCA are only used in small quantities on a per-patient basis; however, the acquisition of contrast-enhanced MRI examinations worldwide results in the use of many thousands of litres of GBCA per year. Data shows that these GBCA are present in sewage water, surface water, and drinking water in many regions of the world. Therefore, there is growing concern regarding the environmental impact of GBCA because of their ubiquitous presence in the aquatic environment. To address the problem of GBCA in the water system as a whole, collaboration is necessary between all stakeholders, including the producers of GBCA, medical professionals and importantly, the consumers of drinking water, i.e. the patients. This paper aims to make healthcare professionals aware of the opportunity to take the lead in making informed decisions about the use of GBCA and provides an overview of the different options for action.In this paper, we first provide a summary on the metabolism and clinical use of GBCA, then the environmental fate and observations of GBCA, followed by measures to reduce the use of GBCA. The environmental impact of GBCA can be reduced by (1) measures focusing on the application of GBCA by means of weight-based contrast volume reduction, GBCA with higher relaxivity per mmol of Gd, contrast-enhancing sequences, and post-processing; and (2) measures that reduce the waste of GBCA, including the use of bulk packaging and collecting residues of GBCA at the point of application.Critical relevance statement This review aims to make healthcare professionals aware of the environmental impact of GBCA and the opportunity for them to take the lead in making informed decisions about GBCA use and the different options to reduce its environmental burden.Key points• Gadolinium-based contrast agents are found in sources of drinking water and constitute an environmental risk.• Radiologists have a wide spectrum of options to reduce GBCA use without compromising diagnostic quality.• Radiology can become more sustainable by adopting such measures in clinical practice.

The Facilitators and Barriers of the Implementation of a Clinical Decision Support System for Breast Cancer Multidisciplinary Team Meetings-An Interview Study.

BACKGROUND: AI-driven clinical decision support systems (CDSSs) hold promise for multidisciplinary team meetings (MDTMs). This study aimed to uncover the hurdles and aids in implementing CDSSs during breast cancer MDTMs. METHODS: Twenty-four core team members from three hospitals engaged in semi-structured interviews, revealing a collective interest in experiencing CDSS workflows in clinical practice. All interviews were audio recorded, transcribed verbatim and analyzed anonymously. A standardized approach, 'the framework method', was used to create an analytical framework for data analysis, which was performed by two independent researchers. RESULTS: Positive aspects included improved data visualization, time-saving features, automated trial matching, and enhanced documentation transparency. However, challenges emerged, primarily concerning data connectivity, guideline updates, the accuracy of AI-driven suggestions, and the risk of losing human involvement in decision making. Despite the complexities involved in CDSS development and integration, clinicians demonstrated enthusiasm to explore its potential benefits. CONCLUSIONS: Acknowledging the multifaceted nature of this challenge, insights into the barriers and facilitators identified in this study offer a potential roadmap for smoother future implementations. Understanding these factors could pave the way for more effective utilization of CDSSs in breast cancer MDTMs, enhancing patient care through informed decision making.

Optimized, Person-Centered Workflow Design for a High-Throughput Breast MRI Screening Facility-A Simulation Study.

OBJECTIVES: This project aims to model an optimal scanning environment for breast magnetic resonance imaging (MRI) screening based on real-life data to identify to what extent the logistics of breast MRI can be optimized. MATERIALS AND METHODS: A novel concept for a breast MRI screening facility was developed considering layout of the building, workflow steps, used resources, and MRI protocols. The envisioned screening facility is person centered and aims for an efficient workflow-oriented design. Real-life data, collected from existing breast MRI screening workflows, during 62 scans in 3 different hospitals, were imported into a 3D simulation software for designing and testing new concepts. The model provided several realistic, virtual, logistical pathways for MRI screening and their outcome measures: throughput, waiting times, and other relevant variables. RESULTS: The total average appointment time in the baseline scenario was 25:54 minutes, with 19:06 minutes of MRI room occupation. Simulated improvements consisted of optimizing processes and resources, facility layout, and scanning protocol. In the simulation, time spent in the MRI room was reduced by introducing an optimized facility layout, dockable tables, and adoption of an abbreviated MRI scanning protocol. The total average appointment time was reduced to 19:36 minutes, and in this scenario, the MRI room was occupied for 06:21 minutes. In the most promising scenario, screening of about 68 people per day (10 hours) on a single MRI scanner could be feasible, compared with 36 people per day in the baseline scenario. CONCLUSIONS: This study suggests that by optimizing workflow MRI for breast screening total appointment duration and MRI occupation can be reduced. A throughput of up to 6 people per hour may be achieved, compared with 3 people per hour in the current setup.

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.

Dynamic Computed Tomography Angiography for capturing vessel wall motion: A phantom study for optimal image reconstruction.

BACKGROUND: Reliably capturing sub-millimeter vessel wall motion over time, using dynamic Computed Tomography Angiography (4D CTA), might provide insight in biomechanical properties of these vessels. This may improve diagnosis, prognosis, and treatment decision making in vascular pathologies. PURPOSE: The aim of this study is to determine the most suitable image reconstruction method for 4D CTA to accurately assess harmonic diameter changes of vessels. METHODS: An elastic tube (inner diameter 6 mm, wall thickness 2 mm) was exposed to sinusoidal pressure waves with a frequency of 70 beats-per-minute. Five flow amplitudes were set, resulting in increasing sinusoidal diameter changes of the elastic tube, measured during three simulated pulsation cycles, using ECG-gated 4D CTA on a 320-detector row CT system. Tomographic images were reconstructed using one of the following three reconstruction methods: hybrid iterative (Hybrid-IR), model-based iterative (MBIR) and deep-learning based (DLR) reconstruction. The three reconstruction methods where based on 180 degrees (half reconstruction mode) and 360 degrees (full reconstruction mode) raw data. The diameter change, captured by 4D CTA, was computed based on image registration. As a reference metric for diameter change measurement, a 9 MHz linear ultrasound transducer was used. The sum of relative absolute differences (SRAD) between the ultrasound and 4D CTA measurements was calculated for each reconstruction method. The standard deviation was computed across the three pulsation cycles. RESULTS: MBIR and DLR resulted in a decreased SRAD and standard deviation compared to Hybrid-IR. Full reconstruction mode resulted in a decreased SRAD and standard deviations, compared to half reconstruction mode. CONCLUSIONS: 4D CTA can capture a diameter change pattern comparable to the pattern captured by US. DLR and MBIR algorithms show more accurate results than Hybrid-IR. Reconstruction with DLR is >3 times faster, compared to reconstruction with MBIR. Full reconstruction mode is more accurate than half reconstruction mode.

Pricing and cost-saving potential for deep-learning computer-aided lung nodule detection software in CT lung cancer screening.

OBJECTIVE: An increasing number of commercial deep learning computer-aided detection (DL-CAD) systems are available but their cost-saving potential is largely unknown. This study aimed to gain insight into appropriate pricing for DL-CAD in different reading modes to be cost-saving and to determine the potentially most cost-effective reading mode for lung cancer screening. METHODS: In three representative settings, DL-CAD was evaluated as a concurrent, pre-screening, and second reader. Scoping review was performed to estimate radiologist reading time with and without DL-CAD. Hourly cost of radiologist time was collected for the USA (€196), UK (€127), and Poland (€45), and monetary equivalence of saved time was calculated. The minimum number of screening CTs to reach break-even was calculated for one-time investment of €51,616 for DL-CAD. RESULTS: Mean reading time was 162 (95% CI: 111-212) seconds per case without DL-CAD, which decreased by 77 (95% CI: 47-107) and 104 (95% CI: 71-136) seconds for DL-CAD as concurrent and pre-screening reader, respectively, and increased by 33-41 s for DL-CAD as second reader. This translates into €1.0-4.3 per-case cost for concurrent reading and €0.8-5.7 for pre-screening reading in the USA, UK, and Poland. To achieve break-even with a one-time investment, the minimum number of CT scans was 12,300-53,600 for concurrent reader, and 9400-65,000 for pre-screening reader in the three countries. CONCLUSIONS: Given current pricing, DL-CAD must be priced substantially below €6 in a pay-per-case setting or used in a high-workload environment to reach break-even in lung cancer screening. DL-CAD as pre-screening reader shows the largest potential to be cost-saving. CRITICAL RELEVANCE STATEMENT: Deep-learning computer-aided lung nodule detection (DL-CAD) software must be priced substantially below 6 euro in a pay-per-case setting or must be used in high-workload environments with one-time investment in order to achieve break-even. DL-CAD as a pre-screening reader has the greatest cost savings potential. KEY POINTS: • DL-CAD must be substantially below €6 in a pay-per-case setting to reach break-even. • DL-CAD must be used in a high-workload screening environment to achieve break-even. • DL-CAD as a pre-screening reader shows the largest potential to be cost-saving.

Impact of time-to-treatment on survival for early-stage non-small cell lung cancer in The Netherlands-a nationwide observational cohort study.

Background: Varied outcomes on the relation between time-to-treatment and survival in early-stage non-small cell lung cancer (NSCLC) patients are reported. We examined this relation in a large multicentric retrospective cohort study and identified factors associated with extended time-to-treatment. Methods: We included 9,536 patients with clinical stage I-II NSCLC, diagnosed and treated in 2014-2019, from the Netherlands Cancer Registry that includes nation-wide data. Time-to-treatment was defined as the number of days between first outpatient visit for suspected lung cancer and start of treatment. The effect of extended time-to-treatment beyond the first quartile and survival was studied with Cox proportional hazard regression. Analyses were stratified for stage and type of therapy. Time-to-treatment was adjusted for multiple covariates including performance status and socioeconomic status. Factors associated with treatment delay were identified by multilevel logistic regression. Results: Median time-to-treatment was 47 days [interquartile range (IQR): 34-65] for stage I and 46 days (IQR: 34-62) for stage II. The first quartile extended to 33 days for both stages. Risk of death increased significantly with extended time-to-treatment for surgical treatment of clinical stage II patients [adjusted hazard ratio (aHR) >33 days: 1.36, 95% confidence intervals (CI): 1.09-1.70], but not in stage II patients treated with radiotherapy or in stage I patients. Causes of prolonged time-to-treatment were multifactorial including diagnostic tests, such as endoscopic ultrasound (EUS) or endobronchial ultrasound (EBUS). Conclusions: Clinical stage II patients benefit from fast initiation of surgical treatment. Surprisingly this appears to be accounted for by patients who are clinically stage II but pathologically stage I. Further study is needed on characterizing these patients and the significance of lymph node- or distant micrometastasis in guiding time-to-treatment and treatment strategy.

Deep learning for the detection of benign and malignant pulmonary nodules in non-screening chest CT scans.

BACKGROUND: Outside a screening program, early-stage lung cancer is generally diagnosed after the detection of incidental nodules in clinically ordered chest CT scans. Despite the advances in artificial intelligence (AI) systems for lung cancer detection, clinical validation of these systems is lacking in a non-screening setting. METHOD: We developed a deep learning-based AI system and assessed its performance for the detection of actionable benign nodules (requiring follow-up), small lung cancers, and pulmonary metastases in CT scans acquired in two Dutch hospitals (internal and external validation). A panel of five thoracic radiologists labeled all nodules, and two additional radiologists verified the nodule malignancy status and searched for any missed cancers using data from the national Netherlands Cancer Registry. The detection performance was evaluated by measuring the sensitivity at predefined false positive rates on a free receiver operating characteristic curve and was compared with the panel of radiologists. RESULTS: On the external test set (100 scans from 100 patients), the sensitivity of the AI system for detecting benign nodules, primary lung cancers, and metastases is respectively 94.3% (82/87, 95% CI: 88.1-98.8%), 96.9% (31/32, 95% CI: 91.7-100%), and 92.0% (104/113, 95% CI: 88.5-95.5%) at a clinically acceptable operating point of 1 false positive per scan (FP/s). These sensitivities are comparable to or higher than the radiologists, albeit with a slightly higher FP/s (average difference of 0.6). CONCLUSIONS: The AI system reliably detects benign and malignant pulmonary nodules in clinically indicated CT scans and can potentially assist radiologists in this setting.

Early-stage lung cancer can be diagnosed after identifying an abnormal spot on a chest CT scan ordered for other medical reasons. These spots or lung nodules can be overlooked by radiologists, as they are not necessarily the focus of an examination and can be as small as a few millimeters. Software using Artificial Intelligence (AI) technology has proven to be successful for aiding radiologists in this task, but its performance is understudied outside a lung cancer screening setting. We therefore developed and validated AI software for the detection of cancerous nodules or non-cancerous nodules that would need attention. We show that the software can reliably detect these nodules in a non-screening setting and could potentially aid radiologists in daily clinical practice.

ERS/ESTS/ESTRO/ESR/ESTI/EFOMP statement on management of incidental findings from low dose CT screening for lung cancer.

BACKGROUND: Screening for lung cancer with low radiation dose computed tomography has a strong evidence base, is being introduced in several European countries and is recommended as a new targeted cancer screening programme. The imperative now is to ensure that implementation follows an evidence-based process that will ensure clinical and cost effectiveness. This European Respiratory Society (ERS) task force was formed to provide an expert consensus for the management of incidental findings which can be adapted and followed during implementation. METHODS: A multi-European society collaborative group was convened. 23 topics were identified, primarily from an ERS statement on lung cancer screening, and a systematic review of the literature was conducted according to ERS standards. Initial review of abstracts was completed and full text was provided to members of the group for each topic. Sections were edited and the final document approved by all members and the ERS Science Council. RESULTS: Nine topics considered most important and frequent were reviewed as standalone topics (interstitial lung abnormalities, emphysema, bronchiectasis, consolidation, coronary calcification, aortic valve disease, mediastinal mass, mediastinal lymph nodes and thyroid abnormalities). Other topics considered of lower importance or infrequent were grouped into generic categories, suitable for general statements. CONCLUSIONS: This European collaborative group has produced an incidental findings statement that can be followed during lung cancer screening. It will ensure that an evidence-based approach is used for reporting and managing incidental findings, which will mean that harms are minimised and any programme is as cost-effective as possible.