Impact of Artificial Intelligence Assistance on Chest CT Interpretation Times: A Prospective Randomized Study.

BACKGROUND. Deep learning-based convolutional neural networks have enabled major advances in development of artificial intelligence (AI) software applications. Modern AI applications offer comprehensive multiorgan evaluation. OBJECTIVE. The purpose of this article was to evaluate the impact of an automated AI platform integrated into clinical workflow for chest CT interpretation on radiologists' interpretation times when evaluated in a real-world clinical setting. METHODS. In this prospective single-center study, a commercial AI software solution was integrated into clinical workflow for chest CT interpretation. The software provided automated analysis of cardiac, pulmonary, and musculoskeletal findings, including labeling, segmenting, and measuring normal structures as well as detecting, labeling, and measuring abnormalities. AI-annotated images and autogenerated summary results were stored in the PACS and available to interpreting radiologists. A total of 390 patients (204 women, 186 men; mean age, 62.8 ± 13.3 [SD] years) who underwent out-patient chest CT between January 19, 2021, and January 28, 2021, were included. Scans were randomized using 1:1 allocation between AI-assisted and non-AI-assisted arms and were clinically interpreted by one of three cardiothoracic radiologists (65 scans per arm per radiologist; total of 195 scans per arm) who recorded interpretation times using a stopwatch. Findings were categorized according to review of report impressions. Interpretation times were compared between arms. RESULTS. Mean interpretation times were significantly shorter in the AI-assisted than in the non-AI-assisted arm for all three readers (289 ± 89 vs 344 ± 129 seconds, p < .001; 449 ± 110 vs 649 ± 82 seconds, p < .001; 281 ± 114 vs 348 ± 93 seconds, p = .01) and for readers combined (328 ± 122 vs 421 ± 175 seconds, p < .001). For readers combined, the mean difference was 93 seconds (95% CI, 63-123 seconds), corresponding with a 22.1% reduction in the AI-assisted arm. Mean interpretation time was also shorter in the AI-assisted arm compared with the non-AI-assisted arm for contrast-enhanced scans (83 seconds), noncontrast scans (104 seconds), negative scans (84 seconds), positive scans without significant new findings (117 seconds), and positive scans with significant new findings (92 seconds). CONCLUSION. Cardiothoracic radiologists exhibited a 22.1% reduction in chest CT interpretations times when they had access to results from an automated AI support platform during real-world clinical practice. CLINICAL IMPACT. Integration of the AI support platform into clinical workflow improved radiologist efficiency.

Automated detection of lung nodules and coronary artery calcium using artificial intelligence on low-dose CT scans for lung cancer screening: accuracy and prognostic value.

BACKGROUND: Artificial intelligence (AI) in diagnostic radiology is undergoing rapid development. Its potential utility to improve diagnostic performance for cardiopulmonary events is widely recognized, but the accuracy and precision have yet to be demonstrated in the context of current screening modalities. Here, we present findings on the performance of an AI convolutional neural network (CNN) prototype (AI-RAD Companion, Siemens Healthineers) that automatically detects pulmonary nodules and quantifies coronary artery calcium volume (CACV) on low-dose chest CT (LDCT), and compare results to expert radiologists. We also correlate AI findings with adverse cardiopulmonary outcomes in a retrospective cohort of 117 patients who underwent LDCT. METHODS: A total of 117 patients were enrolled in this study. Two CNNs were used to identify lung nodules and CACV on LDCT scans. All subjects were used for lung nodule analysis, and 96 subjects met the criteria for coronary artery calcium volume analysis. Interobserver concordance was measured using ICC and Cohen's kappa. Multivariate logistic regression and partial least squares regression were used for outcomes analysis. RESULTS: Agreement of the AI findings with experts was excellent (CACV ICC = 0.904, lung nodules Cohen's kappa = 0.846) with high sensitivity and specificity (CACV: sensitivity = .929, specificity = .960; lung nodules: sensitivity = 1, specificity = 0.708). The AI findings improved the prediction of major cardiopulmonary outcomes at 1-year follow-up including major adverse cardiac events and lung cancer (AUCMACE = 0.911, AUCLung Cancer = 0.942). CONCLUSION: We conclude the AI prototype rapidly and accurately identifies significant risk factors for cardiopulmonary disease on standard screening low-dose chest CT. This information can be used to improve diagnostic ability, facilitate intervention, improve morbidity and mortality, and decrease healthcare costs. There is also potential application in countries with limited numbers of cardiothoracic radiologists.

Comparison of Artificial Intelligence-Based Fully Automatic Chest CT Emphysema Quantification to Pulmonary Function Testing.

OBJECTIVE. The purpose of this study was to evaluate an artificial intelligence (AI)-based prototype algorithm for fully automated quantification of emphysema on chest CT compared with pulmonary function testing (spirometry). MATERIALS AND METHODS. A total of 141 patients (72 women, mean age ± SD of 66.46 ± 9.7 years [range, 23-86 years]; 69 men, mean age of 66.72 ± 11.4 years [range, 27-91 years]) who underwent both chest CT acquisition and spirometry within 6 months were retrospectively included. The spirometry-based Tiffeneau index (TI; calculated as the ratio of forced expiratory volume in the first second to forced vital capacity) was used to measure emphysema severity; a value less than 0.7 was considered to indicate airway obstruction. Segmentation of the lung based on two different reconstruction methods was carried out by using a deep convolution image-to-image network. This multilayer convolutional neural network was combined with multilevel feature chaining and depth monitoring. To discriminate the output of the network from ground truth, an adversarial network was used during training. Emphysema was quantified using spatial filtering and attenuation-based thresholds. Emphysema quantification and TI were compared using the Spearman correlation coefficient. RESULTS. The mean TI for all patients was 0.57 ± 0.13. The mean percentages of emphysema using reconstruction methods 1 and 2 were 9.96% ± 11.87% and 8.04% ± 10.32%, respectively. AI-based emphysema quantification showed very strong correlation with TI (reconstruction method 1, ρ = -0.86; reconstruction method 2, ρ = -0.85; both p < 0.0001), indicating that AI-based emphysema quantification meaningfully reflects clinical pulmonary physiology. CONCLUSION. AI-based, fully automated emphysema quantification shows good correlation with TI, potentially contributing to an image-based diagnosis and quantification of emphysema severity.

A diffusion model-free framework with echo time dependence for free-water elimination and brain tissue microstructure characterization.

PURPOSE: The compartmental nature of brain tissue microstructure is typically studied by diffusion MRI, MR relaxometry or their correlation. Diffusion MRI relies on signal representations or biophysical models, while MR relaxometry and correlation studies are based on regularized inverse Laplace transforms (ILTs). Here we introduce a general framework for characterizing microstructure that does not depend on diffusion modeling and replaces ill-posed ILTs with blind source separation (BSS). This framework yields proton density, relaxation times, volume fractions, and signal disentanglement, allowing for separation of the free-water component. THEORY AND METHODS: Diffusion experiments repeated for several different echo times, contain entangled diffusion and relaxation compartmental information. These can be disentangled by BSS using a physically constrained nonnegative matrix factorization. RESULTS: Computer simulations, phantom studies, together with repeatability and reproducibility experiments demonstrated that BSS is capable of estimating proton density, compartmental volume fractions and transversal relaxations. In vivo results proved its potential to correct for free-water contamination and to estimate tissue parameters. CONCLUSION: Formulation of the diffusion-relaxation dependence as a BSS problem introduces a new framework for studying microstructure compartmentalization, and a novel tool for free-water elimination.

Real valued diffusion-weighted imaging using decorrelated phase filtering.

PURPOSE: Because of the intrinsic low signal-to-noise ratio in diffusion-weighted imaging (DWI), magnitude processing often causes an overestimation of the signal's amplitude. This results in low-estimation accuracy of diffusion models and reduced contrast because of a superposition of the image signal and the noise floor. We adopt a new phase correction (PC) technique that yields real valued diffusion data while maintaining a Gaussian noise distribution. METHODS: We conduct simulations of the noise propagation in the echo-planar imaging reconstruction chain to determine the spatial noise correlation in the image. Using the correlation pattern, optimized filter kernels are derived to estimate the true phase of the signal in each voxel. Furthermore, we adopt an outlier detection technique to replace the real value by the magnitude in case of substantial signal loss resulting from incorrect PC. RESULTS: The benefits of our method are demonstrated on Monte Carlo simulations, DWI data acquired from healthy volunteer experiments, estimated parameters of the diffusion kurtosis imaging model, and the model-free diffusion spectrum imaging technique. The improved PC approach significantly reduces the noise bias and only slightly increases the sensitivity to local phase variations. CONCLUSION: PC can enhance the usefulness of higher b-values, allowing deeper insights into tissue microstructure. Magn Reson Med 77:559-570, 2017. © 2016 International Society for Magnetic Resonance in Medicine.

Model-based denoising in diffusion-weighted imaging using generalized spherical deconvolution.

PURPOSE: Diffusion MRI often suffers from low signal-to-noise ratio, especially for high b-values. This work proposes a model-based denoising technique to address this limitation. METHODS: A generalization of the multi-shell spherical deconvolution model using a Richardson-Lucy algorithm is applied to noisy data. The reconstructed coefficients are then used in the forward model to compute denoised diffusion-weighted images (DWIs). The proposed method operates in the diffusion space and thus is complementary to image-based denoising methods. RESULTS: We demonstrate improved image quality on the DWIs themselves, maps of neurite orientation dispersion and density imaging, and diffusional kurtosis imaging (DKI), as well as reduced spurious peaks in deterministic tractography. For DKI in particular, we observe up to 50% error reduction and demonstrate high image quality using just 30 DWIs. This corresponds to greater than fourfold reduction in scan time if compared to the widely used 140-DWI acquisitions. We also confirm consistent performance in pathological data sets, namely in white matter lesions of a multiple sclerosis patient. CONCLUSION: The proposed denoising technique termed generalized spherical deconvolution has the potential of significantly improving image quality in diffusion MRI. Magn Reson Med 78:2428-2438, 2017. © 2017 International Society for Magnetic Resonance in Medicine.

Distortion correction in diffusion-weighted imaging of the breast: Performance assessment of prospective, retrospective, and combined (prospective + retrospective) approaches.

PURPOSE: To compare the effectiveness of prospective, retrospective, and combined (prospective + retrospective) EPI distortion correction methods in bilateral breast diffusion-weighted imaging (DWI) scans. METHODS: Five healthy female subjects underwent an axial bilateral breast DWI exam with and without prospective B0 inhomogeneity correction using slice-by-slice linear shimming. In each case, an additional b=0 DWI scan was performed with the polarity of the phase-encoding gradient reversed, to generate an estimated B0 map; this map or a separately acquired B0 map was used for retrospective correction, either alone or in combination with the prospective correction. The alignment between an undistorted, anatomical reference scan with similar contrast and the corrected b=0 DWI images with different correction schemes was assessed. RESULTS: The average cross-correlation coefficient between the DWI images and the anatomical reference scan was increased from 0.82 to 0.92 over the five volunteers when combined prospective and retrospective distortion correction was applied. Furthermore, such correction substantially reduced patient-to-patient variation of the image alignment and the variability of the average apparent diffusion coefficient in normal glandular tissue. CONCLUSION: Combined prospective and retrospective distortion correction can provide an efficient way to reduce susceptibility-induced image distortions and enhance the reliability of breast DWI exams. Magn Reson Med 78:247-253, 2017. © 2016 International Society for Magnetic Resonance in Medicine.

Diffusion in realistic biophysical systems can lead to aliasing effects in diffusion spectrum imaging.

PURPOSE: Diffusion spectrum imaging (DSI) is an imaging technique that has been successfully applied to resolve white matter crossings in the human brain. However, its accuracy in complex microstructure environments has not been well characterized. THEORY AND METHODS: Here we have simulated different tissue configurations, sampling schemes, and processing steps to evaluate DSI performances' under realistic biophysical conditions. A novel approach to compute the orientation distribution function (ODF) has also been developed to include biophysical constraints, namely integration ranges compatible with axial fiber diffusivities. RESULTS: Performed simulations identified several DSI configurations that consistently show aliasing artifacts caused by fast diffusion components for both isotropic diffusion and fiber configurations. The proposed method for ODF computation showed some improvement in reducing such artifacts and improving the ability to resolve crossings, while keeping the quantitative nature of the ODF. CONCLUSION: In this study, we identified an important limitation of current DSI implementations, specifically the presence of aliasing due to fast diffusion components like those from pathological tissues, which are not well characterized, and can lead to artifactual fiber reconstructions. To minimize this issue, a new way of computing the ODF was introduced, which removes most of these artifacts and offers improved angular resolution. Magn Reson Med 76:1837-1847, 2016. © 2015 The Authors Magnetic Resonance in Medicine published by Wiley Periodicals, Inc. on behalf of International Society for Magnetic Resonance in Medicine. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Bias and precision analysis of diffusional kurtosis imaging for different acquisition schemes.

PURPOSE: Diffusional kurtosis imaging (DKI) is an approach to characterizing the non-Gaussian fraction of water diffusion in biological tissue. However, DKI is highly susceptible to the low signal-to-noise ratio of diffusion-weighted images, causing low precision and a significant bias due to Rician noise distribution. Here, we evaluate precision and bias using weighted linear least squares fitting of different acquisition schemes including several multishell schemes, a diffusion spectrum imaging (DSI) scheme, as well as a compressed sensing reconstruction of undersampled DSI scheme. METHODS: Monte Carlo simulations were performed to study the three-dimensional distribution of the apparent kurtosis coefficient (AKC). Experimental data were acquired from one healthy volunteer with multiple repetitions, using the same acquisition schemes as for the simulations. RESULTS: The angular distribution of the bias and precision were very inhomogeneous. While axial kurtosis was significantly overestimated, radial kurtosis was underestimated. The precision of radial kurtosis was up to 10-fold lower than axial kurtosis. CONCLUSION: The noise bias behavior of DKI is highly complex and can cause overestimation as well as underestimation of the AKC even within one voxel. The acquisition scheme with three shells, suggested by Poot et al, provided overall the best performance. Magn Reson Med 76:1684-1696, 2016. © 2016 International Society for Magnetic Resonance in Medicine.

Multi-directional anisotropy from diffusion orientation distribution functions.

PURPOSE: To evaluate a model-independent, multi-directional anisotropy (MDA) metric that is analytically and experimentally equivalent to fractional anisotropy (FA) in single-direction diffusivity, but potentially superior to FA in its sensitivity to the underlying anisotropy of multi-directional diffusivity. MATERIALS AND METHODS: An expression for MDA was defined from the orientation distribution function (ODF) and its analytical relation to FA was derived. Simulations of single and crossed double-fibers were performed using a compressed-sensing-accelerated diffusion-spectrum-imaging (CS-DSI) scheme. In vivo brain imaging using CS-DSI was performed on eight healthy subjects. MDA was compared with FA and with another ODF-based metric known as generalized FA (GFA). RESULTS: In simulated single-direction fibers, MDA was shown to be equivalent to FA (from FA = 0.2 to 0.8). In crossed fibers, MDA provided superior differentiation of the underlying anisotropy as compared to FA and GFA. In vivo analysis shows that the MDA was superior to both FA (P = 0.015) and GFA (P = 0.021) in terms of its relative accuracy in crossed fiber regions. CONCLUSION: MDA provides a potentially superior measure of fiber anisotropy relative to conventional FA or GFA, and may be used to improve the assessment of disease in regions with multi-directional brain fibers.

A Fourier-domain algorithm for total-variation regularized phase retrieval in differential X-ray phase contrast imaging.

Phase retrieval in differential X-ray phase contrast imaging involves a one dimensional integration step. In the presence of noise, standard integration methods result in image blurring and streak artifacts. This work proposes a regularized integration method which takes the availability of two dimensional data as well as the integration-specific frequency-dependent noise amplification into account. In more detail, a Fourier-domain algorithm is developed comprising a frequency-dependent minimization of the total variation orthogonal to the direction of integration. For both simulated and experimental data, the novel method yielded strong artefact reduction without increased blurring superior to the results obtained by standard integration methods or regularization techniques in the image domain.

Anatomically aware dual-hop learning for pulmonary embolism detection in CT pulmonary angiograms.

Pulmonary Embolisms (PE) represent a leading cause of cardiovascular death. While medical imaging, through computed tomographic pulmonary angiography (CTPA), represents the gold standard for PE diagnosis, it is still susceptible to misdiagnosis or significant diagnosis delays, which may be fatal for critical cases. Despite the recently demonstrated power of deep learning to bring a significant boost in performance in a wide range of medical imaging tasks, there are still very few published researches on automatic pulmonary embolism detection. Herein we introduce a deep learning based approach, which efficiently combines computer vision and deep neural networks for pulmonary embolism detection in CTPA. Our method brings novel contributions along three orthogonal axes: (1) automatic detection of anatomical structures; (2) anatomical aware pretraining, and (3) a dual-hop deep neural net for PE detection. We obtain state-of-the-art results on the publicly available multicenter large-scale RSNA dataset.

Accuracy of a deep learning-based algorithm for the detection of thoracic aortic calcifications in chest computed tomography and cardiovascular surgery planning.

OBJECTIVES: To assess the accuracy of a deep learning-based algorithm for fully automated detection of thoracic aortic calcifications in chest computed tomography (CT) with a focus on the aortic clamping zone. METHODS: We retrospectively included 100 chest CT scans from 91 patients who were examined on second- or third-generation dual-source scanners. Subsamples comprised 47 scans with an electrocardiogram-gated aortic angiography and 53 unenhanced scans. A deep learning model performed aortic landmark detection and aorta segmentation to derive 8 vessel segments. Associated calcifications were detected and their volumes measured using a mean-based density thresholding. Algorithm parameters (calcium cluster size threshold, aortic mask dilatation) were varied to determine optimal performance for the upper ascending aorta that encompasses the aortic clamping zone. A binary visual rating served as a reference. Standard estimates of diagnostic accuracy and inter-rater agreement using Cohen's Kappa were calculated. RESULTS: Thoracic aortic calcifications were observed in 74% of patients with a prevalence of 27-70% by aorta segment. Using different parameter combinations, the algorithm provided binary ratings for all scans and segments. The best performing parameter combination for the presence of calcifications in the aortic clamping zone yielded a sensitivity of 93% and a specificity of 82%, with an area under the receiver operating characteristic curve of 0.874. Using these parameters, the inter-rater agreement ranged from κ 0.66 to 0.92 per segment. CONCLUSIONS: Fully automated segmental detection of thoracic aortic calcifications in chest CT performs with high accuracy. This includes the critical preoperative assessment of the aortic clamping zone.

Saturation-recovery metabolic-exchange rate imaging with hyperpolarized [1-13C] pyruvate using spectral-spatial excitation.

Within the last decade hyperpolarized [1-13C] pyruvate chemical-shift imaging has demonstrated impressive potential for metabolic MR imaging for a wide range of applications in oncology, cardiology, and neurology. In this work, a highly efficient pulse sequence is described for time-resolved, multislice chemical shift imaging of the injected substrate and obtained downstream metabolites. Using spectral-spatial excitation in combination with single-shot spiral data acquisition, the overall encoding is evenly distributed between excitation and signal reception, allowing the encoding of one full two-dimensional metabolite image per excitation. The signal-to-noise ratio can be flexibly adjusted and optimized using lower flip angles for the pyruvate substrate and larger ones for the downstream metabolites. Selectively adjusting the excitation of the down-stream metabolites to 90° leads to a so-called "saturation-recovery" scheme with the detected signal content being determined by forward conversion of the available pyruvate. In case of repetitive excitations, the polarization is preserved using smaller flip angles for pyruvate. Metabolic exchange rates are determined spatially resolved from the metabolite images using a simplified two-site exchange model. This novel contrast is an important step toward more quantitative metabolic imaging. Goal of this work was to derive, analyze, and implement this "saturation-recovery metabolic exchange rate imaging" and demonstrate its capabilities in four rats bearing subcutaneous tumors.

Impact of retraining a deep learning algorithm for improving guideline-compliant aortic diameter measurements on non-gated chest CT.

PURPOSE/OBJECTIVE: Reliable detection of thoracic aortic dilatation (TAD) is mandatory in clinical routine. For ECG-gated CT angiography, automated deep learning (DL) algorithms are established for diameter measurements according to current guidelines. For non-ECG gated CT (contrast enhanced (CE) and non-CE), however, only a few reports are available. In these reports, classification as TAD is frequently unreliable with variable result quality depending on anatomic location with the aortic root presenting with the worst results. Therefore, this study aimed to explore the impact of re-training on a previously evaluated DL tool for aortic measurements in a cohort of non-ECG gated exams. METHODS & MATERIALS: A cohort of 995 patients (68 ± 12 years) with CE (n = 392) and non-CE (n = 603) chest CT exams was selected which were classified as TAD by the initial DL tool. The re-trained version featured improved robustness of centerline fitting and cross-sectional plane placement. All cases were processed by the re-trained DL tool version. DL results were evaluated by a radiologist regarding plane placement and diameter measurements. Measurements were classified as correctly measured diameters at each location whereas false measurements consisted of over-/under-estimation of diameters. RESULTS: We evaluated 8948 measurements in 995 exams. The re-trained version performed 8539/8948 (95.5%) of diameter measurements correctly. 3765/8948 (42.1%) of measurements were correct in both versions, initial and re-trained DL tool (best: distal arch 655/995 (66%), worst: Aortic sinus (AS) 221/995 (22%)). In contrast, 4456/8948 (49.8%) measurements were correctly measured only by the re-trained version, in particular at the aortic root (AS: 564/995 (57%), sinotubular junction: 697/995 (70%)). In addition, the re-trained version performed 318 (3.6%) measurements which were not available previously. A total of 228 (2.5%) cases showed false measurements because of tilted planes and 181 (2.0%) over-/under-segmentations with a focus at AS (n = 137 (14%) and n = 73 (7%), respectively). CONCLUSION: Re-training of the DL tool improved diameter assessment, resulting in a total of 95.5% correct measurements. Our data suggests that the re-trained DL tool can be applied even in non-ECG-gated chest CT including both, CE and non-CE exams.

Automated quantification of airway wall thickness on chest CT using retina U-Nets - Performance evaluation and application to a large cohort of chest CTs of COPD patients.

PURPOSE: Airway wall thickening is a consequence of chronic inflammatory processes and usually only qualitatively described in CT radiology reports. The purpose of this study is to automatically quantify airway wall thickness in multiple airway generations and assess the diagnostic potential of this parameter in a large cohort of patients with Chronic Obstructive Pulmonary Disease (COPD). MATERIALS AND METHODS: This retrospective, single-center study included a series of unenhanced chest CTs. Inclusion criteria were the mentioning of an explicit COPD GOLD stage in the written radiology report and time period (01/2019-12/2021). A control group included chest CTs with completely unremarkable lungs according to the report. The DICOM images of all cases (axial orientation; slice-thickness: 1 mm; soft-tissue kernel) were processed by an AI algorithm pipeline consisting of (A) a 3D-U-Net for det detection and tracing of the bronchial tree centerlines (B) extraction of image patches perpendicular to the centerlines of the bronchi, and (C) a 2D U-Net for segmentation of airway walls on those patches. The performance of centerline detection and wall segmentation was assessed. The imaging parameter average wall thickness was calculated for bronchus generations 3-8 (AWT3-8) across the lungs. Mean AWT3-8 was compared between five groups (control, COPD Gold I-IV) using non-parametric statistics. Furthermore, the established emphysema score %LAV-950 was calculated and used to classify scans (normal vs. COPD) alone and in combination with AWT3-8. RESULTS: A total of 575 chest CTs were processed. Algorithm performance was very good (airway centerline detection sensitivity: 86.9%; airway wall segmentation Dice score: 0.86). AWT3-8 was statistically significantly greater in COPD patients compared to controls (2.03 vs. 1.87 mm, p < 0.001) and increased with COPD stage. The classifier that combined %LAV-950 and AWT3-8 was superior to the classifier using only %LAV-950 (AUC = 0.92 vs. 0.79). CONCLUSION: Airway wall thickness increases in patients suffering from COPD and is automatically quantifiable. AWT3-8 could become a CT imaging parameter in COPD complementing the established emphysema biomarker %LAV-950. CLINICAL RELEVANCE STATEMENT: Quantitative measurements considering the complete visible bronchial tree instead of qualitative description could enhance radiology reports, allow for precise monitoring of disease progression and diagnosis of early stages of disease.

Diagnostic Accuracy and Performance of Artificial Intelligence in Detecting Lung Nodules in Patients With Complex Lung Disease: A Noninferiority Study.

OBJECTIVES: The aim of the study is to investigate the performance of artificial intelligence (AI) convolutional neural networks (CNN) in detecting lung nodules on chest computed tomography of patients with complex lung disease, and demonstrate its noninferiority when compared against an experienced radiologist through clinically relevant assessments. METHODS: A CNN prototype was used to retrospectively evaluate 103 complex lung disease cases and 40 control cases without reported nodules. Computed tomography scans were blindly evaluated by an expert thoracic radiologist; a month after initial analyses, 20 positive cases were re-evaluated with the assistance of AI. For clinically relevant applications: (1) AI was asked to classify each patient into nodules present or absent and (2) AI results were compared against standard radiology reports. Standard statistics were performed to determine detection performance. RESULTS: AI was, on average, 27 seconds faster than the expert and detected 8.4% of nodules that would have been missed. AI had a sensitivity of 67.7%, similar to an accuracy reported for experienced radiologists. AI correctly classified each patient (nodules present/absent) with a sensitivity of 96.1%. When matched against radiology reports, AI performed with a sensitivity of 89.4%. Control group assessment demonstrated an overall specificity of 82.5%. When aided by AI, the expert decreased the average assessment time per case from 2:44 minutes to 35.7 seconds, while reporting an overall increase in confidence. CONCLUSION: In a group of patients with complex lung disease, the sensitivity of AI is similar to an experienced radiologist and the tool helps detect previously missed nodules. AI also helps experts analyze for lung nodules faster and more confidently, a feature that is beneficial to patients and favorable to hospitals due to increased patient load and need for shorter turnaround times.

Performance of a deep learning tool to detect missed aortic dilatation in a large chest CT cohort.

Purpose: Thoracic aortic (TA) dilatation (TAD) is a risk factor for acute aortic syndrome and must therefore be reported in every CT report. However, the complex anatomy of the thoracic aorta impedes TAD detection. We investigated the performance of a deep learning (DL) prototype as a secondary reading tool built to measure TA diameters in a large-scale cohort. Material and methods: Consecutive contrast-enhanced (CE) and non-CE chest CT exams with "normal" TA diameters according to their radiology reports were included. The DL-prototype (AIRad, Siemens Healthineers, Germany) measured the TA at nine locations according to AHA guidelines. Dilatation was defined as >45 mm at aortic sinus, sinotubular junction (STJ), ascending aorta (AA) and proximal arch and >40 mm from mid arch to abdominal aorta. A cardiovascular radiologist reviewed all cases with TAD according to AIRad. Multivariable logistic regression (MLR) was used to identify factors (demographics and scan parameters) associated with TAD classification by AIRad. Results: 18,243 CT scans (45.7% female) were successfully analyzed by AIRad. Mean age was 62.3 ± 15.9 years and 12,092 (66.3%) were CE scans. AIRad confirmed normal diameters in 17,239 exams (94.5%) and reported TAD in 1,004/18,243 exams (5.5%). Review confirmed TAD classification in 452/1,004 exams (45.0%, 2.5% total), 552 cases were false-positive but identification was easily possible using visual outputs by AIRad. MLR revealed that the following factors were significantly associated with correct TAD classification by AIRad: TAD reported at AA [odds ratio (OR): 1.12, p < 0.001] and STJ (OR: 1.09, p = 0.002), TAD found at >1 location (OR: 1.42, p = 0.008), in CE exams (OR: 2.1-3.1, p < 0.05), men (OR: 2.4, p = 0.003) and patients presenting with higher BMI (OR: 1.05, p = 0.01). Overall, 17,691/18,243 (97.0%) exams were correctly classified. Conclusions: AIRad correctly assessed the presence or absence of TAD in 17,691 exams (97%), including 452 cases with previously missed TAD independent from contrast protocol. These findings suggest its usefulness as a secondary reading tool by improving report quality and efficiency.

Accelerated diffusion spectrum imaging in the human brain using compressed sensing.

We developed a novel method to accelerate diffusion spectrum imaging using compressed sensing. The method can be applied to either reduce acquisition time of diffusion spectrum imaging acquisition without losing critical information or to improve the resolution in diffusion space without increasing scan time. Unlike parallel imaging, compressed sensing can be applied to reconstruct a sub-Nyquist sampled dataset in domains other than the spatial one. Simulations of fiber crossings in 2D and 3D were performed to systematically evaluate the effect of compressed sensing reconstruction with different types of undersampling patterns (random, gaussian, Poisson disk) and different acceleration factors on radial and axial diffusion information. Experiments in brains of healthy volunteers were performed, where diffusion space was undersampled with different sampling patterns and reconstructed using compressed sensing. Essential information on diffusion properties, such as orientation distribution function, diffusion coefficient, and kurtosis is preserved up to an acceleration factor of R = 4.

Reduction of missed thoracic findings in emergency whole-body computed tomography using artificial intelligence assistance.

BACKGROUND: Radiology reporting of emergency whole-body computed tomography (CT) scans is time-critical and therefore involves a significant risk of pathology under-detection. We hypothesize a relevant number of initially missed secondary thoracic findings that would have been detected by an artificial intelligence (AI) software platform including several pathology-specific AI algorithms. METHODS: This retrospective proof-of-concept-study consecutively included 105 shock-room whole-body CT scans. Image data was analyzed by platform-bundled AI-algorithms, findings were reviewed by radiology experts and compared with the original radiologist's reports. We focused on secondary thoracic findings, such as cardiomegaly, coronary artery plaques, lung lesions, aortic aneurysms and vertebral fractures. RESULTS: We identified a relevant number of initially missed findings, with their quantification based on 105 analyzed CT scans as follows: up to 25 patients (23.8%) with cardiomegaly or borderline heart size, 17 patients (16.2%) with coronary plaques, 34 patients (32.4%) with aortic ectasia, 2 patients (1.9%) with lung lesions classified as "recommended to control" and 13 initially missed vertebral fractures (two with an acute traumatic origin). A high number of false positive or non-relevant AI-based findings remain problematic especially regarding lung lesions and vertebral fractures. CONCLUSIONS: We consider AI to be a promising approach to reduce the number of missed findings in clinical settings with a necessary time-critical radiological reporting. Nevertheless, algorithm improvement is necessary focusing on a reduction of "false positive" findings and on algorithm features assessing the finding relevance, e.g., fracture age or lung lesion malignancy.