Algorithmic transparency and interpretability measures improve radiologists' performance in BI-RADS 4 classification.

OBJECTIVE: To evaluate the perception of different types of AI-based assistance and the interaction of radiologists with the algorithm's predictions and certainty measures. METHODS: In this retrospective observer study, four radiologists were asked to classify Breast Imaging-Reporting and Data System 4 (BI-RADS4) lesions (n = 101 benign, n = 99 malignant). The effect of different types of AI-based assistance (occlusion-based interpretability map, classification, and certainty) on the radiologists' performance (sensitivity, specificity, questionnaire) were measured. The influence of the Big Five personality traits was analyzed using the Pearson correlation. RESULTS: Diagnostic accuracy was significantly improved by AI-based assistance (an increase of 2.8% ± 2.3%, 95 %-CI 1.5 to 4.0 %, p = 0.045) and trust in the algorithm was generated primarily by the certainty of the prediction (100% of participants). Different human-AI interactions were observed ranging from nearly no interaction to humanization of the algorithm. High scores in neuroticism were correlated with higher persuasibility (Pearson's r = 0.98, p = 0.02), while higher consciousness and change of accuracy showed an inverse correlation (Pearson's r = -0.96, p = 0.04). CONCLUSION: Trust in the algorithm's performance was mostly dependent on the certainty of the predictions in combination with a plausible heatmap. Human-AI interaction varied widely and was influenced by personality traits. KEY POINTS: • AI-based assistance significantly improved the diagnostic accuracy of radiologists in classifying BI-RADS 4 mammography lesions. • Trust in the algorithm's performance was mostly dependent on the certainty of the prediction in combination with a reasonable heatmap. • Personality traits seem to influence human-AI collaboration. Radiologists with specific personality traits were more likely to change their classification according to the algorithm's prediction than others.

Dark-field chest x-ray imaging: first experience in patients with alpha1-antitrypsin deficiency.

BACKGROUND: Spirometry and conventional chest x-ray have limitations in investigating early emphysema, while computed tomography, the reference imaging method in this context, is not part of routine patient care due to its higher radiation dose. In this work, we investigated a novel low-dose imaging modality, dark-field chest x-ray, for the evaluation of emphysema in patients with alpha1-antitrypsin deficiency. METHODS: By exploiting wave properties of x-rays for contrast formation, dark-field chest x-ray visualises the structural integrity of the alveoli, represented by a high signal over the lungs in the dark-field image. We investigated four patients with alpha1-antitrypsin deficiency with a novel dark-field x-ray prototype and simultaneous conventional chest x-ray. The extent of pulmonary function impairment was assessed by pulmonary function measurement and regional emphysema distribution was compared with CT in one patient. RESULTS: We show that dark-field chest x-ray visualises the extent of pulmonary emphysema displaying severity and regional differences. Areas with low dark-field signal correlate with emphysematous changes detected by computed tomography using a threshold of -950 Hounsfield units. The airway parameters obtained by whole-body plethysmography and single breath diffusing capacity of the lungs for carbon monoxide demonstrated typical changes of advanced emphysema. CONCLUSIONS: Dark-field chest x-ray directly visualised the severity and regional distribution of pulmonary emphysema compared to conventional chest x-ray in patients with alpha1-antitrypsin deficiency. Due to the ultra-low radiation dose in comparison to computed tomography, dark-field chest x-ray could be beneficial for long-term follow-up in these patients.

Lung nodule detection in chest X-rays using synthetic ground-truth data comparing CNN-based diagnosis to human performance.

We present a method to generate synthetic thorax radiographs with realistic nodules from CT scans, and a perfect ground truth knowledge. We evaluated the detection performance of nine radiologists and two convolutional neural networks in a reader study. Nodules were artificially inserted into the lung of a CT volume and synthetic radiographs were obtained by forward-projecting the volume. Hence, our framework allowed for a detailed evaluation of CAD systems' and radiologists' performance due to the availability of accurate ground-truth labels for nodules from synthetic data. Radiographs for network training (U-Net and RetinaNet) were generated from 855 CT scans of a public dataset. For the reader study, 201 radiographs were generated from 21 nodule-free CT scans with altering nodule positions, sizes and nodule counts of inserted nodules. Average true positive detections by nine radiologists were 248.8 nodules, 51.7 false positive predicted nodules and 121.2 false negative predicted nodules. The best performing CAD system achieved 268 true positives, 66 false positives and 102 false negatives. Corresponding weighted alternative free response operating characteristic figure-of-merits (wAFROC FOM) for the radiologists range from 0.54 to 0.87 compared to a value of 0.81 (CI 0.75-0.87) for the best performing CNN. The CNN did not perform significantly better against the combined average of the 9 readers (p = 0.49). Paramediastinal nodules accounted for most false positive and false negative detections by readers, which can be explained by the presence of more tissue in this area.

Calcium scoring using virtual non-contrast images from a dual-layer spectral detector CT: comparison to true non-contrast data and evaluation of proportionality factor in a large patient collective.

OBJECTIVE: Determination of coronary artery calcium scoring (CACS) in non-contrast computed tomography (CT) images has been shown to be an important prognostic factor in coronary artery disease (CAD). The objective of this study was to evaluate the accuracy of CACS from virtual non-contrast (VNC) imaging generated from spectral data in comparison to standard (true) non-contrast (TNC) imaging in a representative patient cohort with clinically approved software. METHODS: One hundred three patients referred to coronary CTA with suspicion of CAD were investigated on a dual-layer spectral detector CT (SDCT) scanner. CACS was calculated from both TNC and VNC images by software certified for medical use. Patients with a CACS of 0 were excluded from analysis. RESULTS: The mean age of the study population was 61 ± 11 years with 48 male patients (67%). Inter-quartile range of clinical CACS was 22-282. Correlation of measured CACS from true- and VNC images was high (0.95); p < 0.001. The slope was 3.83, indicating an underestimation of VNC CACS compared to TNC CACS by that factor. Visual analysis of the Bland-Altman plot of CACS showed good accordance with both methods after correction of VNC CACS by the abovementioned factor. CONCLUSIONS: In clinical diagnostics of CAD, the determination of CACS is feasible using VNC images generated from spectral data obtained on a dual-layer spectral detector CT. When multiplied by a correction factor, results were in good agreement with the standard technique. This could enable radiation dose reductions by obviating the need for native scans typically used for CACS. KEY POINTS: • Calcium scoring is feasible from contrast-enhanced CT images using a dual-layer spectral detector CT scanner. • When multiplied by a correction factor, calcium scoring from virtual non-contrast images shows good agreement with the standard technique. • Omitting native scans for calcium scoring could enable radiation dose reduction.

Coronary calcium scoring assessed on native screening chest CT imaging as predictor for outcome in COVID-19: An analysis of a hospitalized German cohort.

BACKGROUND: Since the outbreak of the COVID-19 pandemic, a number of risk factors for a poor outcome have been identified. Thereby, cardiovascular comorbidity has a major impact on mortality. We investigated whether coronary calcification as a marker for coronary artery disease (CAD) is appropriate for risk prediction in COVID-19. METHODS: Hospitalized patients with COVID-19 (n = 109) were analyzed regarding clinical outcome after native computed tomography (CT) imaging for COVID-19 screening. CAC (coronary calcium score) and clinical outcome (need for intensive care treatment or death) data were calculated following a standardized protocol. We defined three endpoints: critical COVID-19 and transfer to ICU, fatal COVID-19 and death, composite endpoint critical and fatal COVID-19, a composite of ICU treatment and death. We evaluated the association of clinical outcome with the CAC. Patients were dichotomized by the median of CAC. Hazard ratios and odds ratios were calculated for the events death or ICU or a composite of death and ICU. RESULTS: We observed significantly more events for patients with CAC above the group's median of 31 for critical outcome (HR: 1.97[1.09,3.57], p = 0.026), for fatal outcome (HR: 4.95[1.07,22.9], p = 0.041) and the composite endpoint (HR: 2.31[1.28,4.17], p = 0.0056. Also, odds ratio was significantly increased for critical outcome (OR: 3.01 [1.37, 6.61], p = 0.01) and for fatal outcome (OR: 5.3 [1.09, 25.8], p = 0.02). CONCLUSION: The results indicate a significant association between CAC and clinical outcome in COVID-19. Our data therefore suggest that CAC might be useful in risk prediction in patients with COVID-19.

A robust convolutional neural network for lung nodule detection in the presence of foreign bodies.

Lung cancer is a major cause of death worldwide. As early detection can improve outcome, regular screening is of great interest, especially for certain risk groups. Besides low-dose computed tomography, chest X-ray is a potential option for screening. Convolutional network (CNN) based computer aided diagnosis systems have proven their ability of identifying nodules in radiographies and thus may assist radiologists in clinical practice. Based on segmented pulmonary nodules, we trained a CNN based one-stage detector (RetinaNet) with 257 annotated radiographs and 154 additional radiographs from a public dataset. We compared the performance of the convolutional network with the performance of two radiologists by conducting a reader study with 75 cases. Furthermore, the potential use for screening on patient level and the impact of foreign bodies with respect to false-positive detections was investigated. For nodule location detection, the architecture achieved a performance of 43 true-positives, 26 false-positives and 22 false-negatives. In comparison, performance of the two readers was 42 ± 2 true-positives, 28 ± 0 false-positives and 23 ± 2 false-negatives. For the screening task, we retrieved a ROC AUC value of 0.87 for the reader study test set. We found the trained RetinaNet architecture to be only slightly prone to foreign bodies in terms of misclassifications: out of 59 additional radiographs containing foreign bodies, false-positives in two radiographs were falsely detected due to foreign bodies.

Acceleration of chemical shift encoding-based water fat MRI for liver proton density fat fraction and T2* mapping using compressed sensing.

OBJECTIVES: To evaluate proton density fat fraction (PDFF) and T2* measurements of the liver with combined parallel imaging (sensitivity encoding, SENSE) and compressed sensing (CS) accelerated chemical shift encoding-based water-fat separation. METHODS: Six-echo Dixon imaging was performed in the liver of 89 subjects. The first acquisition variant used acceleration based on SENSE with a total acceleration factor equal to 2.64 (acquisition labeled as SENSE). The second acquisition variant used acceleration based on a combination of CS with SENSE with a total acceleration factor equal to 4 (acquisition labeled as CS+SENSE). Acquisition times were compared between acquisitions and proton density fat fraction (PDFF) and T2*-values were measured and compared separately for each liver segment. RESULTS: Total scan duration was 14.5 sec for the SENSE accelerated image acquisition and 9.3 sec for the CS+SENSE accelerated image acquisition. PDFF and T2* values did not differ significantly between the two acquisitions (paired Mann-Whitney and paired t-test P>0.05 in all cases). CS+SENSE accelerated acquisition showed reduced motion artifacts (1.1%) compared to SENSE acquisition (12.3%). CONCLUSION: CS+SENSE accelerates liver PDFF and T2*mapping while retaining the same quantitative values as an acquisition using only SENSE and reduces motion artifacts.

Volume versus standard coils in the treatment of intracranial aneurysms.

BACKGROUND: Volume coils were developed to improve occlusion rates of intracranial aneurysms. Previous studies have shown increased packing density and comparable occlusion rates, but subgroup analyses of aneurysm size have not been carried out. OBJECTIVE: To evaluate the safety and efficacy of the Penumbra Coil 400 (PC400) system in treating intracranial aneurysms compared with standard diameter coils. METHODS: A monocentric retrospective case review of 260 aneurysms in 233 patients was carried out. In 37 aneurysms the PC400 system was used, while 223 aneurysms were treated with conventional coils. Previously treated aneurysms and aneurysms treated with flow diverters were excluded. Aneurysm and procedure characteristics, packing density, postprocedural and follow-up occlusion grades as well as coil compaction were evaluated. RESULTS: Aneurysms treated with PC400 coils had higher volume (218.9 vs 47.1 mm(3), p<0.001), wider necks (3.0 vs 2.5 mm, p=0.005), and greater dome/neck ratio (2.0 vs 1.6, p=0.001) in comparison with aneurysms treated with conventional coils. Compared with controls, in the PC400 group we achieved higher packing densities (43.2% vs 34.4%, p<0.001; in aneurysms ≥7 mm 42.2% vs 27.8%, p<0.001). On follow-up angiography we observed less coil compaction (23.8% vs 64.3%, p=0.003) and less aneurysm recurrence (14.3% vs 40.5%, p=0.046) in aneurysms ≥7 mm when using the PC400 system. CONCLUSIONS: Use of the PC400 system as opposed to conventional coils suggests that the PC400 system is safe and effective in treating intracranial aneurysms. Despite having been applied in a potentially more difficult-to-treat group, the use of PC400 was associated with less coil compaction and aneurysm recurrence in aneurysms ≥7 mm.