Artificial intelligence in BreastScreen Norway: a retrospective analysis of a cancer-enriched sample including 1254 breast cancer cases.

OBJECTIVES: To compare results of selected performance measures in mammographic screening for an artificial intelligence (AI) system versus independent double reading by radiologists. METHODS: In this retrospective study, we analyzed data from 949 screen-detected breast cancers, 305 interval cancers, and 13,646 negative examinations performed in BreastScreen Norway during the period from 2010 to 2018. An AI system scored the examinations from 1 to 10, based on the risk of malignancy. Results from the AI system were compared to screening results after independent double reading. AI score 10 was set as the threshold. The results were stratified by mammographic density. RESULTS: A total of 92.7% of the screen-detected and 40.0% of the interval cancers had an AI score of 10. Among women with a negative screening outcome, 9.1% had an AI score of 10. For women with the highest breast density, the AI system scored 100% of the screen-detected cancers and 48.6% of the interval cancers with an AI score of 10, which resulted in a sensitivity of 80.9% for women with the highest breast density for the AI system, compared to 62.8% for independent double reading. For women with screen-detected cancers who had prior mammograms available, 41.9% had an AI score of 10 at the prior screening round. CONCLUSIONS: The high proportion of cancers with an AI score of 10 indicates a promising performance of the AI system, particularly for women with dense breasts. Results on prior mammograms with AI score 10 illustrate the potential for earlier detection of breast cancers by using AI in screen-reading. KEY POINTS: • The AI system scored 93% of the screen-detected cancers and 40% of the interval cancers with AI score 10. • The AI system scored all screen-detected cancers and almost 50% of interval cancers among women with the highest breast density with AI score 10. • About 40% of the screen-detected cancers had an AI score of 10 on the prior mammograms, indicating a potential for earlier detection by using AI in screen-reading.

Estimating cardiac output from coronary CT angiography: an individualized compartment model in comparison to the Stewart-Hamilton method.

Background: Attenuation is correlated with the concentration of contrast medium (CM) in the arteries. The cardiac output (CO) affects the concentration of CM in the circulatory system; therefore, CO affects the time-density curve (TDC). Thus, estimating CO using TDC from test-bolus images acquired in computed tomography (CT) is possible. In this study, we compare two methods of estimating CO, namely, an individualized mathematical compartment model, integrating patient, contrast, and scanning factors with TDC, and the Stewart-Hamilton method based on the area under the curve of the TDC. Materials and methods: Attenuation in the aorta was measured during test-bolus in 40 consecutive patients with a clinical indication for coronary CT angiography (CCTA). Each participant underwent cardiac magnetic resonance imaging following CCTA to validate the estimated CO. The individual compartment model used TDC in conjunction with scanning and patient-specific parameters to estimate the concentration of CM and CO over time. This was compared to the CO calculated from the area under the curve using the Stewart-Hamilton method. Results: Both CO estimated with our individualized compartment model (r = 0.66, p < 0.01) and the Stewart-Hamilton method (r = 0.53, p < 0.01) were moderately correlated with CO measured with cardiac MRI. Body surface area (BSA) and time to peak (TTP) affected the accuracy of our model. Lower BSA resulted in overestimation, and lower TTP resulted in CO underestimation, respectively. We found no gender-specific difference in the accuracy of our model when correcting for BSA. The Stewart-Hamilton method performed better with a more complete TDC, whereas the compartment model performed better overall with a partial TDC. Conclusion: The TDC acquired in CCTA allows for CO estimation. Both the Stewart-Hamilton method and our mathematical compartment model show moderate correlation when applied to our data, although each method has its strengths and limitations. If the majority of the TDC is known, the Stewart-Hamilton method may be more reliable, but an individual compartment model is preferable when there are insufficient data points in the TDC. Regardless, both methods can potentially increase the diagnostic information acquired from a CCTA, which is increasingly recommended in clinical guidelines.

Metric based virtual simulation training for endovascular thrombectomy improves interventional neuroradiologists' simulator performance.

OBJECTIVE: Metric based virtual reality simulation training may enhance the capability of interventional neuroradiologists (INR) to perform endovascular thrombectomy. As pilot for a national simulation study we examined the feasibility and utility of simulated endovascular thrombectomy procedures on a virtual reality (VR) simulator. METHODS: Six INR and four residents participated in the thrombectomy skill training on a VR simulator (Mentice VIST 5G). Two different case-scenarios were defined as benchmark-cases, performed before and after VR simulator training. INR performing endovascular thrombectomy clinically were also asked to fill out a questionnaire analyzing their degree of expectation and general attitude towards VR simulator training. RESULTS: All participants improved in mean total procedure time for both benchmark-cases. Experts showed significant improvements in handling errors (case 2), a reduction in contrast volume used (case 1 and 2), and fluoroscopy time (case 1 and 2). Novices showed a significant improvement in steps finished (case 2), a reduction in fluoroscopy time (case 1), and radiation used (case 1). Both, before and after having performed simulation training the participating INR had a positive attitude towards VR simulation training. CONCLUSION: VR simulation training enhances the capability of INR to perform endovascular thrombectomy on the VR simulator. INR have generally a positive attitude towards VR simulation training. Whether the VR simulation training translates to enhanced clinical performance will be evaluated in the ongoing Norwegian national simulation study.

Machine Learning Algorithms Versus Thresholding to Segment Ischemic Regions in Patients With Acute Ischemic Stroke.

OBJECTIVE: Computed tomography (CT) scan is a fast and widely used modality for early assessment in patients with symptoms of a cerebral ischemic stroke. CT perfusion (CTP) is often added to the protocol and is used by radiologists for assessing the severity of the stroke. Standard parametric maps are calculated from the CTP datasets. Based on parametric value combinations, ischemic regions are separated into presumed infarct core (irreversibly damaged tissue) and penumbra (tissue-at-risk). Different thresholding approaches have been suggested to segment the parametric maps into these areas. The purpose of this study is to compare fully-automated methods based on machine learning and thresholding approaches to segment the hypoperfused regions in patients with ischemic stroke. METHODS: We test two different architectures with three mainstream machine learning algorithms. We use parametric maps as input features, and manual annotations made by two expert neuroradiologists as ground truth. RESULTS: The best results are produced with random forest (RF) and Single-Step approach; we achieve an average Dice coefficient of 0.68 and 0.26, respectively for penumbra and core, for the three groups analysed. We also achieve an average in volume difference of 25.1 ml for penumbra and 7.8 ml for core. CONCLUSIONS: Our best RF-based method outperforms the classical thresholding approaches, to segment both the ischemic regions in a group of patients regardless of the severity of vessel occlusion. SIGNIFICANCE: A correct visualization of the ischemic regions will guide treatment decisions better.