Three artificial intelligence data challenges based on CT and ultrasound.

PURPOSE: The 2020 edition of these Data Challenges was organized by the French Society of Radiology (SFR), from September 28 to September 30, 2020. The goals were to propose innovative artificial intelligence solutions for the current relevant problems in radiology and to build a large database of multimodal medical images of ultrasound and computed tomography (CT) on these subjects from several French radiology centers. MATERIALS AND METHODS: This year the attempt was to create data challenge objectives in line with the clinical routine of radiologists, with less preprocessing of data and annotation, leaving a large part of the preprocessing task to the participating teams. The objectives were proposed by the different organizations depending on their core areas of expertise. A dedicated platform was used to upload the medical image data, to automatically anonymize the uploaded data. RESULTS: Three challenges were proposed including classification of benign or malignant breast nodules on ultrasound examinations, detection and contouring of pathological neck lymph nodes from cervical CT examinations and classification of calcium score on coronary calcifications from thoracic CT examinations. A total of 2076 medical examinations were included in the database for the three challenges, in three months, by 18 different centers, of which 12% were excluded. The 39 participants were divided into six multidisciplinary teams among which the coronary calcification score challenge was solved with a concordance index > 95%, and the other two with scores of 67% (breast nodule classification) and 63% (neck lymph node calcifications).

Automatic coronary artery calcium scoring from unenhanced-ECG-gated CT using deep learning.

PURPOSE: The purpose of this study was to develop and evaluate an algorithm that can automatically estimate the amount of coronary artery calcium (CAC) from unenhanced electrocardiography (ECG)-gated computed tomography (CT) cardiac volume acquisitions by using convolutional neural networks (CNN). MATERIALS AND METHODS: The method used a set of five CNN with three-dimensional (3D) U-Net architecture trained on a database of 783 CT examinations to detect and segment coronary artery calcifications in a 3D volume. The Agatston score, the conventional CAC scoring, was then computed slice by slice from the resulting segmentation mask and compared to the ground truth manually estimated by radiologists. The quality of the estimation was assessed with the concordance index (C-index) on CAC risk category on a separate testing set of 98 independent CT examinations. RESULTS: The final model yielded a C-index of 0.951 on the testing set. The remaining errors of the method were mainly observed on small-size and/or low-density calcifications, or calcifications located near the mitral valve or ring. CONCLUSION: The deep learning-based method proposed here to compute automatically the CAC score from unenhanced-ECG-gated cardiac CT is fast, robust and yields accuracy similar to those of other artificial intelligence methods, which could improve workflow efficiency, eliminating the time spent on manually selecting coronary calcifications to compute the Agatston score.

Usefulness of a single-parameter tool for the prediction of large vessel occlusion in acute stroke.

BACKGROUND: In acute stroke, large vessel occlusion (LVO) should be promptly identified to guide patient's transportation directly to comprehensive stroke centers (CSC) for mechanical thrombectomy (MT). In many cases, prehospital multi-parameter scores are used by trained emergency teams to identify patients with high probability of LVO. However, in several countries, the first aid organization without intervention of skilled staff precludes the on-site use of such scores. Here, we assessed the accuracy of LVO prediction using a single parameter (i.e. complete hemiplegia) obtained by bystander's telephone-based witnessing. PATIENTS AND METHODS: This observational, single-center study included consecutive patients who underwent intravenous thrombolysis at the primary stroke center and/or were directly transferred to a CSC for MT, from January 1, 2015 to March 1, 2020. We defined two groups: patients with initial hemiplegia (no movement in one arm and leg and facial palsy) and patients without initial hemiplegia, on the basis of a bystander's witnessing. RESULTS: During the study time, 874 patients were included [mean age 73 years (SD 13.8), 56.7% men], 320 with initial hemiplegia and 554 without. The specificity of the hemiplegia criterion to predict LVO was 0.88, but its sensitivity was only 0.53. CONCLUSION: Our results suggest that the presence of hemiplegia as witnessed by a bystander can predict LVO with high specificity. This single criterion could be used for decision-making about direct transfer to CSC for MT when the absence of emergency skilled staff precludes the patient's on-site assessment, especially in regions distant from a CSC.

Mechanical Recanalization after Transfer from a Distant Primary Stroke Center: Effectiveness and Future Directions.

INTRODUCTION: Little is known about the effectiveness of endovascular treatment (EVT) in patients with acute ischemic stroke (AIS) due to large vessel occlusion (LVO) admitted to a primary stroke center (PSC). The aim of this study was to assess EVT effectiveness after transfer from a PSC to a distant (156 km apart; 1.5 hour by car) comprehensive stroke center (CSC), and to discuss perspectives to improve access to EVT, if indicated. PATIENTS AND METHOD: Analysis of the data collected in a 6-year prospective registry of patients admitted to a PSC for AIS due to LVO and selected for transfer to a distant CSC for EVT. The rate of transfer, futile transfer, EVT, reperfusion (thrombolysis in cerebral infarction score ≥2b-3), and relevant time measures were determined. RESULTS: Among the 529 patients eligible, 278 (52.6%) were transferred and 153 received EVT (55% of transferred patients) followed by reperfusion in 115 (overall reperfusion rate: 21.7%). Median times (interquartile range) were: 90 minutes (76-110) for PSC-door-in to PSC-door-out, 88 minutes (65-104) for PSC-door-out to CSC-door-in, 262 minutes (239-316) for PSC-imaging to reperfusion, and 393 minutes (332-454) for symptom onset to reperfusion. At 3 months, rates of favorable outcome (modified Rankin Scale 0-2) were not significantly different between patients eligible for EVT (42.4%), transferred patients (49.1%) and patients who underwent EVT (34.1%). DISCUSSION AND CONCLUSIONS: Our study suggests that transfer to a distant CSC is associated with reduced access to early EVT. These results argue in favor of on-site EVT at high volume PSCs that are distant from the CSC.

Target Door-to-Needle Time for Tissue Plasminogen Activator Treatment with Magnetic Resonance Imaging Screening Can Be Reduced to 45 min.

OBJECTIVE: The purpose of this study was to demonstrate that the median door-to-needle (DTN) time for intravenous tissue plasminogen activator (tPA) treatment can be reduced to 45 min in a primary stroke centre with MRI-based screening for acute ischaemic stroke (AIS). METHODS: From February 2015 to February 2017, the stroke unit of Perpignan general hospital, France, implemented a quality-improvement (QI) process. During this period, patients who received tPA within 4.5 h after AIS onset were included in the QI cohort. Their clinical characteristics and timing metrics were compared each semester and also with those of 135 consecutive patients with AIS treated by tPA during the 1-year pre-QI period (pre-QI cohort). RESULTS: In the QI cohort, 274 patients (92.5%) underwent MRI screening. While the demographic and baseline characteristics were not significantly different between cohorts, the median DTN time was significantly lower in the QI than in the pre-QI cohort (52 vs. 84 min; p < 0.00001). Within the QI cohort, the median DTN time for each semester decreased from 65 to 44 min (p < 0.00001) and the proportion of treated patients with a DTN time ≤45 min increased from 25 to 58.9% (p < 0.0001). Overall, DTN time improvement was associated with a better outcome at 3 months (patients with a modified Rankin Scale score between 0 and 2: 61.8% in the QI vs. 39.3% in the pre-QI cohort; p < 0.0001). CONCLUSIONS: A QI process can reduce the DTN within 45 min with MRI as a screening tool.