Delineation of the left ventricular cavity, myocardium, and right ventricle from cardiac magnetic resonance images (multi-slice 2-D cine MRI) is a common clinical task to establish diagnosis. The automation of the corresponding tasks has thus been the subject of intense research over the past decades. In this paper, we introduce the "Automatic Cardiac Diagnosis Challenge" dataset (ACDC), the largest publicly available and fully annotated dataset for the purpose of cardiac MRI (CMR) assessment. The dataset contains data from 150 multi-equipments CMRI recordings with reference measurements and classification from two medical experts. The overarching objective of this paper is to measure how far state-of-the-art deep learning methods can go at assessing CMRI, i.e., segmenting the myocardium and the two ventricles as well as classifying pathologies. In the wake of the 2017 MICCAI-ACDC challenge, we report results from deep learning methods provided by nine research groups for the segmentation task and four groups for the classification task. Results show that the best methods faithfully reproduce the expert analysis, leading to a mean value of 0.97 correlation score for the automatic extraction of clinical indices and an accuracy of 0.96 for automatic diagnosis. These results clearly open the door to highly accurate and fully automatic analysis of cardiac CMRI. We also identify scenarios for which deep learning methods are still failing. Both the dataset and detailed results are publicly available online, while the platform will remain open for new submissions.

In the above article, incorrect data was inadvertently included in the ''INH'' and ''RIF'' columns of Table S5. The corrected table is now available online.

In newly diagnosed diffuse large B-cell lymphoma (DLBCL), the sensitivity of bone marrow biopsy (BMB) for the detection of bone marrow involvement (BMI) can be low because of sampling error if the BMI is focal and not diffuse. Although 18 F-FDG PET/CT is now recommended for initial staging of DLBCL, its role regarding BMI is not well defined. This study evaluated whether 18 F-FDG PET/CT, in comparison with BMB, is useful for the detection of BMI and predictive of outcome. Methods: From the 142 patients who were referred to our institution for newly diagnosed DLBCL from June 2006 to October 2011, 133 were retrospectively enrolled in our study. All patients underwent whole-body 18 F-FDG PET/CT and a BMB from the iliac crest before any treatment. 18 F-FDG PET/CT was considered positive for BMI in cases of uni-or multifocal bone marrow 18 F-FDG uptake that could not be explained by benign findings on the underlying CT image or history. A final diagnosis of BMI was considered if the BMB was positive or if the positive 18 F-FDG PET/CT was confirmed by guided biopsy or targeted MR imaging or in cases of disappearance of focal bone marrow uptake concomitant with the disappearance of uptake in other lymphoma lesions on 18 F-FDG PET/CT monitoring. Progression-free survival and overall survival were analyzed using the Cox proportional hazards regression model. Results: Thirty-three patients were considered to have BMI. Of these, 8 were positive according to the BMB and 32 were positive according to 18 F-FDG PET/CT. 18 F-FDG PET/CT was more sensitive (94% vs. 24%; P , 0.001), showed a higher negative predictive value (98% vs. 80%), and was more accurate (98% vs. 81%) than BMB. Median follow-up was 24 mo (range, 1-67 mo). Twentynine patients (22%) experienced recurrence or disease progression during follow-up, and 20 patients died (15%). In multivariate analysis, only the International Prognostic Index and the 18 F-FDG PET/CT bone marrow status were independent predictors of progressionfree survival (P 5 0.005 and 0.02, respectively), whereas only the International Prognostic Index remained an independent predictor of overall survival (P 5 0.004). Conclusion: Assessment of BMI with 18 F-FDG PET/CT provides better diagnostic performance and prognostic stratification in newly diagnosed DLBCL than does BMB.