Global-Local attention network with multi-task uncertainty loss for abnormal lymph node detection in MR images

•We propose a novel network for the universal abnormal lymph node detection in MR images, which has great clinical value for the diagnosis of numerous diseases.•We design a global-local context module to encode the image global and local scale context information for the detection and utilize the channel attention mechanism to weight different contexts.•We introduce a multi-task uncertainty loss to adaptively balance the losses of different tasks, which can effectively alleviate the burden for tuning the loss weights by hand.•We build a large-scale MRI abnormal lymph node dataset, which includes a total of 821 abnormal abdominal lymph nodes of 41 types from 584 different patients. Moreover, 123 images with complete 3D volume annotations are delineated by an experienced radiologist. [Display omitted] Accurate and reliable detection of abnormal lymph nodes in magnetic resonance (MR) images is very helpful for the diagnosis and treatment of numerous diseases. However, it is still a challenging task due to similar appearances between abnormal lymph nodes and other tissues. In this paper, we propose a novel network based on an improved Mask R-CNN framework for the detection of abnormal lymph nodes in MR images. Instead of laboriously collecting large-scale pixel-wise annotated training data, pseudo masks generated from RECIST bookmarks on hand are utilized as the supervision. Different from the standard Mask R-CNN architecture, there are two main innovations in our proposed network: 1) global-local attention which encodes the global and local scale context for detection and utilizes the channel attention mechanism to extract more discriminative features and 2) multi-task uncertainty loss which adaptively weights multiple objective loss functions based on the uncertainty of each task to automatically search the optimal solution. For the experiments, we built a new abnormal lymph node dataset with 821 RECIST bookmarks of 41 different types of abnormal abdominal lymph nodes from 584 different patients. The experimental results showed the superior performance of our algorithm over compared state-of-the-art approaches.