A New 2.5D Representation for Lymph Node Detection using Random Sets of Deep Convolutional Neural Network Observations
This work addresses a challenging clinical diagnostic problem for medical imaging by providing a more accurate automated detection system for lymph nodes, with incremental improvements over existing approaches.
The paper tackled automated lymph node detection in CT scans by proposing a 2.5D representation using random sets of deep CNN observations, achieving sensitivities of 70% at 3 false-positives per volume for mediastinal LNs and 83% at 3 FP/vol. for abdominal LNs, which significantly improves over prior state-of-the-art methods.
Automated Lymph Node (LN) detection is an important clinical diagnostic task but very challenging due to the low contrast of surrounding structures in Computed Tomography (CT) and to their varying sizes, poses, shapes and sparsely distributed locations. State-of-the-art studies show the performance range of 52.9% sensitivity at 3.1 false-positives per volume (FP/vol.), or 60.9% at 6.1 FP/vol. for mediastinal LN, by one-shot boosting on 3D HAAR features. In this paper, we first operate a preliminary candidate generation stage, towards 100% sensitivity at the cost of high FP levels (40 per patient), to harvest volumes of interest (VOI). Our 2.5D approach consequently decomposes any 3D VOI by resampling 2D reformatted orthogonal views N times, via scale, random translations, and rotations with respect to the VOI centroid coordinates. These random views are then used to train a deep Convolutional Neural Network (CNN) classifier. In testing, the CNN is employed to assign LN probabilities for all N random views that can be simply averaged (as a set) to compute the final classification probability per VOI. We validate the approach on two datasets: 90 CT volumes with 388 mediastinal LNs and 86 patients with 595 abdominal LNs. We achieve sensitivities of 70%/83% at 3 FP/vol. and 84%/90% at 6 FP/vol. in mediastinum and abdomen respectively, which drastically improves over the previous state-of-the-art work.