Towards Single-phase Single-stage Detection of Pulmonary Nodules in Chest CT Imaging

Abstract

Detection of pulmonary nodules in chest CT imaging plays a crucial role in early diagnosis of lung cancer. Manual examination is highly time-consuming and prone to errors, calling for computer-aided detection, both to improve detection efficiency and reduce misdiagnosis. Over the years, a large number of such systems have been proposed, which mostly followed a two-phase paradigm with: 1) candidate detection and 2) false positive reduction. Recently, deep learning has become a dominant force in algorithm development. As for candidate detection, prior state-of-the-art was mainly based on the two-stage Faster R-CNN framework, which starts with an initial sub-net to generate a set of classagnostic region proposals, followed by a second sub-net to perform classification and bounding-box regression. In contrast, we abandon the conventional two-phase paradigm and two-stage framework altogether, and propose to train a single network to achieve end-to-end nodule detection instead, without transfer learning or further post-processing. Our feature learning model is a modification of the ResNet and feature pyramid network combined, powered by RReLU activation. The major challenge is the condition of extreme inter-class and intra-class sample imbalance, where positive samples are overwhelmed by a vast negative pool largely composed of easily discriminative samples and a handful of hard samples.Direct training on all samples can seriously undermine training efficacy. We propose a patch-based sampling strategy over a set of regularly updating anchors, which is able to narrow sampling scope to all positives and only hard negatives, effectively addressing this issue. As a result, our approach substantially outperforms prior art in terms of both accuracy and speed. Finally, the prevailing evaluation method is a FROC analysis over [1/8, 1/4, 1/2, 1, 2, 4, 8] false positives per scan, which is far from ideal in real clinical environments. Regarding practical considerations, we suggest FROC over [1,2,4] false positives as a better metric.

3D Aggregated Faster R-CNN for General Lesion Detection

“…Second, there are many other factors that also affect the performance such as re-shuffle and cropping strategies, hyper-parameter settings and the implementation details. This is illustrated perfectly by the large performance gap between the two Res18-like networks [10], [11]. Effective modifications include substituting the ReLU activation and NMS to Randomized ReLU and soft-NMS respectively as well as leveraging the multi-scale technique.…”

“…Recently, CNN based approaches dominate this task. These CNNs can either be 2D [7], [8] or 3D [9], [10], [11]. In general, 3D CNNs often deliver better performance than 2D CNNs.…”

“…Though they both achieved very good accuracy performance, many drawbacks exist. For instance, in [11], the authors traded off much inference speed for accuracy by adopting very complicated anchor box settings and soft Non-Maximum Suppression (NMS). In [7], the 3D object detection task is reduced to a 2D task with the "center slice" known in advance.…”

“…Compared with plain 2D CNNs, a considerable performance boost is yielded. Concurrently, 3D U-Net variants [10], [11] became popular and achieved huge success. These variants differ in building blocks and the data strategy.…”

“…As a result, it is not possible to feed the whole image into the whole network (even with the help of semantic segmentation). To address this issue, people resort [9], [10], [11] to the divideand-conquer strategy: using sub-cubes instead of the whole Fig. 2.…”