Automatic Polyp Recognition in Colonoscopy Images Using Deep Learning and Two-Stage Pyramidal Feature Prediction

Polyp recognition in colonoscopy images is crucial for early colorectal cancer detection and treatment. However, the current manual review requires undivided concentration of the gastroenterologist and is prone to diagnostic errors. In this article, we present an effective, two-stage approach called PLPNet, where the abbreviation PLP stands for the word polyp, for automated pixel-accurate polyp recognition in colonoscopy images using very deep convolutional neural networks (CNNs). Compared to hand-engineered approaches and previous neural network architectures, our PLPNet model improves recognition accuracy by adding a polyp proposal stage that predicts the location box with polyp presence. Several schemes are proposed to ensure the model's performance. First of all, we construct a polyp proposal stage as an extension of the faster R-CNN, which performs as a region-level polyp detector to recognize the lesion area as a whole and constitutes stage I of PLPNet. Second, stage II of PLPNet is built in a fully convolutional fashion for pixelwise segmentation. We define a feature sharing strategy to transfer the learned semantics of polyp proposals to the segmentation task of stage II, which is proven to be highly capable of guiding the learning process and improve recognition accuracy. Additionally, we design skip schemes to enrich the feature scales and thus allow the model to generate detailed segmentation predictions. For accurate recognition, the advanced residual nets and feature pyramids are adopted to seek deeper and richer semantics at all network levels. Finally, we construct a two-stage framework for training and run our model convolutionally via a single-stream network at inference time to efficiently output the polyp mask. Experimental results on public data sets of GIANA Challenge demonstrate the accuracy gains of our approach, which surpasses previous state-of-the-art methods on the polyp segmentation task (74.7 Jaccard Index) and establishes new top results in the polyp localization challenge (81.7 recall). Note to Practitioners-Given the current manual review of colonoscopy is laborious and time-consuming, computational methods that can assist automatic polyp recognition will enhance the outcome both in terms of efficiency and diagnostic accuracy of colonoscopy. This article suggests a new approach using a very deep convolutional neural network (CNN) architecture for polyp recognition, which gains accuracy from deeper and richer representations. The method, called PLPNet, can effectively detect polyps in colonoscopy images and generate high-quality segmentation masks in a pixel-to-pixel manner. We evaluate the proposed framework on publicly available data sets, and we show by experiments that our method surpasses the state-of-the-art polyp recognition results. The finding of this article corroborates that CNNs with very deep architecture and richer semantics are highly efficient in medical image learning and inference. We believe that the proposed method will facilitate potential computer-aided applications in clinical practice, in that it can enhance medical decision-making in cancer detection and imaging.