CaraNet: context axial reverse attention network for segmentation of small medical objects

Abstract

Purpose: Segmenting medical images accurately and reliably is important for disease diagnosis and treatment. It is a challenging task because of the wide variety of objects' sizes, shapes, and scanning modalities. Recently, many convolutional neural networks have been designed for segmentation tasks and have achieved great success. Few studies, however, have fully considered the sizes of objects; thus, most demonstrate poor performance for small object segmentation. This can have a significant impact on the early detection of diseases.Approach: We propose a context axial reverse attention network (CaraNet) to improve the segmentation performance on small objects compared with several recent state-of-the-art models. CaraNet applies axial reserve attention and channel-wise feature pyramid modules to dig the feature information of small medical objects. We evaluate our model by six different measurement metrics.Results: We test our CaraNet on segmentation datasets for brain tumor (BraTS 2018) and polyp (Kvasir-SEG, CVC-ColonDB, CVC-ClinicDB, CVC-300, and ETIS-LaribPolypDB). Our CaraNet achieves the top-rank mean Dice segmentation accuracy, and results show a distinct advantage of CaraNet in the segmentation of small medical objects. Conclusions:We proposed CaraNet to segment small medical objects and outperform state-ofthe-art methods.

WS-SfMLearner: self-supervised monocular depth and ego-motion estimation on surgical videos with unknown camera parameters

“…The cost volume method can still be improved by learning depth plane distribution rather than using linear distribution directly such as using sinusoidal activation function [33], which will be considered in our future work. Additionally, addressing the challenge from moving objects, such as surgical tools within the scene, numerous segmentation methods [34][35][36][37][38] have been designed to segment and recognize medical objects. Incorporating these segmentation techniques into WS-SfMLearner to alleviate dynamic uncertainties and refine depth prediction precision for the surgical background would an interesting direction.…”

BGNet: Boundary‐guided network for polyp segmentation

“…Lou et al proposed CaraNet 30 . CaraNet 30 fuses shallow and deep features through a context axial reverse attention module to enhance the learning of global and local semantics. However, it only combines dual‐scale features without utilizing multiscale information.…”

“…To address the difficulties in the segmentation process, researchers have proposed various network architectures for optimization. Lou et al proposed CaraNet 30 . CaraNet 30 fuses shallow and deep features through a context axial reverse attention module to enhance the learning of global and local semantics.…”

SAMSNeRF: segment anything model (SAM) guided dynamic surgical scene reconstruction by neural radiance field (NeRF)

“…To address these uncertainties, segmentation networks [8][9] can be used to filter the background and foreground in pixel-level detail. We apply the Segment Anything Model (SAM) [10] to segment the surgical tool using a rough bounding box as input prompt, which provides a more accurate segmentation mask compared to some other CNNbased methods [11]. Our goal in this paper is to build a model to reconstruct the surgical scene and track the movement of surgical tools.…”