Abstract:Caenorhabditis elegans (C. elegans) is an excellent model organism because of its short lifespan and high degree of homology with human genes, and it has been widely used in a variety of human health and disease models. However, the segmentation of C. elegans remains challenging due to the following reasons: 1) the activity trajectory of C. elegans is uncontrollable, and multiple nematodes often overlap, resulting in blurred boundaries of C. elegans. This makes it impossible to clearly study the life trajectory of a certain nematode; and 2) in the microscope images of overlapping C. elegans, the translucent tissues at the edges obscure each other, leading to inaccurate boundary segmentation. To solve these problems, a Bilayer Segmentation-Recombination Network (BR-Net) for the segmentation of C. elegans instances is proposed. The network consists of three parts: A Coarse Mask Segmentation Module (CMSM), a Bilayer Segmentation Module (BSM), and a Semantic Consistency Recombination Module (SCRM). The CMSM is used to extract the coarse mask, and we introduce a Unified Attention Module (UAM) in CMSM to make CMSM better aware of nematode instances. The Bilayer Segmentation Module (BSM) segments the aggregated C. elegans into overlapping and non-overlapping regions. This is followed by integration by the SCRM, where semantic consistency regularization is introduced to segment nematode instances more accurately. Finally, the effectiveness of the method is verified on the C. elegans dataset. The experimental results show that BR-Net exhibits good competitiveness and outperforms other recently proposed instance segmentation methods in processing C. elegans occlusion images.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is **the inaccurate segmentation problem caused by the overlap of Caenorhabditis elegans (C. elegans) in microscope images**. Specifically, for the following two main reasons, traditional segmentation methods are difficult to accurately segment overlapping C. elegans: 1. **Uncontrollable activity trajectories**: The behavior trajectories of C. elegans are not controlled, and multiple nematodes often overlap, resulting in blurred boundaries and making it impossible to clearly study the life trajectories of specific nematodes. 2. **Mutual occlusion of edge tissues**: In microscope images, the transparent parts of the edge tissues of overlapping C. elegans occlude each other, resulting in inaccurate boundary segmentation. To solve these problems, the paper proposes a new bilayer segmentation - recombination network (Bilayer Segmentation - Recombination Network, BR - Net), which is specifically used to handle the segmentation problem of C. elegans instances. BR - Net consists of three modules: 1. **Coarse Mask Segmentation Module (CMSM)**: It is used to extract coarse masks and introduce the Unified Attention Module (UAM) to better identify nematode instances. 2. **Bilayer Segmentation Module (BSM)**: It divides the aggregated C. elegans into overlapping and non - overlapping regions. 3. **Semantic Consistency Recombination Module (SCRM)**: By introducing semantic consistency regularization, it segments nematode instances more accurately. Through these modules, BR - Net can more accurately segment C. elegans instances in complex overlapping scenarios, thereby improving the research accuracy of nematode behavior and morphological characteristics. Experimental results show that BR - Net shows good competitiveness in processing C. elegans overlapping images and is superior to other recently proposed instance segmentation methods. ### Formula Summary - **Multi - task loss function (Coarse Mask Segmentation Loss)**: \[ L_{\text{coarse}} = L_{\text{cls}} + L_{\text{reg}} + L_{\text{cmask}} \] - \(L_{\text{cls}}\): Classification cross - entropy loss \[ L_{\text{cls}} = -\frac{1}{K} \sum_{k = 1}^{K} \sum_{i = 1}^{A_k} y_{i,c} \cdot \log(p_{i,c}) \] - \(L_{\text{reg}}\): Smooth L1 loss \[ L_{\text{reg}} = \text{smooth}_L1(x)=\begin{cases} 0.5x^2 & \text{if } |x| < 1\\ |x| - 0.5 & \text{if } |x| \geq 1 \end{cases} \] - \(L_{\text{cmask}}\): Pixel - level cross - entropy loss \[ L_{\text{cmask}} = -(y_i \cdot \log(p_i)+(1 - y_i) \cdot \log(1 - p_i)) \] - **Bilayer Segmentation Module loss function (Decomposition Loss)**: \[ L_{\text{dec}}=\frac{1}{K} \sum_{k = 1}^{K} \frac{1}{A_k} \sum_{i = 1}^{A_k}(L_{\text{...}}) \]

A Bilayer Segmentation-Recombination Network for Accurate Segmentation of Overlapping C. elegans

WSM-MIL: A Weakly Supervised Segmentation Method with Multiple Instance Learning for C. Elegans Image

Mask R-CNN Based C. Elegans Detection with a DIY Microscope

Automated segmentation and recognition of C. elegans whole-body cells

Modeling the Evolution of Biological Neural Networks Based on Caenorhabditis Elegans Connectomes Across Development.

Deep Neural Networks to Register and Annotate the Cells of the C. elegans Nervous System

DoNet: Deep De-overlapping Network for Cytology Instance Segmentation

Celeganser: Automated Analysis of Nematode Morphology and Age

Establishment of morphological atlas of Caenorhabditis elegans embryo with cellular resolution using deep-learning-based 4D segmentation

ERDUnet: An Efficient Residual Double-coding Unet for Medical Image Segmentation

ARMS Net: Overlapping chromosome segmentation based on Adaptive Receptive field Multi-Scale network

High-Resolution Imaging and Morphological Phenotyping of C. elegans through Stable Robotic Sample Rotation and Artificial Intelligence-Based 3-Dimensional Reconstruction

Deep Reinforcement Learning of Cell Movement in the Early Stage of C. elegans Embryogenesis

The Application Research of Deep Neural Networks in Colonic Polyp Segmentation

Multilayer network analysis of C. elegans: Looking into the locomotory circuitry

CA2.5-Net Nuclei Segmentation Framework with a Microscopy Cell Benchmark Collection

BDG-Net: Boundary Distribution Guided Network for Accurate Polyp Segmentation

MSGSE-Net: Multi-scale guided squeeze-and-excitation network for subcortical brain structure segmentation

Segmentation and Classification of Two-Channel C. Elegans Nucleus-Labeled Fluorescence Images

See Elegans: Simple-to-use, accurate, and automatic 3D detection of neural activity from densely packed neurons

Rapid detection and recognition of whole brain activity in a freely behaving Caenorhabditis elegans