Research on Morphometric Methods for Larimichthys crocea Based on YOLOv11-CBAM X-Ray Imaging

Yuan‐Wei Yao, Guangde Qiao, Shengmao Zhang, Chong Wu, Zuli Wu, Tianfei Cheng, Hanfeng ZhengFishes 2025 · DOI

TL;DR

This paper introduces a novel, automated X-ray imaging pipeline for precise morphometric analysis of Large Yellow Croaker fish. By combining the YOLOv11 instance segmentation model with a CBAM attention mechanism, the researchers achieve highly accurate segmentation of internal anatomical structures, overcoming limitations of traditional manual dissection. This non-destructive method offers a standardized and efficient approach for aquaculture management and biological research.

What Problem Does This Paper Address?

Traditional methods for analyzing fish morphometrics, especially for commercially important species like Large Yellow Croaker (Larimichthys crocea), often involve manual dissection. This process is not only time-consuming and labor-intensive but also destructive, leading to potential biases due to tissue deformation and loss of original spatial configuration. Furthermore, these methods struggle with quantifying fine internal structures like otoliths or air bladders. While X-ray imaging offers a non-invasive alternative, existing analysis techniques often require manual region identification or rely on less precise object detection, hindering accurate quantification of subtle anatomical features. This paper aims to bridge this gap by developing a standardized, non-destructive, and automated workflow for detailed internal morphometric analysis using deep learning on X-ray images.

Methods

The core of this research lies in an improved instance segmentation model. The authors leverage the YOLOv11 architecture, known for its strong multi-scale feature extraction and decoupled head for improved localization. To enhance its performance on complex backgrounds and fine-grained targets common in X-ray images, they integrate the Convolutional Block Attention Module (CBAM) into the YOLOv11 framework, creating the YOLOv11-CBAM model. CBAM adaptively recalibrates feature maps by focusing on informative channels and spatial regions, effectively highlighting key anatomical cues while suppressing background noise.

For quantitative analysis, a crucial step is converting pixel-based segmentation masks into biologically meaningful measurements. This is achieved through a novel coin-based calibration pipeline. A coin of known area is placed in each X-ray image, allowing for the calculation of a scale factor to convert pixel counts into absolute areas. This ensures that measurements are comparable and accurate across different imaging conditions.

Datasets

The study utilized a custom dataset comprising 160 X-ray images of Large Yellow Croaker. These images were meticulously annotated to identify five key anatomical categories: whole fish, air bladder, spine, eyes, and otoliths. The dataset was split into training (70%), validation (20%), and testing (10%) subsets for model development and evaluation.

Key Insights and Ideas

  • CBAM's Impact on Fine Structures: Integrating CBAM significantly improved the segmentation of small and low-contrast structures like the air bladder and spine, which are critical for understanding fish health and buoyancy.
  • Standardized Non-Destructive Phenotyping: The proposed workflow offers a robust and reproducible method for morphometric analysis without requiring dissection, opening doors for large-scale studies.
  • Coin Calibration for Real-World Metrics: The coin-based calibration method is a practical solution for converting image data into biologically relevant area measurements, ensuring consistency across different imaging setups.
  • YOLOv11-CBAM as a Balanced Solution: The YOLOv11-CBAM model strikes a good balance between segmentation accuracy and real-time performance, making it suitable for practical applications.

Key Results

The YOLOv11-CBAM model demonstrated superior performance compared to the baseline YOLOv11, achieving higher mAP50 and mAP50–95 scores. Specifically, the model showed notable improvements in segmenting small structures. After calibration, the mean area ratios for the air bladder, otoliths, eyes, and spine were found to be 7.72%, 0.59%, 2.20%, and 8.48%, respectively, with acceptable standard deviations, indicating the robustness of the method. The study also highlighted that the spine and air bladder had relatively low variability in their proportional areas, suggesting consistency in these traits, while otoliths and eyes showed greater variability, likely due to their small size and imaging constraints.

Limitations and Future Work

The authors acknowledge that the dataset size (160 images) is limited and primarily derived from a single aquaculture facility and imaging system. This restricts the generalizability of the findings to broader aquaculture applications without further validation. Future work should focus on expanding the dataset with multi-farm, multi-season, and multi-angle acquisitions to capture more diversity. Integrating other imaging modalities like micro-CT for 3D validation and developing more robust calibration techniques for ultra-small targets (like otoliths) are also suggested. The researchers also propose exploring cross-species generalization of the segmentation framework, given the conserved nature of many internal anatomical structures across teleost lineages.

Summary updated

Abstract

Traditional morphometric analysis of Large Yellow Croaker (Larimichthys crocea) relies heavily on manual dissection, which is time-consuming, labor-intensive, and prone to subjectivity. To address these limitations, we propose an automated quantitative approach based on deep-learning–driven instance segmentation. A dataset comprising 160 X-ray images of L. crocea was established, encompassing five anatomical categories: whole fish, air bladder, spine, eyes, and otoliths. Building upon the baseline YOLOv11-Seg model, we integrated a lightweight Convolutional Block Attention Module (CBAM) to construct an improved YOLOv11-CBAM network, thereby enhancing segmentation accuracy for complex backgrounds and fine-grained targets. Experimental results demonstrated that the modified model achieved superior performance in both mAP50 and mAP50–95 compared with the baseline, with particularly notable improvements in the segmentation of small-scale structures such as the air bladder and spine. By introducing coin-based calibration, pixel counts were converted into absolute areas and relative proportions. The measured area ratios of the air bladder, otoliths, eyes, and spine were 7.72%, 0.59%, 2.20%, and 8.48%, respectively, with standard deviations remaining within acceptable ranges, thus validating the robustness of the proposed method. Collectively, this study establishes a standardized, efficient, and non-destructive workflow for X-ray image-based morphometric analysis, providing practical applications for aquaculture management, germplasm conservation, and fundamental biological research.
BibTeX
@inproceedings{yao2025research,
  author = {Yuan‐Wei Yao and Guangde Qiao and Shengmao Zhang and Chong Wu and Zuli Wu and Tianfei Cheng and Hanfeng Zheng},
  title = {{Research on Morphometric Methods for Larimichthys crocea Based on YOLOv11-CBAM X-Ray Imaging}},
  booktitle = {Fishes},
  year = {2025},
  doi = {10.3390/fishes10120641},
}
Figure