Automatic Chromosomal Abnormality Detection Using Varifocal-Net with CNN Chromosomal Abnormality Detection
International Journal of Medical Toxicology and Forensic Medicine,
Vol. 16 (2026),
1 Dey 2026
,
Page 1-7
https://doi.org/10.22037/ijmtfm.v16.51792
Abstract
Background: Chromosomal abnormalities are a major cause of hereditary disorders, congenital anomalies, and developmental impairments. Conventional karyotype analysis relies on manual inspection by cytogenetic experts, making it time-consuming and prone to subjective interpretation.
Methods: To overcome these shortcomings, this article illustrates an automated chromosomal abnormality detection framework based on a Varifocal-Net–integrated Convolutional Neural Network (CNN) architecture. The proposed model adopts a dual-scale learning strategy, where a global-scale network captures overall chromosomal morphology, and a local-scale network extracts fine-grained structural features such as banding patterns and centromere regions. Deep feature learning is achieved using dual VGG-16 backbones enhanced with residual connections and multi-task learning. The system has undergone training and validation on a curated and expert-verified karyotype image dataset.
Results: Significant classification of performance is demonstrated by findings on experiments achieving 99.04% accuracy, 98.63% sensitivity, and 100% specificity, outperforming baseline CNN and residual architectures.
Conclusion: The inferences depicts that the proposed framework offers a reliable and efficient solution for automated analysis of karyotype evidencing promising potential for clinical decision-support applications.
- Chromosomal abnormality detection, Karyotyping, Convolutional neural network, Varifocal-Net, Deep learning, Cytogenetics
How to Cite
References
[1] Abid F, Hamami L. Deep learning-based chromosome classification: A survey. Artif Intell Med. 2021;117:102083. [DOI: 10.1016/j.artmed.2021.102083]
[2] Swati G, Gupta G, Yadav M, Sharma M, Vig L, et al. Siamese neural networks for chromosome classification. Expert Systems with Applications. 2021; 176:114860. [DOI: 10.1016/j.eswa.2021.114860]
[3] Chen X, et al. Automated karyotype analysis using multi-scale convolutional neural networks. Biomedical Signal Processing and Control. 2022; 72:103327. [DOI: 10.1016/j.bspc.2021.103327]
[4] Shafiq M, Gu Z. Deep residual learning for image recognition: A survey. Applied Sciences. 2022; 12:8972. [DOI: 10.3390/app12188972]
[5] Li J, et al. Attention-guided CNNs for fine-grained medical image classification. IEEE Transactions on Medical Imaging. 2023; 42(4):987–998. [DOI: 10.1109/TMI.2022.3221234]
[6] Zhou T, Ye X, Lu H, Zheng X, Qiu S, Liu Y, et al. Densely connected convolutional networks for medical imaging. Medical Image Analysis. 2022; 75:102304. [DOI: 10.1016/j.media.2021.102304]
[7] Liu Y, et al. Weakly supervised deep learning for cytogenetic image analysis. Pattern Recognition. 2024; 145:109923. [DOI: 10.1016/j.patcog.2023.109923]
[8] Wang C, et al. Fully automatic karyotyping via deep convolutional neural networks. IEEE Access. 2024; 12:46081–92. [DOI: 10.1109/ACCESS.2024.3380829]
[9] Chen X, Wang J, Liu Y, Zhang C, Li H, Zhao F, et al. Varifocal learning for fine-grained visual recognition. IEEE Trans Image Process. 2022;31:3456–3468. [DOI: 10.1109/TIP.2022.3168735]
- Abstract Viewed: 92 times
- PDF Downloaded: 100 times