ALN: An Anatomy-Aware Network For Unified Full-Volume Liver And Tumour Segmentation In CT

Syeda Husna Mehanoor; Supriya L P; Arun Madhu; Neetha B S; Dharma Teja Lanka; Mihir Harishbhai Rajyaguru; Reji R; Divya Saleela; Jubilant J Kizhakkethottam

doi:10.29284/jk9y5e12

Authors

Syeda Husna Mehanoor Associate Professor, Malla Reddy Technical Campus, Malla Reddy Vishwavidyapeeth (Deemed to be University), Telangana, India
Supriya L P Associate Professor, Saveetha School of Engineering, Saveetha Institute of Medical and Technical Science, Saveetha University, Chennai, India
Arun Madhu Assistant Professor, Saintgits College of Engineering, Kerala, India
Neetha B S Assistant Professor, Fatima Mata National College, Kerala, India
Dharma Teja Lanka Assistant Professor, VNR Vignana Jyothi Institute of Engineering & Technology, Telangana, India
Mihir Harishbhai Rajyaguru Assistant Professor, Madhuben and Bhanubhai Patel Institute of Technology (MBIT) - The Charutar Vidya Mandal (CVM) University, Gujarat, India
Reji R Professor, Thangal Kunju Musaliar Institute of Technology, Kerala, India
Divya Saleela Assistant Professor, College of Engineering Aranmula, Kerala, India
Jubilant J Kizhakkethottam Professor, SRM Institute of Science and Technology, Chennai, India

DOI:

https://doi.org/10.29284/jk9y5e12

Keywords:

Liver tumour segmentation; computed tomography; full volume learning; Medical Segmentation Decathlon.

Abstract

Accurate joint segmentation of the liver and intra hepatic tumours from computed tomography images remains challenging due to extreme class imbalance, scale disparity, and structural ambiguity of sparse lesions. Existing methods frequently prioritise liver delineation, rely on cascaded pipelines that propagate errors, or optimise solely for overlap based metrics that fail to preserve tumour morphology and suppress false positive leakage. This work proposes ALN, an anatomy aware liver network for unified full volume liver tumour segmentation that jointly models organ level context and tumour specific structure within a single end to end framework. ALN employs a shared volumetric backbone with dual semantic and structural heads and is optimised using a composite objective that integrates multi class and class specific Dice supervision, tumour focused focal loss, explicit background suppression, and structure aware constraints to stabilise learning under severe sparsity. Evaluation on the Medical Segmentation Decathlon Task03 Liver dataset using volume level inference demonstrates near ceiling liver delineation with a Dice score of 0.98 ± 0.01 and robust tumour segmentation with a Dice score of 0.86 ± 0.04, accompanied by balanced tumour precision and recall of 0.88 ± 0.05 and 0.85 ± 0.06, respectively. Ablation analysis confirms that each component of the composite objective and the structural supervision head contributes to improved tumour segmentation stability and balanced precision–recall behaviour.

References

1. M. Y. Ansari et al., “Practical utility of liver segmentation methods in clinical surgeries and interventions,” BMC Med Imaging, vol. 22, no. 1, p. 97, May 2022, doi: 10.1186/s12880-022-00825-2.

2. M. C. Brunese, A. Rocca, A. Santone, M. Cesarelli, L. Brunese, and F. Mercaldo, “Explainable and Robust Deep Learning for Liver Segmentation Through U-Net Network,” Diagnostics, vol. 15, no. 7, p. 878, Mar. 2025, doi: 10.3390/diagnostics15070878.

3. C. Grapă et al., “Turning the Tide—Artificial Intelligence in the Evolving Landscape of Liver Cancer,” Cancers, vol. 17, no. 18, p. 3003, Sep. 2025, doi: 10.3390/cancers17183003.

4. F. Alshomrani, “Recent Advances in Magnetic Resonance Imaging for the Diagnosis of Liver Cancer: A Comprehensive Review,” Diagnostics, vol. 15, no. 16, p. 2016, Aug. 2025, doi: 10.3390/diagnostics15162016.

5. A. Gotra et al., “Liver segmentation: indications, techniques and future directions,” Insights Imaging, vol. 8, no. 4, pp. 377–392, Aug. 2017, doi: 10.1007/s13244-017-0558-1.

6. M. Y. Ansari et al., “Practical utility of liver segmentation methods in clinical surgeries and interventions,” BMC Med Imaging, vol. 22, no. 1, p. 97, May 2022, doi: 10.1186/s12880-022-00825-2.

7. H. Ling, S. K. Zhou, Y. Zheng, B. Georgescu, M. Suehling, and D. Comaniciu, “Hierarchical, learning-based automatic liver segmentation,” in 2008 IEEE Conference on Computer Vision and Pattern Recognition, Anchorage, AK, USA: IEEE, Jun. 2008, pp. 1–8. doi: 10.1109/CVPR.2008.4587393.

8. M. Moghbel, S. Mashohor, R. Mahmud, and M. I. B. Saripan, “Review of liver segmentation and computer assisted detection/diagnosis methods in computed tomography,” Artif Intell Rev, vol. 50, no. 4, pp. 497–537, Dec. 2018, doi: 10.1007/s10462-017-9550-x.

9. X. Liu, L. Qu, Z. Xie, Y. Shi, and Z. Song, “Deep Mutual Learning among Partially Labeled Datasets for Multi-Organ Segmentation,” IEEE Trans. Med. Imaging, pp. 1–1, 2025, doi: 10.1109/TMI.2025.3614853.

10. S. Hariharan, D. Anandan, M. Krishnamoorthy, V. Kukreja, N. Goyal, and S.-Y. Chen, “Advancements in Liver Tumor Detection: A Comprehensive Review of Various Deep Learning Models,” CMES, vol. 142, no. 1, pp. 91–122, 2025, doi: 10.32604/cmes.2024.057214.

11. K. Sethia et al., “Advances in liver, liver lesion, hepatic vasculature, and biliary segmentation: a comprehensive review of traditional and deep learning approaches,” Artif Intell Rev, vol. 58, no. 10, p. 299, Jul. 2025, doi: 10.1007/s10462-025-11310-x.

12. M. Moghbel, S. Mashohor, R. Mahmud, and M. I. B. Saripan, “Review of liver segmentation and computer assisted detection/diagnosis methods in computed tomography,” Artif Intell Rev, vol. 50, no. 4, pp. 497–537, Dec. 2018, doi: 10.1007/s10462-017-9550-x.

13. Y. S. Velichko, N. Gennaro, M. Karri, M. Antalek, and U. Bagci, “A Comprehensive Review of Deep Learning Approaches for Magnetic Resonance Imaging Liver Tumor Analysis,” Advances in Clinical Radiology, vol. 5, no. 1, pp. 1–15, Sep. 2023, doi: 10.1016/j.yacr.2023.06.001.

14. K. S. S. and V. K. R. S., “Literature survey on deep learning methods for liver segmentation from CT images: a comprehensive review,” Multimed Tools Appl, vol. 83, no. 28, pp. 71833–71862, Feb. 2024, doi: 10.1007/s11042-024-18388-5.

15. S. Hariharan, D. Anandan, M. Krishnamoorthy, V. Kukreja, N. Goyal, and S.-Y. Chen, “Advancements in Liver Tumor Detection: A Comprehensive Review of Various Deep Learning Models,” CMES, vol. 142, no. 1, pp. 91–122, 2025, doi: 10.32604/cmes.2024.057214.

16. M. Ciecholewski and M. Kassjański, “Computational Methods for Liver Vessel Segmentation in Medical Imaging: A Review,” Sensors, vol. 21, no. 6, p. 2027, Mar. 2021, doi: 10.3390/s21062027.

17. A. Gotra et al., “Liver segmentation: indications, techniques and future directions,” Insights Imaging, vol. 8, no. 4, pp. 377–392, Aug. 2017, doi: 10.1007/s13244-017-0558-1.

18. M. Ciecholewski and M. Kassjański, “Computational Methods for Liver Vessel Segmentation in Medical Imaging: A Review,” Sensors, vol. 21, no. 6, p. 2027, Mar. 2021, doi: 10.3390/s21062027.

19. H. B. Jenssen et al., “Challenges and artificial intelligence solutions for clinically optimal hepatic venous vessel segmentation,” Biomedical Signal Processing and Control, vol. 106, p. 107822, Aug. 2025, doi: 10.1016/j.bspc.2025.107822.

20. T. Heimann et al., “Comparison and Evaluation of Methods for Liver Segmentation From CT Datasets,” IEEE Trans. Med. Imaging, vol. 28, no. 8, pp. 1251–1265, Aug. 2009, doi: 10.1109/TMI.2009.2013851.

21. P. F. Christ et al., “Automatic Liver and Tumor Segmentation of CT and MRI Volumes using Cascaded Fully Convolutional Neural Networks,” 2017, arXiv. doi: 10.48550/ARXIV.1702.05970.

22. J. Xie et al., “Lesion segmentation method for multiple types of liver cancer based on balanced dice loss,” Medical Physics, vol. 52, no. 5, pp. 3059–3074, May 2025, doi: 10.1002/mp.17624.

23. Y. Wan, L. Zhang, and M. Wang, “RIS-UNet: A Multi-Level Hierarchical Framework for Liver Tumor Segmentation in CT Images,” Entropy, vol. 27, no. 7, p. 735, Jul. 2025, doi: 10.3390/e27070735.

24. M. Antonelli et al., “The Medical Segmentation Decathlon,” Nat Commun, vol. 13, no. 1, p. 4128, Jul. 2022, doi: 10.1038/s41467-022-30695-9.