Citation :

Masthan Pasha, Kishore Kumar ATA, V Vijaya Kishore, "A Lightweight Topology Preserving and Stain Adaptive Hybrid Network for Lung and Colon Histopathology Recognition," International Journal of Electronics and Communication Engineering, vol. 13, no. 4, pp. 109-118, 2026. Crossref, https://doi.org/10.14445/23488549/IJECE-V13I4P108

Abstract

Precise and accurate patch-level classification of lung and colon histopathology images remains an open problem due to the similarity in histopathology patches amongst carcinoma types, stain variability amongst institutions, and tile duplication in publicly available datasets such as LC25000. Most of the CNN-based approaches suffer from exaggerated estimates of generalization performance, either due to cross-fold duplication, loss of spatial topological structure present within histopathology images, or ineffective usage of hand-crafted representations in interaction with deep representations. In order to alleviate these challenges, this work proposes PatchGNN-FiLM for the effective amalgamation of four different strategies: convolutional tokenization, globally reasoning within Transformers, spatial refinement using graph neural networks, and adaptive stain transfer in feature-wise linear modulation. To provide a more realistic estimate of generalization performance, this work proposes a novel data duplication-aware and perceptual hash-based approach to cross-fold division, with focused emphasis on avoiding data leaks. The proposed architecture proceeds with the spatially coherent tokenization of VGG-16 feature maps to form a regular lattice structure of spatial dimension 4×4. This would enable Transformers to reason about space, as well as topological message passing with GCNs. A manually designed feature branch is used to produce chroma-texture representations, which are then applied as modifiers to input embeddings in FiLM blocks. The LC25000 test sets indicate that the model is performing in a continuously excellent manner for the test folds with 97.09% accuracy, 0.9717 macro F1, 0.9653 Cohen’s κ values, and a small ECE = 0.0163 for the calibration error. The confusion matrices check that the model is performing satisfactorily with large confusion margins only for similar subtypes of carcinoma. The ablation study for all test sets confirms that a significant benefit for the model’s performance was achieved through all architectural components from the Transformer reasoning layers to the optimization of the GCN topology and the FiLM stain features adaptation. The external validation for the Kather Texture 2016 competition set (5,000 tiles) further confirms the model’s robustness for domain transfer, achieving 95.20% accuracy and a macro F1 score of 0.9484.

Keywords

Histopathology Classification, PatchGNN-FiLM, Transformer and Graph Neural Networks, Stain-Adaptive Feature Modulation, Duplicate-Aware Cross-Validation.

References

1. Andrew A. Borkowski et al., “Lung and Colon Cancer Histopathological Image Dataset (LC25000),” arXiv preprint, 2019.
   [CrossRef] [Google Scholar] [Publisher Link]
2. Jie Ji et al., “Automated Lung and Colon Cancer Classification using Histopathological Images,” Biomedical Engineering and Computational Biology, vol. 15, pp. 1-8, 2024.
   [CrossRef] [Google Scholar] [Publisher Link]
3. A. Hasib Uddin et al., “Colon and Lung Cancer Classification from Multi-Modal Images using Resilient and Efficient Neural Network Architectures,” Heliyon, vol. 10, no. 9, pp. 1-23, 2024.
   [CrossRef] [Google Scholar] [Publisher Link]
4. Omneya Attallah, Muhammet Fatih Aslan, and Kadir Sabanci, “A Framework for Lung and Colon Cancer Diagnosis via Lightweight Deep Learning Models and Transformation Methods,” Diagnostics, vol. 12, no. 12, pp. 1-23, 2022.
   [CrossRef] [Google Scholar] [Publisher Link]
5. Sheeba Santhosh, A. Vimala Juliet, and G. Hari Krishnan, “Impact of Electrodes Separation Distance on Bio-Impedance Diagnosis,” Biomedical and Pharmacology Journal, vol. 14, no. 1, pp. 141-146, 2021.
   [CrossRef] [Google Scholar] [Publisher Link]
6. G. Hari Krishnan, G. Umashankar, and S. Abraham, “Cerebrovascular Disorder Diagnosis using MR Angiography,” Biomedical Research India, vol. 27, no. 3, pp. 773-775, 2016.
   [Google Scholar] [Publisher Link]
7. A. Nagarjuna Reddy, G. Hari Krishnan, and D. Raghuram, “Real Time Patient Health Monitoring using Raspberry Pi,” Research Journal of Pharmaceutical, Biological and Chemical Sciences, vol. 7, no. 6, pp. 570-575, 2016.
   [Google Scholar] [Publisher Link]
8. G. Hari Krishnan, R. Ananda Natarajan, and Anima Nanda, “Microcontroller Based Non-Invasive Diagnosis of Knee Joint Diseases,” International Conference on Information Communication and Embedded Systems (ICICES2014), Chennai, India, pp. 1-3, 2014.
   [CrossRef] [Google Scholar] [Publisher Link]
9. Mohammed Al-Jabbar et al., “Histopathological Analysis for Detecting Lung and Colon Cancer Malignancies using Hybrid Systems with Fused Features,” Bioengineering, vol. 10, no. 3, pp. 1-25, 2023.
   [CrossRef] [Google Scholar] [Publisher Link]
10. Abdulkream A. Alsulami et al., “Identification of Anomalies in Lung and Colon Cancer Using Computer Vision-Based Swin Transformer with Ensemble Model on Histopathological Images,” Bioengineering, vol. 11, no. 10, pp. 1-22, 2024.
    [CrossRef] [Google Scholar] [Publisher Link]
11. G. Hari Krishnan, R. Ananda Natarajan, and Anima Nanda, “Comparative Study of Rheumatoid Arthritis Diagnosis using Two Method,” Biomedical and Pharmacology Journal, vol. 7, no. 1, pp. 379-381, 2014.
    [CrossRef] [Google Scholar] [Publisher Link]
12. Mehedi Masud et al., “A Machine Learning approach to Diagnosing Lung and Colon Cancer using a Deep Learning-Based Classification Framework,” Sensors, vol. 21, no. 3, pp. 1-20, 2021.
    [CrossRef] [Google Scholar] [Publisher Link]
13. Md. Alamin Talukder et al., “Machine Learning-Based Lung and Colon Cancer Detection Using Deep Feature Extraction and Ensemble Learning,” Expert Systems with Applications, vol. 205, 2022.
    [CrossRef] [Google Scholar] [Publisher Link]
14. Shtwai Alsubai et al., “Transfer Learning Based Approach for Lung and Colon Cancer Detection using Local Binary Pattern Features and Explainable Artificial Intelligence (AI) Techniques,” PeerJ Computer Science, pp. 1-21, 2024.
    [CrossRef] [Google Scholar] [Publisher Link]
15. Jakob Nikolas Kather et al., “Multi-Class Texture Analysis in Colorectal Cancer Histology,” Scientific Reports, vol. 6, pp. 1-11, 2016.
    [CrossRef] [Google Scholar] [Publisher Link]
16. Min-Jen Tsai, and Yu-Han Tao, “Deep Learning Techniques for the Classification of Colorectal Cancer Tissue,” Electronics, vol. 10, no. 14, pp. 1-26, 2021.
    [CrossRef] [Google Scholar] [Publisher Link]
17. Lakpa Dorje Tamang, and Byung Wook Kim, “Deep Learning Approaches to Colorectal Cancer Diagnosis: A Review,” Applied Sciences, vol. 11, no. 22, pp. 1-19, 2021.
    [CrossRef] [Google Scholar] [Publisher Link]
18. G. H. Hari Krishnan, R. Ananda Natarajan, and A. Nanda, “Impact of Upper Limb Joint Fluid Variation on Inflammatory Diseases Diagnosis,” Journal of Electrical Engineering and Technology, vol. 9, no. 6, pp. 2114-2117, 2014.
    [Google Scholar] [Publisher Link]
19. R.J. Hemalatha et al., “Computerized Breast Cancer Detection System,” Biosciences Biotechnology Research Asia, vol. 11, no. 2, pp. 907-910, 2014.
    [CrossRef] [Google Scholar] [Publisher Link]
20. V. Guru Anand et al., “Predicting Grade of Prostate Cancer using Image Analysis Software,” Trendz in Information Sciences & Computing(TISC2010), Chennai, India, pp. 122-124, 2010.
    [CrossRef] [Google Scholar] [Publisher Link]
21. G. Mohandass, R. Ananda Natarajan, and G. Hari Krishnan, “Comparative Analysis of Optical Coherence Tomography Retinal Images using Multidimensional and Cluster Methods,” Biomedical Research, vol. 26, no. 2, pp. 273-285, 2015.
    [Google Scholar] [Publisher Link]
22. G. Hari Krishnan et al., “Development of Magnetic Control System for Electric Wheel Chair Using Tongue,” Intelligent Computing, Communication and Devices, pp. 635-641, 2014.
    [CrossRef] [Google Scholar] [Publisher Link]
23. N. Ilangovan, and G. Hari Krishnan, “Wheel Chair Movement Control using Human Input: Comparative Study Approach,” Research Journal of Pharmaceutical, Biological and Chemical Sciences, vol. 6, no. 3, pp. 568-570, 2015.
    [Google Scholar]
24. Sheeba Santhosh, A. Vimala Juliet, and G. Hari Krishnan, “Predictive Analysis of Identification and Disease Condition Monitoring using Bioimpedance Data,” Journal of Ambient Intelligence and Humanized Computing, vol. 12, pp. 2955-2963, 2021.
    [CrossRef] [Google Scholar] [Publisher Link]
25. G. Hari Krishnan et al., “CNN based Image Processing for Crack Detection on HT Insulator’s Surface,” 2023 International Conference on Sustainable Computing and Smart Systems (ICSCSS), Coimbatore, India, pp. 618-621, 2023.
    [CrossRef] [Google Scholar] [Publisher Link]
26. T. Sudhakar et al., “Drug Retrieving System in Hospitals Using Robotics,” Biomedical and Pharmacology Journal, vol. 13, no. 3, pp. 1239-1244, 2020.
    [CrossRef] [Google Scholar] [Publisher Link]
27. Sheeba Santhosh, A. Vimala Juliet, and G. Hari Krishnan, “Simulation of Signal Generation and Measuring Circuit and Real Time IoT based Electrical Bio Impedance Cardiac Monitoring System,” Intelligent Computing, Information and Control Systems, pp. 701-706, 2019.
    [CrossRef] [Google Scholar] [Publisher Link]
28. L. Margreat, and G. Hari Krishnan, Statistical Approach for Diagnosis of Diseases Using Histopathology Data, International Journal of Pharma and Bio Sciences, vol. 6, no. 2, pp. B199-B203, 2015. [Online]. Available: https://www.researchgate.net/publication/282845103_Statistical_approach_for_diagnosis_of_diseases_using_histopathology_data
29. G. Radhakrishna Rao, and G. Hari Krishnan, “Comparative Study of Pacemaker Energy Harvesting Techniques,” Research Journal of Pharmaceutical, Biological and Chemical Sciences, vol. 6, no. 1, pp. 1545-1547, 2015.
    [Google Scholar]
30. P. Nandhini, G. Hari Krishnan, and G. Umashankar, “Home based Telemedicine System for Respiratory Disorder Patients,” International Journal of Pharma and Bio Sciences, vol. 6, no. 4, pp. B227-B231, 2015.
    [Google Scholar] [Publisher Link]