Clinical Feature-Based Polycystic Ovary Syndrome Detection Using an Interpretable Kolmogorov-Arnold Network and Hybrid Feature Selection
International Journal of Medical Toxicology and Forensic Medicine,
Vol. 16 (2026),
1 January 2026
,
Page 1-9
https://doi.org/10.22037/ijmtfm.v16.53481
Abstract
Background: Polycystic Ovary Syndrome (PCOS) is a prevalent hormonal disorder affecting women and can lead to menstrual irregularities, hormonal imbalance, infertility, and other health complications. The availability of clinical and physical parameters has encouraged the application of Artificial Intelligence (AI) and deep learning (DL) techniques for automated PCOS detection. However, existing approaches still face challenges in selecting the most relevant clinical features and learning complex nonlinear relationships between patient variables while maintaining the interpretability of predictions. To address these issues, this paper introduces an Interpretable Kolmogorov-Arnold Network-based framework for Polycystic Ovary Syndrome detection (IKAN-PCOS).
Methods: The proposed framework applies a hybrid mRMR-Boruta feature selection approach to identify relevant clinical attributes while reducing redundant information. A Tabular Kolmogorov-Arnold Network is then employed to learn nonlinear relationships among the selected clinical features and classify PCOS and non-PCOS cases. During model training, the AdaPlus optimizer is used to improve convergence and learning performance. Furthermore, SHAP and LIME-based explainable artificial intelligence techniques are incorporated to provide global and local explanations of the model predictions by identifying the contribution of individual clinical features. The explanations help identify the clinical attributes that most influence PCOS and No PCOS classification and provide a clearer understanding of the model's decision process.
Results: The proposed IKAN-PCOS framework achieved 96.93% accuracy in PCOS and No PCOS classification, demonstrating its effectiveness in accurately detecting PCOS from clinical features while providing interpretable insights into classification decisions.
Conclusion: The proposed IKAN-PCOS model supports accurate and interpretable PCOS classification by integrating effective feature selection, nonlinear learning, and XAI techniques.
- Polycystic Ovary Syndrome, Clinical Data, Kolmogorov-Arnold Network, mRMR-Boruta, AdaPlus Optimizer, Explainable Artificial Intelligence
How to Cite
References
[1] Ramteke M, Raut S. Prediction of polycystic ovary syndrome using machine learning with SFS and Boruta feature selection: an explainable AI approach. Systems Biology in Reproductive Medicine. 2025;71(1):439-60. [DOI: 10.1080/19396368.2025.2560839]
[2] Efty MRH, Rohman MN, Uddin KMM, Rahman MM, Uddin MA. A Two‐Stage Hybrid Feature Selection and Ensemble Learning Framework with Explainable AI for Accurate PCOS Prediction. Health Science Reports. 2026;9(9):e73100. [DOI: 10.1002/hsr2.73100]
[3] Sah S, Myneni M, Sulaiman RB. A Fused Interpretable Architecture for PCOS Detection: Leveraging Deep learning and Ensemble learning with Explainable AI. Procedia Computer Science. 2026;282:369-78. [DOI: 10.1016/j.procs.2026.05.084]
[4] Nayem SI, Efat AH, Surhyth AI. X-decision tree: An explainable AI-based architecture for PCOS prediction with hyperparameter tuning and cross-validation. In: 2025 International Conference on Quantum Photonics, Artificial Intelligence, and Networking (QPAIN); 2025 July. IEEE; p. 1-6. [DOI: 10.1109/QPAIN66474.2025.11171903]
[5] Malode A, Injamuri V, Pawar V, Atkane S. Explainable AI-Based PCOS Detection Using SMOTE-Balanced Clinical Markers. International Journal of Computer Information Systems and Industrial Management Applications. 2026;18(15s):282-96. [DOI: 10.70917/ijcisim-2026-4403]
[6] Athapaththu SHSK, Malkanthi AMC. Explainable AI-Driven Early Screening of Polycystic Ovary Syndrome. In: 2026 6th International Conference on Advanced Research in Computing (ICARC); 2026 February. IEEE; p. 1-6. [DOI: 10.1109/ICARC68737.2026.11454090]
[7] Vijay M, Nandhini N, Mallikarjun I, Kumar GJ, Mounika K, Lavanya LL. Explainable AI-based Predictive Model for Early Diagnosis of Polycystic Ovary Syndrome (PCOS). In: 2026 Second International Conference on Intelligent Systems for Communication, IoT and Security (ICISCoIS); 2026 January. IEEE; p. 1-7. [DOI: 10.1109/ICISCoIS62701.2026.11447799]
[8] Babu P, Laxmanan DL. Explainable AI-Driven LeNet-Deep Maxout Network-Based Polycystic Ovary Syndrome Detection Using Ultrasound Image. Journal of Bionic Engineering. 2026:1-30. [DOI: 10.1007/s42235-026-00916-8]
[9] Malik N, Srivastava I, Sachdeva R. Machine Learning for Early Diagnosis of PCOS Using Explainable AI. In: International Conference on Innovative Computing and Communication. Cham: Springer Nature Switzerland; 2026 February. p. 284-95. [DOI: 10.1007/978-3-032-28310-8_20]
[10] Baalebail B, Vrishank HJ, Rao ABR, Shetty AJ, Kodipalli A, Thomas M. Deep Learning-Driven Multi-Modal Diagnosis of Polycystic Ovary Syndrome (PCOS) with Explainable AI. In: 2026 Contemporary Computing Innovations Conference (CCIC); 2026 February. IEEE; p. 1-6. [DOI: 10.1109/CCIC68129.2026.11485971]
[11] Jaganathan G, Natesan S. Blockchain and explainable-AI integrated system for Polycystic Ovary Syndrome (PCOS) detection. PeerJ Computer Science. 2025;11:e2702. [DOI: 10.7717/peerj-cs.2702]
[12] Patel K, Rai S, Nischal J, Pandey D, Shrivastava RK, Hazra D. Polycystic Ovary Syndrome (PCOS) Detection Using Deep Learning and Explainable AI. In: 2025 2nd International Conference on Advanced Computing and Emerging Technologies (ACET); 2025 November. IEEE; p. 1-6. [DOI: 10.1109/ACET67282.2025.11430323]
[13] Shah IB, Yadav PK. Automated PCOS Detection Using Fine-Grained Deep Feature Extraction and Explainable AI: A Transformer-Based Ensemble Approach. International Journal of Image, Graphics and Signal Processing. 2026;18(1):88. [DOI: 10.5815/ijigsp.2026.01.06]
[14] Polycystic Ovary Syndrome (PCOS) Dataset. Kaggle. [Link]
[15] Peng H, Long F, Ding C. Feature selection based on mutual information criteria of max-dependency, max-relevance, and min-redundancy. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2005;27(8):1226-38. [DOI: 10.1109/TPAMI.2005.159]
[16] Farzanehpour M, Soleimanjahi H, Hassan Z, Amanzadeh A, Ghaemi A, Fazeli M. HSP70 modified response against HPV based tumor. European Review for Medical & Pharmacological Sciences. 2013;17(2):228. [Link]
[17] Guan L. AdaPlus: Integrating Nesterov momentum and precise stepsize adjustment on AdamW basis. In: ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP); 2024 April. IEEE; p. 5210-4. [DOI: 10.1109/ICASSP48485.2024.10447337]
[18] Salih AM, Raisi‐Estabragh Z, Galazzo IB, Radeva P, Petersen SE, Lekadir K, Menegaz G. A perspective on explainable artificial intelligence methods: SHAP and LIME. Advanced Intelligent Systems. 2025;7(1):2400304. [DOI: 10.1002/aisy.20]
[19] Ahmad R, Maghrabi LA, Khaja IA, Maghrabi LA, Ahmad M. SMOTE-based automated PCOS prediction using lightweight deep learning models. Diagnostics. 2024;14(19):2225. [DOI: 10.3390/diagnostics14192225]
[20] Rani V, Venkatachalam C, M KRK. Explainable Artificial Intelligence with Deep Representation Learning Model for Polycystic Ovary Syndrome Detection Using Clinical Features. Journal of Innovative Image Processing. 2026;8(3):916-37. [DOI: 10.36548/jiip.2026.3.009]
- Abstract Viewed: 0 times
- 53481 Downloaded: 0 times