Explainable AI for Early Identification of Polycystic Ovary Syndrome using a Calibrated Neural Network

View at publisher Summarise with VIRAT
Authors: S. A. Shalini, Priya Vedhanayagam
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

One of the most widespread hormonal illnesses affecting women of reproductive age is polycystic ovarian syndrome. It is often linked to Infertility, hormonal disorders, and other metabolic conditions. This paper proposes an seamless machine learning system for the timely diagnosis of polycystic ovary syndrome using clinical data. In the preprocessing step, removed the unnecessary characteristics like serial numbers and patient IDs, handling missing values, and encoded categorical variables. Feature selection was performed using the SelectKBest technique, picking the top 25 features most relevant to PCOS prediction. Several classification algorithms, like Extreme Gradient Boosting, Random Forest, Multi Layer Perceptron, and Categorical Boosting were trained and tested. Hyperparameter optimization was conducted using RandomizedSearchCV to improve model accuracy. The model was examined using the F1-score, recall, accuracy, and precision measures. Among those models the MLP classifier displayed the highest performance, with an accuracy of 92% after tuning. To improve the reliability of probabilistic outputs, model calibration was implemented using CalibratedClassifierCV with isotonic regression and improved the final accuracy to 93%. Furthermore, Local interpretable model agnostic explanations were employed to give instance-level interpretability, providing transparency and supporting clinical decision making.

Fields of Study & Subjects

Artificial Intelligence in Healthcare Ovarian function and disorders Explainable Artificial Intelligence (XAI)

Publication details

Journal / source
CRC Press eBooks
Pages
527–532
Publication date
31 March 2026
Language
EN
Document type
Book chapter