The proposed GPT2-based table-to-text framework provides a practical and clinically interpretable approach for disease prediction from limited structured healthcare data and demonstrates strong potential for early risk detection, transparent clinical decision support, and reliable deployment in real-world low-resource healthcare environments.
Abstract
Background and Objective: Early and reliable disease prediction from structured clinical data remains challenging when datasets are small, highly imbalanced, and contain limited positive disease cases. Conventional machine learning (ML) and deep learning approaches often struggle to capture clinically meaningful relationships under such low-data representation conditions due to weak statistical associations between features and prediction targets. This study proposes a clinically grounded GPT2-based table-to-text framework for disease prediction using structured healthcare datasets, motivated by the contextual reasoning capability of GPT models to better capture clinically meaningful relationships when statistical learning alone becomes insufficient due to limited data availability. Methods & Materials: Structured clinical records were transformed into physician-style textual descriptions and enriched through GPT4-generated medical paraphrasing to improve minority-class representation while preserving clinical meaning. Both the original and generated clinical texts were used to fine-tune a GPT2 model across four public healthcare datasets, including heart disease, heart failure, chronic kidney disease, and thyroid cancer recurrence. Gradient-based explainable AI analysis was additionally incorporated to identify clinically important features influencing prediction outcomes. Results: The proposed framework demonstrated consistently strong predictive performance with average precision, specificity, sensitivity, and F1-score of 0.96, 0.97, 0.96, and 0.96, respectively. The model achieved improved sensitivity, stronger generalization, and more stable predictive behavior compared with traditional ML, deep learning, transformer-based, and GAN-augmented approaches. Importantly, the framework consistently emphasized clinically meaningful variables even under severe class imbalance where conventional ML and neural network models often struggled. Conclusions: The proposed GPT2-based table-to-text framework provides a practical and clinically interpretable approach for disease prediction from limited structured healthcare data. By integrating contextual clinical reasoning with explainable prediction mechanisms, the framework demonstrates strong potential for early risk detection, transparent clinical decision support, and reliable deployment in real-world low-resource healthcare environments.
A hybrid framework for a comprehensive evaluation of machine learning and transformer-based models for cardiovascular risk prediction by integrating structured clinical data with unstructured clinical narratives is included in this study. The structured data includes demographic and diagnostic variables like patient information and test results, while unstructured data is derived from electronic health records such as clinical notes. Through preprocessing, feature engineering, and semantic embedding using transformer-based models, the system leverages the complementary strengths of both data types. By analysing and processing this unstructured information, this project aims to improve the predictions for heart disease. The results show that it is feasible to make better predictions and gain valuable insights by merging these two types of data. Such analysis is essential because structured data alone often overlooks fine-grained clinical indicators found in narrative texts. Integrating unstructured data allows for a more holistic view of patient health, leading to earlier detection, personalized interventions, and improved decision-making in clinical settings.
The reviewed studies demonstrated that Deep learning approaches, particularly convolutional neural networks, transformer-based models, and multimodal frameworks, showed improved predictive accuracy when large and diverse datasets were available.
Tehreem Khan, Tayyaba Usman, Ifrah Khalid et al.· Discover Artificial Intellig...· 0 citations
Clinical biostatistics has traditionally relied on regression-based models to explain associations, estimate risks, and support medical decision-making. The growth of electronic health records, imaging data, omics data, and follow-up data has made clinical information larger, less tidy, and more difficult to model with only classical methods. Machine learning is increasingly used in this setting because it can capture non-linear patterns and handle high-dimensional predictors. This narrative review summarizes applications of supervised learning, unsupervised learning, and deep learning in clinical diagnosis, prognosis prediction, patient stratification, and biomarker discovery. Literature was selected from PubMed, Web of Science, and Google Scholar, with emphasis on studies and reporting guidelines published from 2019 to 2025. This review finds that machine learning is useful when the clinical question is clear, data quality is acceptable, and validation is strict. However, a stronger algorithm does not automatically become a better clinical tool. Main limitations include weak interpretability, biased training data, poor transportability, overreliance on Area Under the Curve (AUC), and incomplete reporting. Machine learning should therefore be treated as a complement to traditional biostatistics rather than a simple replacement.
Xiuyi Wei· Journal of Clinical Technolo...· 0 citations
It is indicated that a rigorously conducted methodology and interpretability in machine learning development are crucial in creating machine learning solutions in healthcare decision support, which is the pathway to real applications in diabetes risk assessment.
T. Khan, M. Saeed, Majid Hussain et al.· Scientific Reports· 0 citations
Cardiovascular diseases (CVDs) are among the leading causes of mortality worldwide. Early detection and risk stratification are critical for preventive care. Traditional machine learning (ML) models can predict heart disease risk but often lack interpretability and fail to integrate with real-time clinical data. Recent advances in fine-tuned large language models (Custom GPT) offer natural language explanations but are limited by insufficient interoperability with heterogeneous healthcare data sources. This study aims to design and evaluate a Model Context Protocol (MCP)–enabled Custom GPT framework that integrates ML-based predictive models with external healthcare systems—including EHRs, laboratory APIs, and wearable devices—to deliver context-aware, explainable, and clinically actionable heart disease risk predictions. The experimental evaluation was conducted on a validated cardiovascular dataset containing 303 patient records and 14 clinically relevant attributes derived from publicly available clinical repositories. Experimental evaluation demonstrated improved predictive accuracy (approximately 88% with the XGBoost ensemble) and robustness compared to standalone models. MCP integration enabled dynamic contextual awareness, reduced latency in tool orchestration, and enriched interpretability through RAG-based explanations. Clinician and patient evaluations confirmed enhanced usability and transparency. This approach paves the way for broader adoption of agentic AI in clinical workflows.
Neha Gupta, Bhawna Singla· Journal of Electronic &...· 0 citations
Prognostic modelling is an essential tool in clinical sciences. A machine learning approach can efficiently help with disease prediction and prevention by integrating large amounts of structured patient records. This study develops a deep learning framework to assist in early disease detection and clinical decision-making. A hospital patient dataset was used to develop a predictive model for 13 chronic conditions. The dataset was preprocessed, including the use of a denoising autoencoder neural network for imputations. To improve the predictive ability of the model and reduce overfitting, a novel multi-model approach was used in which ten additional neural networks were trained separately and used to predict the conditions through bootstrapping and bagging techniques. The program was able to predict the 13 conditions with an average accuracy of 96.95% and a micro-averaged F1 score of 0.91 on an independent test set, with the F1 score calculated to account for the presence of class imbalance. This program can potentially aid in enabling early disease prediction, facilitating timely treatment, and improving quality of life while lowering treatment costs and helping hospitals, emergency rooms, caregivers, and individuals with diagnosis predictions, prompt decision-making, and resource prioritization, especially in times of limited healthcare resources.
