Revolutionizing Liver Disease Diagnosis Through AI Innovation
Researchers from Odisha University of Technology and Research (OUTR) in Bhubaneswar, India, and Prince Sattam bin Abdulaziz University (PSAU) in Saudi Arabia have unveiled a groundbreaking hybrid artificial intelligence (AI) model for predicting liver disease with an impressive 95.49% accuracy. This development marks a significant milestone in medical diagnostics, particularly for early detection where traditional methods often fall short due to high costs and limited specialist availability.
The model, detailed in the February 2026 issue of Engineering, Technology & Applied Science Research, combines deep learning techniques with boosting algorithms to analyze patient data efficiently. It promises to make screening faster, more affordable, and accessible even in resource-constrained settings like rural Indian clinics or government hospitals.
The Rising Burden of Liver Disease in India
Liver diseases, including metabolic dysfunction-associated steatotic liver disease (MASLD, formerly known as non-alcoholic fatty liver disease or NAFLD), represent a silent epidemic in India. Recent studies from the Phenome India cohort, published in The Lancet Regional Health, reveal that nearly 40% of Indian adults may be affected by MASLD, driven by rising obesity, diabetes, and sedentary lifestyles. Urban areas like Bengaluru, Chennai, and Hyderabad show even higher rates, with up to 41% prevalence among adults.
Traditional diagnostics such as liver biopsies or advanced imaging are invasive, expensive, and require expert interpretation, leading to late-stage detections. In India, where liver cancer and cirrhosis claim thousands of lives annually, early prediction tools are crucial. The World Health Organization notes that liver diseases contribute to over 2 million global deaths yearly, with India bearing a disproportionate share due to its population size and lifestyle shifts.
Behind the Research: A Collaborative Effort Between OUTR and PSAU
The research team comprises experts from OUTR's Department of Computer Science and Engineering and PSAU's College of Computer Engineering and Sciences. Key contributors include Sanjit Kumar Dash and Nitish Agrawal from OUTR, alongside Mohammed Altaf Ahmed (corresponding author), Suleman Alnatheer, and Qutubuddin Mohammed from PSAU.
Sanjit Kumar Dash emphasized, "The model is aimed at supporting early diagnosis, as many liver diseases show no symptoms until they become severe. Early detection can improve treatment outcomes and reduce pressure on the healthcare system." This international collaboration highlights OUTR's growing prowess in AI-driven research, bolstered by its Center of Excellence in Artificial Intelligence (CoE-AI) established with Tech Mahindra.
OUTR, formerly College of Engineering and Technology (CET) Bhubaneswar, is rapidly advancing as a hub for tech innovation, planning additional CoEs in robotics and design by March 2026. Such partnerships not only elevate Indian higher education's global standing but also foster knowledge exchange in critical fields like healthcare AI.
Dataset and Preprocessing: Building a Robust Foundation
The model leverages the Indian Liver Patient Dataset (ILPD) from the UCI Machine Learning Repository, comprising 583 records with 10 key features: age, gender, total bilirubin, direct bilirubin, alkaline phosphatase, alanine aminotransferase (ALP), aspartate aminotransferase (AST), total proteins, albumin, and albumin-globulin ratio (A/G ratio). Notably, 416 cases represent liver patients, while 167 are non-patients, creating a class imbalance.
- Feature engineering introduced medically insightful ratios, such as direct-to-total bilirubin, to capture subtle pathological changes.
- SMOTE-Tomek resampling balanced the dataset by oversampling minority (healthy) cases and undersampling majority (diseased) ones, minimizing bias.
- Data normalization ensured consistent scales across features, enhancing model training stability.
This preprocessing pipeline is vital for real-world applicability, where datasets often suffer from imbalances reflective of disease prevalence in India.
Unpacking the Hybrid Machine Learning Model
The core innovation is a hybrid framework integrating deep learning's Multi-Layer Perceptron Neural Network (MLPNN) with a soft voting classifier ensemble of Extreme Gradient Boosting (XGBoost or XGB) and Light Gradient Boosting Machine (LGBM). Here's how it works step-by-step:
- Input Layer: Patient features fed into MLPNN for non-linear pattern extraction via multiple hidden layers with ReLU activation.
- Deep Learning Phase: MLPNN learns hierarchical representations, capturing complex interactions like enzyme-bilirubin correlations.
- Boosting Ensemble: Outputs combined with XGB and LGBM, which iteratively correct errors using gradient descent on decision trees.
- Soft Voting: Final prediction via averaged probabilities from all models, reducing variance and boosting reliability.
- Output: Binary classification (diseased/non-diseased) with probability scores.
This architecture outperforms standalone models, as boosting handles tabular data efficiently while deep learning excels in feature abstraction.
Impressive Performance Metrics and Benchmarks
The hybrid model achieved 95.49% accuracy, surpassing prior efforts like random forest (94.6%) or single boosting models. Key metrics include:
| Metric | Value |
|---|---|
| Accuracy | 95.49% |
| Precision | 98.4% |
| Specificity | 98.50% |
| ROC-AUC | High (validated via curve) |
Confusion matrix analysis showed minimal false positives/negatives, with strong generalization on hold-out sets. Compared to global AI liver models (e.g., 92-97% for specific subtypes), this sets a new benchmark for general liver disease prediction using routine blood tests.
Transformative Implications for Indian Healthcare
In India, where nearly 40% face MASLD risk and specialist shortages plague rural areas, this model deploys on standard computers—no GPUs needed. It supports telemedicine, enabling primary care physicians to triage patients swiftly, reducing referrals to overburdened tertiary centers.
Mohammed Altaf Ahmed noted, "The hybrid approach helps the model capture complex medical patterns while remaining efficient, and can serve as a decision-support tool in settings with limited diagnostic facilities." For higher education, it exemplifies how universities like OUTR drive societal impact through applied AI research. Explore research jobs in AI healthcare at platforms like AcademicJobs.com.
Read the full research paperOUTR's Growing Leadership in AI and Higher Education Research
OUTR's CoE-AI facilitates such breakthroughs, training students in machine learning for healthcare. The university's evolution from CET to a research powerhouse underscores Odisha's tech ambitions. Collaborations like this with PSAU enhance global ties, preparing graduates for higher ed jobs in AI. Recent achievements include AI for solar panels and skin disorder prediction, positioning OUTR as a key player.
Students and faculty can leverage resources like academic CV tips for advancing in this field.
Challenges, Ethical Considerations, and Future Outlook
While promising, researchers caution that real-world accuracy may vary due to dataset limitations and lab variations. Future work includes multi-hospital validations and integration with wearables for continuous monitoring. Ethical AI deployment requires clinician oversight to avoid over-reliance.
- Benefits: Cost savings (vs. biopsies), scalability, equity in access.
- Risks: Data privacy, bias if unaddressed, need for regulatory approval.
- Solutions: Federated learning for privacy, diverse datasets, clinical trials.
Sanjit Kumar Dash advocates, "Collaboration between computer scientists, clinicians, and policymakers is essential." As India invests in AI via IndiaAI Mission, expect more such innovations from universities.
Photo by julien Tromeur on Unsplash
Actionable Insights for Researchers, Clinicians, and Students
For aspiring AI researchers in Indian higher ed, replicate this by accessing ILPD via UCI and tools like scikit-learn, XGBoost. Clinicians: Pilot in outpatient settings. Students: Join OUTR-inspired programs; check university jobs or rate my professor for mentors.
To thrive, consider postdoc advice. This model paves the way for AI in tackling India's health challenges.
Times of India coverageVisit higher ed jobs, rate my professor, and higher ed career advice for opportunities.







