Breakthrough in AI-Driven Liver Disease Prediction from Odisha Researchers

Researchers at Odisha University of Technology and Research (OUTR) in Bhubaneswar have collaborated with experts from Prince Sattam bin Abdulaziz University (PSAU) in Saudi Arabia to create a groundbreaking hybrid artificial intelligence (AI) model for predicting liver disease. Published in the February 2026 issue of Engineering, Technology & Applied Science Research, this model achieves an impressive 95.49% accuracy, setting a new benchmark in early detection capabilities. Liver disease, a silent killer affecting millions, often goes undetected until advanced stages, but this innovation promises to change that by leveraging accessible computing power for widespread screening.

The study highlights how Indian higher education institutions like OUTR are at the forefront of international AI research, addressing pressing public health challenges through interdisciplinary approaches. By combining deep learning and machine learning techniques, the model processes routine blood test data to flag potential risks efficiently, making it ideal for resource-constrained settings in India.

Meet the Research Team Behind the Innovation

The project brings together talented researchers from two continents. Leading from OUTR are Sanjit Kumar Dash, Nitish Agrawal, and Rahul Agarwalla, all affiliated with the Department of Computer Science and Engineering. Their Saudi counterparts include Mohammed Altaf Ahmed as the corresponding author from the College of Computer Engineering and Sciences at PSAU, along with Suleman Alnatheer and Qutubuddin Mohammed from related departments at the same university.

Sanjit Kumar Dash emphasized the model's potential: "Early detection can improve treatment outcomes and reduce pressure on the healthcare system." This collaboration exemplifies how universities in India are fostering global partnerships to tackle healthcare issues, enhancing research output and skill development for students and faculty alike. For aspiring researchers, opportunities in higher education research jobs at institutions like OUTR are expanding rapidly.

The Growing Burden of Liver Disease in India

India faces a staggering liver health crisis. Recent studies, including one from The Lancet Regional Health published in early 2026, reveal that nearly 40% of adults may have metabolic dysfunction-associated steatotic liver disease (MASLD), previously known as non-alcoholic fatty liver disease. Urban areas like Bengaluru, Chennai, and Hyderabad report prevalence rates as high as 40-41%, driven by rising obesity, diabetes, and sedentary lifestyles.

Government estimates peg the overall fatty liver prevalence at 9-53%, positioning India among the global top three hotspots. Traditional diagnostics like biopsies and imaging are invasive, expensive, and scarce in rural areas, where over 65% of the population resides. This hybrid AI model steps in as a non-invasive, cost-effective alternative using standard blood parameters, aligning perfectly with India's National Digital Health Mission goals.

Such research from Indian universities not only addresses local needs but also contributes to global knowledge, with OUTR's work underscoring the role of higher education in public health innovation.

How the Hybrid AI Model Works: A Step-by-Step Breakdown

The model's ingenuity lies in its hybrid architecture, integrating a Multi-Layer Perceptron Neural Network (MLPNN)—a type of deep learning neural network—with a soft voting classifier comprising Extreme Gradient Boosting (XGBoost) and Light Gradient Boosting Machine (LightGBM). Here's the process:

• Data Input: Starts with the Indian Liver Patient Dataset (ILPD) from UCI Machine Learning Repository, featuring 583 records and 10 core attributes like age, total bilirubin, direct bilirubin, alkaline phosphatase, alanine aminotransferase (ALP, ALT), aspartate aminotransferase (AST), total proteins, albumin, albumin-to-globulin ratio, and selector (disease status).
• Feature Engineering: Researchers added derived ratios, such as direct-to-total bilirubin and albumin-to-globulin, to capture subtle physiological patterns linked to liver dysfunction.
• Preprocessing: Handled class imbalance (more diseased cases) using Synthetic Minority Over-sampling Technique combined with Tomek links (SMOTE-Tomek), ensuring balanced training without data loss.
• Model Training: MLPNN extracts non-linear features from engineered data, fed into the voting classifier where XGBoost and LightGBM predictions are averaged softly for final output.
• Prediction Output: Classifies patients as healthy or at risk, runnable on everyday laptops.

This step-by-step fusion makes the model robust against noisy real-world data, outperforming standalone algorithms.

Dataset Insights and Preprocessing Techniques

The ILPD dataset, curated from Indian patients at the Andhra Pradesh Government Hospital, mirrors real Indian demographics with diverse age groups and metabolic profiles. Its 416 diseased vs. 167 healthy cases posed imbalance risks, mitigated adeptly by SMOTE-Tomek, which oversamples minorities and removes noisy boundaries.

Feature selection emphasized clinical relevance: bilirubin levels indicate liver detoxification capacity, while enzymes like ALT/AST signal inflammation. Derived ratios enhance interpretability, helping clinicians trust AI decisions. Validation via k-fold cross-validation ensured generalizability.

Impressive Performance Metrics and Benchmarks

The model shines with 95.49% accuracy, but deeper metrics reveal its strength:

ROC-AUC curves and confusion matrices confirm low false positives/negatives. Compared to baselines like plain XGBoost (92-93%) or LSTM (90%), the hybrid surges ahead by 2-5%, per similar studies. Mohammed Altaf Ahmed noted: "The hybrid approach captures complex patterns efficiently."

Real-World Implications for Indian Healthcare

In India, where liver cirrhosis claims over 300,000 lives annually, this model could democratize screening. Deployable via mobile apps or hospital software, it supports telemedicine in rural Odisha and beyond, reducing reliance on scarce specialists. Cost savings: a blood test (~₹500) plus AI analysis vs. ₹10,000+ for FibroScan.

• Benefits: Early intervention cuts progression to cancer/hepatitis complications.
• Risks: Over-reliance without clinician oversight; data privacy under DPDP Act.
• Stakeholders: ICMR, MoHFW could integrate into Ayushman Bharat Digital Mission.

For higher ed, it inspires curricula in AI-health tracks, with OUTR leading by example.

Challenges, Limitations, and Path Forward

While promising, limitations include ILPD's single-hospital origin and lack of imaging/genomics. Real-world variances in lab standards or comorbidities may dip performance to 85-90%. Future: Multi-center trials, federated learning for privacy, integration with wearables.

Researchers call for clinician-AI-policy collaborations. In Indian higher ed, funding via IndiaAI Mission (₹10,372 Cr) could scale such projects.

Boosting AI Research in Indian Higher Education

OUTR's feat aligns with trends: India ranks 3rd in AI publications (Stanford 2025), with universities like IITs/OUTR driving healthcare AI. Collaborations like this enhance NIRF rankings, attract scholarships and postdoc positions. Students can explore academic CV tips for such fields.

Govt initiatives like RIE2030 echo Singapore's, but India's focus on affordable AI positions it globally.

Photo by Fotos on Unsplash

Looking Ahead: Transforming Lives Through University-Led Innovation

This OUTR-PSAU model heralds a future where AI from Indian classrooms saves lives. Explore Rate My Professor for mentors, higher ed jobs, career advice, or university jobs. Share your thoughts below and join the conversation on advancing research.