Active research streams.
From portable clinics and EHR digitization to explainable AI, diagnostics, speech, and digital twins — research designed to leave the lab and reach the field.
Ten projects, one mission.
Every stream serves the same goal: clinical-grade healthcare for communities beyond the reach of conventional infrastructure.
SocialTech Lab develops Portable Health Clinic system technologies
Portable health clinic systems have the potential to revolutionize healthcare delivery in remote and underserved areas of the world. At the forefront of this innovation is the Social Tech Lab, which is developing core technology for portable health clinics that can be easily transported and set up in any location. These clinics are being piloted in various countries in Asia and Africa, including Bangladesh, India, Thailand, Pakistan, Nepal, Cambodia, Malaysia, Indonesia, and Liberia. By bringing healthcare services to communities that lack access to traditional medical facilities, portable health clinics can significantly improve health outcomes and quality of life for people in these regions. The Social Tech Lab’s commitment to developing and piloting this technology in multiple countries demonstrates its dedication to improving global health equity and providing healthcare access to all.
Transformation of Healthcare Data
This research focuses on converting analog health records into standardized and digitized formats to create unified Electronic Health Records (EHR). By developing scalable methods to digitize handwritten or paper-based medical data, this work aims to enable easier storage, retrieval, and analysis of patient information.
Visualization of Healthcare Data
This research explores effective methods to visually represent healthcare data, enabling medical professionals and policymakers to quickly interpret complex health trends and make informed decisions for early intervention and improved care delivery.
Personalization of Triage
This study uses explainable AI to analyze health checkup data from the Portable Health Clinic system, aiming to identify individual risk factors and support personalized triage decisions based on each patient's unique health profile. It seeks to make AI-driven decisions more transparent and reliable.
Visualization of Healthcare Data
Due to the ongoing digital transformation in healthcare, there is a shift towards digitizing healthcare data. However, digitalization alone is not sufficient for doctors to efficiently check patient history and make clinical decisions. Our research focuses on developing smart and efficient healthcare data visualization techniques to improve the performance of healthcare service delivery.
Personalization of Triage
Machine learning aids decision-making, but models are often opaque. Explainable AI aims to interpret
NCD Risk Prediction
models and identify factors influencing decisions. This study analyzes health checkup data from a digital healthcare system called Portable Health Clinic. The research argues for personalized triage using explainable AI.
Healthcare Information From Speech
In recent years, the prevalence of diabetes worldwide has been increasing significantly, and it is estimated that the number of diabetes patients will rise to 783 million (with a prevalence rate of 12.2% among individuals aged 20-79) by the year 2045. Diabetes risks complications and has far-reaching implications for society, finance, and healthcare.
Speaker Diarization for Clinical Conversations
This research aims to develop a speech-based system that extracts structured healthcare information from doctor-patient conversations, focusing on challenges unique to low-resource languages. It investigates linguistics and technological barriers, designs, and end-to-end extraction architecture, and predictive clinical support, and new healthcare-specific linguistic resources to enhance accuracy in real clinical practice.
Digitization of Analog Medical Report
This research focuses on speaker diarization system tailored for doctor-patient conversations, especially with low-resource and accent-diverse clinical audio settings. It investigates issues such as overlapping speech, evaluation metrics and audio dynamics. This work designs an ASR enabled Diarization Pipeline that identifies speakers and their speech segments for downstream clinical information extraction, enabling EHR enrichment with conversational speech resources for healthcare applications.
Non-Invasive AI-Driven Diagnostics
Medical reports (Hematology, Clinical Chemistry, Microbiology, etc) are provided as paper documents in developing countries and stored as scanned images in EHRs in developed nations, hindering analysis. This research employs vision language models to classify, structure, and standardize scanned medical reports for integration into the Portable Health Clinic system effectively.
Cognitive-AI Healthcare Interfaces
The research focuses on non-invasive platelet count estimation using optical sensing and machine-learning model. It is integrating biomedical signal acquisition, AI-based feature extraction, and clinical validation to create affordable, portable diagnostics tools. The goal is enabling accessible, painless hematological monitoring in diverse settings, improving early detection and expanding healthcare reach.
Predictive Digital Twins
The research analysis cognitive-load factors in healthcare interfaces and encodes validated HCI principles into generative AI design systems. Through context engineering, prompt optimization, and experimental evaluation, it aims to produce adaptive clinical UIs that significantly reduce medical input errors, improve practitioner efficiency, and support safer, more intuitive digital healthcare environments.
Contextual Explainable Prescribing Research
The research develops a predictive Human Digital Twin framework to combat NCDs in youth. By fusing clinical, wearable, and lifestyle data with AI, it creates a personalized 3D avatar visualizing future health risks. This interactive "Future Me" empowers users with actionable insights, shifting healthcare from reactive treatment to proactive prevention.
Development of Revised ResNet-50 for Diabetic Retinopathy Detection
This research builds explainable prescription recommendation systems for patients with polypharmacy by learning longitudinal patient-context representations from structured records and clinical text. It retrieves clinically similar cohorts, analyzes risks such as adverse drug events, and generates transparent, case-based medication suggestions that clinicians can inspect, trust, and adapt in practice.
The 6th SocialTech Summit and Conference on Healthcare, SDGs, and Social Business
This research focuses on enhancing Diabetic Retinopathy detection by improving the ResNet-50 architecture with attention modules, refined small-lesion feature extraction, class-imbalance-aware loss functions, and medical-specific data augmentation, aiming to increase sensitivity and efficiency for real-world clinical deployment in diverse healthcare screening and diagnostic settings worldwide.deployment.
Biography:
The research develops a predictive Human Digital Twin framework to combat NCDs in youth. By fusing clinical, wearable, and lifestyle data with AI, create a personalized 3D avatar visualizing future health risks. This interactive "Future Me" empowers users with actionable insights, shifting healthcare from reactive treatment to proactive prevention.
View More
SocialTech Lab develops digital healthcare innovations-AI, digital twins, lifelong medical records, EHR standardization, and health risk prediction-to achieve universal health coverage and sustainable, data-driven healthcare systems globally.