Applied research · Digital health · Seeking funding
AI-Powered Non-Invasive Disease Detection for Bangladesh
An AI system for early disease screening from non-invasive signals — medical images, voice, cough, vitals, and wearables — with physician oversight and low-cost community access.
- Healthcare AI
- Medical imaging
- Seeking funding
An AI system for early disease screening from non-invasive signals — medical images, voice, cough, vitals, and wearables — with physician oversight and low-cost community access.
Non-invasive screening pipeline
- Step 1
Capture signals
Collect images, cough/voice, breathing, skin features, vitals, and wearable readings via phone or clinic devices.
- Step 2
AI risk screen
Multimodal models produce preliminary risk scores with calibrated uncertainty — not final diagnoses.
- Step 3
Physician referral
High-risk cases route to qualified doctors through community clinics or telemedicine with consent and audit trails.
Bangladesh context
Bangladesh context
Noncommunicable diseases account for nearly 71% of deaths in Bangladesh, making early detection of cardiovascular disease, diabetes, chronic respiratory illness, kidney disease, and cancer a national priority. The WHO Bangladesh notes growing efforts to integrate NCD and mental-health care into emergency preparedness and primary services. Many rural and peri-urban communities still face long travel times, limited specialist access, and uneven diagnostic infrastructure — while smartphone penetration and community-clinic networks create new channels for low-cost, non-invasive screening at scale.
Research objectives
- Build multimodal AI models that fuse medical images, audio (voice/cough/breath), vitals, and wearable signals for early risk screening.
- Prioritize high-burden NCDs relevant to Bangladeshi populations and care pathways.
- Enable low-cost deployment via smartphones, portable devices, community clinics, and telemedicine workflows.
- Ensure physician-in-the-loop review, informed consent, privacy, fairness across demographic groups, and regulatory alignment.
Methodology
- 01
Establish ethical data-collection protocols with anonymization, consent, and institutional review for urban and rural cohorts.
- 02
Curate and label multimodal datasets for target conditions; benchmark baseline and fusion models.
- 03
Develop risk-scoring pipelines with calibrated uncertainty and explainability for clinical review.
- 04
Pilot screening workflows in community-clinic and telemedicine settings with qualified physician referral paths.
- 05
Evaluate sensitivity, specificity, equity across subgroups, usability, and cost per screen in field pilots.
Expected outputs
- A prototype non-invasive screening platform for selected high-burden conditions.
- Multimodal model benchmarks and validation reports on Bangladeshi cohorts.
- Mobile and clinic-facing interfaces for capture, risk display, and physician referral.
- Governance documentation for consent, privacy, fairness, and healthcare compliance.
Seeking funding
MIRAI Lab is seeking funding to build the multimodal screening prototype, collect ethically governed validation data, and run community-clinic pilots with licensed medical partners.
Overview
This project will develop an AI-powered system for the early detection of diseases using non-invasive data such as medical images, voice, cough and breathing patterns, facial or skin features, vital signs, and wearable-sensor readings. It will initially focus on high-burden conditions in Bangladesh, including cardiovascular disease, diabetes, chronic respiratory disease, kidney disease, and selected cancers.
The system will be trained and clinically validated using ethically collected, anonymized data from diverse Bangladeshi populations, including urban and rural patients. It will support low-cost screening through smartphones, portable medical devices, community clinics, and telemedicine services. AI findings will serve as preliminary risk assessments—not final diagnoses—and high-risk patients will be referred to qualified physicians. The project aims to expand early screening in underserved areas while maintaining patient consent, privacy, fairness, clinical oversight, and compliance with Bangladesh's healthcare regulations.
Team
Project lead: MIRAI Lab
Publications and datasets
Research papers, models, and open datasets will be shared here as they are released.
Project updates
Milestones, fieldwork notes, and lab announcements will appear on our notices page.
View notices →