Multi-Modal Deep Learning

Knee Osteoarthritis Prediction and Progression

A unified decision-support system integrating medical imaging, clinical biomarkers, and wearable IoT sensors for early detection and long-term monitoring of Knee Osteoarthritis.

KneeCare Technology
Vibration Analysis
AI-Driven Grading

Research Domain

Bridging clinical assessment, medical imaging, and continuous patient monitoring through advanced AI.

Literature Survey

Analysis of AI in KOA diagnosis using clinical databases, blood biomarkers, and CNN-based radiographic assessment (Kellgren-Lawrence grading).

Research Gap

Conventional methods rely on single-modality info and lack continuous monitoring, making real-life progression tracking difficult.

Research Problem

High cost of MRI/X-ray and impracticality of frequent hospital visits for elderly patients requiring long-term monitoring.

Research Objectives

Developing low-cost, explainable monitoring systems using IoT sensors and multimodal data fusion for clinical decision support.

Methodology

Phased approach: Clinical prediction (XGBoost), Binary Classification (YOLOv8), Severity Grading (EfficientNetB0), and IoT Wearable Monitoring.

Technologies Used

Python, TensorFlow, YOLOv8, XGBoost, ESP32, MPU6050 (Vibration), and MLX90614 (Thermal) sensors.

Project Milestones

Following our journey from the initial proposal to the final assessment. Each milestone represents a key phase in our research, design, and development process.

January 2026
Milestone 1

Project Proposal

Initial scope definition and hypothesis creation. Secured ethical approval from the National Hospital of Sri Lanka for patient data anonymization and research.

Marks: 10%
April 2026
Milestone 2

Progress Presentation 1

Focus on binary classification models (YOLOv8) and literature survey completion. Initial preprocessing of X-ray and MRI datasets from hospital archives.

Marks: 15%
July 2026
Milestone 3

Progress Presentation 2

Implementation of severity grading using EfficientNetB0 and late-fusion of clinical biomarkers via XGBoost. Integration of IoT sensor calibration for vibration analysis.

Marks: 15%
October 2026
Milestone 4

Final Assessment

Comprehensive evaluation of the unified multimodal decision-support system. Final thesis submission and performance benchmarking against KL grading standards.

Marks: 40%
November 2026
Milestone 5

Viva & Demonstration

Live demonstration of the wearable IoT subsystem and the web-based doctor monitoring portal. Formal oral examination by the clinical board.

Marks: 20%

Project Documentation

Formal reports and research papers produced during the study.

Project Charter

Project governance and scope management.

Download

Proposal Document

Detailed research methodology and hypothesis.

Download

Check List Documents

Requirement traceability and ethical compliance.

Download

Final Document

Complete thesis with experimental results.

Download

Past Presentations

Visual slides for each evaluation phase.

Proposal Presentation

View Slides

Progress Presentation-1

View Slides

Progress Presentation-2

View Slides

Final Presentation

View Slides

Research Team

The innovators behind the KneeCare platform.

Supervisor

Jenny Krishara

Supervisor

Co-supervisor

Thamali Dassanayake

Co-supervisor

Fernando W.D.A

Fernando W.D.A

Data Scientist

IT22223708

Jayasinghe J.M.N.S.

Jayasinghe J.M.N.S.

IoT Developer

IT22582942​

Perera B.B.A.R

Perera B.B.A.R

Deep Learning Specialist

IT22606792​

Gamage D.M.G.P.K

Gamage D.M.G.P.K

Lead Researcher

IT22188472​

Contact Us

For research collaborations or inquiries regarding the KneeCare system.

+94 11 234 5678
info@kneecare.edu
National Hospital, Colombo, Sri Lanka