BradykinesiaCam — AI-Powered Parkinson's Disease Finger Tap Assessment System

BradykinesiaCam — Computer Vision System for Parkinson's Bradykinesia Assessment

BradykinesiaCam is a Django-based clinical decision support platform that uses computer vision and signal processing to assess bradykinesia — the slowness of movement that is one of the cardinal symptoms of Parkinson's Disease. By analyzing short finger-tapping videos recorded on any standard camera, the system produces UPDRS Part III (item 3.4) aligned scores covering Speed, Amplitude, and Rhythm, and delivers a composite severity rating that classifies whether the subject is a Healthy Control or a Parkinson's Disease patient. For PD patients, it further determines medication state as ON, OFF, or UNCERTAIN based on motor kinematic data.

This project is an excellent choice for final year students in Computer Science, Biomedical Engineering, and Artificial Intelligence programs who want to build a real-world clinical AI application. If you are looking for final year projects with source code that stand out in viva and dissertation evaluations, BradykinesiaCam delivers both technical depth and healthcare relevance in a single package.

What is Bradykinesia and Why Does It Matter

Bradykinesia refers to the progressive slowness, reduced amplitude, and irregular rhythm of voluntary movements observed in Parkinson's Disease. Clinicians typically assess it using the MDS-UPDRS Part III scale, which relies on manual observation — a process that is time-consuming, subjective, and inconsistent across examiners. BradykinesiaCam automates this assessment using a video-based pipeline that extracts precise kinematic measurements from finger-tapping tasks, achieving a Pearson correlation of 0.740 with clinical UPDRS scores.

Core Features

• MediaPipe Hand Landmark Extraction: Each video frame is processed through MediaPipe Hands' 21-point skeleton model. The system tracks the Euclidean distance between thumb tip (landmark 4) and index tip (landmark 8) across all frames to compute the tap aperture series at the native recording frame rate.
• SciPy Peak Detection for Tap Events: The aperture signal is inverted so finger closures become peaks, and scipy.signal.find_peaks isolates individual tap cycles. This produces accurate tap timestamps even with minor hand tremor or lighting variation.
• UPDRS-Aligned Scoring Engine: Three independent scores are computed — Speed (taps per second), Amplitude (maximum aperture normalized to hand span), and Rhythm (coefficient of variation of inter-tap intervals). A weighted composite score (0.35 Speed + 0.35 Amplitude + 0.30 Rhythm) maps directly to the 0–4 UPDRS severity scale.
• Disease State Classification: A composite score of 1.5 or above predicts Parkinson's Disease; below 1.5 predicts Healthy Control. The model reaches 91% sensitivity and 97% specificity on the validation set.
• Medication State Detection: For patients classified as PD, the system further determines whether they are in a medication ON state (composite score of 1.5 or below), OFF state (above 2.5), or an UNCERTAIN window in between.
• Asynchronous Video Processing with Celery and Redis: Uploaded videos are queued as Celery tasks, allowing the web interface to remain responsive during computationally intensive CV processing. A graceful fallback to synchronous processing is available for development environments without Redis.
• PDF Clinical Report Generation: Each analysis session produces a downloadable PDF report built with ReportLab, covering all kinematic metrics, UPDRS scores, classification outcomes, and recording metadata.
• Patient Management Dashboard: Clinicians or researchers can manage multiple patients, upload video sessions per patient, and track longitudinal assessment history through a clean Tailwind CSS and Alpine.js interface.
• Demo Fixture Data: Three pre-computed patient records are included as Django fixtures, making it straightforward to explore the full UI without recording videos during initial setup.
• PostgreSQL-Ready Architecture: The project ships with SQLite for local development and is pre-configured for a PostgreSQL switch via DATABASE_URL in the environment file, making it deployment-ready on Railway, Render, or a custom VPS.

Technology Stack

• Backend: Django 4.2, Django REST Framework
• Async Processing: Celery, Redis
• Computer Vision: MediaPipe Hands 0.10, OpenCV
• Signal Processing: SciPy, NumPy
• PDF Reports: ReportLab
• Frontend: Tailwind CSS, Alpine.js, Chart.js
• Database: SQLite (development), PostgreSQL (production)

Applications and Use Cases

• Telemedicine Screening: Patients can record finger-tap videos at home and upload them for remote neurological assessment, reducing the need for in-clinic visits for routine monitoring.
• Medication Response Tracking: The ON/OFF medication state classifier provides an objective tool for tracking how well a patient's dopaminergic medication is working across different times of the day.
• Clinical Research: Researchers studying Parkinson's motor biomarkers can use the kinematic export data and PDF reports to build longitudinal datasets without manual annotation.
• Medical AI Final Year Project: This is one of the most technically differentiated AI final year projects available on the CodeAj Marketplace, combining computer vision, signal processing, clinical scoring systems, and asynchronous web architecture in a single coherent codebase.
• Hospital Prototype Deployment: With its PostgreSQL-ready backend, Nginx media serving guide, and Celery process manager support, the project can be adapted into a working hospital prototype for neurological outpatient departments.

Performance Benchmarks

• Sensitivity: 91%
• Specificity: 97%
• UPDRS Pearson Correlation: r = 0.740

What You Get with This Project

When you purchase BradykinesiaCam from the CodeAj Marketplace, you receive the complete Django source code with all CV engine modules, pre-configured settings, demo fixture data, and a detailed setup guide. Our team also offers project setup sessions with full source code explanation, custom project reports, IEEE-format research papers, and presentation slides as add-on services. Students looking for Python final year projects with source code and strong academic documentation support will find everything they need in one place.

Add-On Services Available

• Idea Implementation — Custom project creation tailored to your college requirements
• Project Setup and Source Code Explanation sessions via video call
• Custom project report, IEEE research paper, and presentation slides

BradykinesiaCam is a clinical decision support tool. All results must be reviewed and interpreted by a qualified healthcare professional before any clinical use.