AI-Powered Neurocognitive Assessment & Stroke-Care Automation

Summary

INNERLUXES engineered an advanced neurocognitive assessment platform designed to support stroke patients through continuous monitoring, automated testing, and AI-driven evaluation. The solution uses a 3D animated model to guide patients through specific movements and commands, while integrated AI interprets facial expressions, body posture, response time, and movement accuracy in real time. Doctors and hospital staff manage assessments through a dedicated interface that captures patient details, initiates examinations, and generates final reports — enabling reliable neurological evaluation without disrupting standard patient care.

The Client

The Client is a healthcare organization led by a neurology specialist and academic, dedicated to modernizing neurological evaluations. Its goal is to automate cognitive and motor assessments for stroke patients using objective, data-driven methods to ensure timely detection of neurological decline.

The Approach

INNERLUXES implemented a structured, scalable approach to build a dual-interface system that connects clinicians and patients in real time. Core needs included:

• Real-time neurocognitive assessment at a patient's bedside.
• A 3D interactive model in the patient's native language (English, Gujarati, Spanish).
• Capturing and interpreting facial expressions, body movements, and verbal responses.
• Reducing nurses' and clinicians' manual workload.
• Ensuring accuracy, consistency, and minimal setup time.

The Solution

INNERLUXES developed a synchronized two-screen application:

1. Patient-side module

• A 3D animated model instructs the patient to perform specific movements.
• Commands are delivered visually and in the patient's chosen language.
• A camera tracks the patient's actions and expressions in real time.

2. Doctor/staff-side application

• Secure login with access to patient demographic and provider details.
• Ability to select patients and initiate examinations.
• Real-time visibility into patient responses.
• Automated scoring using facial and body landmark data.
• Final report generation for each completed exam.

The AI-driven assessment workflow ensures precision, minimal disruption to care, and continuous, reliable data collection.

The Impact

• AI-driven analysis reduced assessment time by 60–70% and improved scoring reliability through pose detection, facial-landmark mapping, and LSTM-based interpretation.
• The multilingual 3D model increased patient responsiveness by up to 40%.
• Clinicians gained clear, data-driven insights into behavioral changes, movement accuracy, and exam progression.
• The system reduced manual nurse involvement, minimized errors, and supported faster, repeatable assessments.
• Organized digital records support long-term continuity of care.

Technologies and Tools

Python 3.8.10, MS SQL 2014, OpenCV (video processing, frame extraction), MediaPipe (face & body landmark detection), PoseNet (body joint detection), LSTM neural networks (movement sequence assessment), speech recognition (audio-to-text), Windows-based MVC architecture, VS Code & Jupyter Notebook.