When AI Helps You Breathe Right: UHK Researchers Develop an Application for Proper Inhalation Technique
Can Artificial Intelligence (AI) assist in treating asthma or chronic obstructive pulmonary disease (COPD)? Proper inhalation technique is critical to treatment success, yet many patients could get it wrong. Researchers here at the Faculty of Informatics and Management (FIM UHK) are developing an AI-driven video analysis prototype to identify and correct these mistakes in real time.
A Smart "Second Pair of Eyes"
Inhalation therapy may look simple at first glance, but it requires precise timing and coordination: preparing the device, exhaling fully beforehand, maintaining correct posture and seal, timing, inhalation depth and speed according to the inhaler type, and holding one’s breath afterwards. Even a high-quality medication could fail if it cannot reach deep into the airways.
To tackle this challenge, an interdisciplinary project led by Dr. Karel Mls from the Department of Information Technologies at FIM UHK is developing a web-based application utilizing computer vision to evaluate inhalation techniques.
"In this phase, it is not a commercial product or a replacement for a doctor or pharmacist. It is a research prototype testing whether computer vision and motion analysis can serve as a supportive tool in clinical practice," explains Dr. Karel Mls. "We want to use AI as a 'second pair of eyes' that helps healthcare professionals and patients spot subtle errors that might otherwise go unnoticed."
Bridging Computer Science and Healthcare
The multi-year project, supported by the Czech Health Research Council (AZV), bridges informatics, pharmacy, and pulmonology. FIM UHK collaborates closely with:
- Department of Social and Clinical Pharmacy, Faculty of Pharmacy in Hradec Králové (Charles University)
- Department of Pneumology, University Hospital Hradec Králové
While clinical partners provide expert annotations and clinical ground truth, FIM UHK's team focuses on technical architecture, user interface design, video processing, and landmark tracking (such as face, hands, and upper body tracking using tools like MediaPipe).
How It Works
- Device Selection & Recording: The user opens the application via a standard laptop, tablet, or smartphone camera and selects their inhaler type (e.g., pressurized metered-dose inhaler, dry powder inhaler, or soft mist inhaler).
- Phase Recognition: The system tracks visible visual markers such as body orientation, hand movement, positioning of the mouthpiece, and inhalation
- Actionable Feedback: Rather than providing a medical diagnosis, the tool evaluates execution against clinical guidelines, highlighting specific errors such as early inhalation, incorrect speed, or missing breath-holds.
The Road Ahead
Because clinical AI requires rigorous validation, privacy compliance, and high precision across varied real-world lighting and camera conditions, testing will proceed across multiple phases, ranging from technical functionality to expert clinical review and patient usability testing.
The ultimate vision is a digital assistant capable of guiding patients empathetically through their routines, maximizing therapy efficacy without requiring changes to medication.
