A team led by Dr. Chang Che-Wei at Far Eastern Memorial Hospital has developed an AI-based burn diagnosis platform and software that leverages computer vision and deep learning technologies. This innovative system can quickly and accurately calculate the burn area and provide immediate treatment recommendations, significantly improving the efficiency of burn diagnosis and treatment.

Traditionally, healthcare professionals rely on visual estimation to assess burn areas. However, studies show that inter-physician variability in burn area assessment can be as high as 20%, leading to potential errors in treatment plans. These errors may result in miscalculated fluid resuscitation, delayed treatment, or worse patient outcomes.

AI Delivers Precise and Objective Data

To overcome the limitations of traditional methods, the platform integrates deep learning technology with a vast database of burn images. This database includes individuals of different ages, genders, skin tones, and body types, as well as burns of varying depths, sizes, and shapes. The images were collected under diverse lighting conditions and from multiple angles, allowing the AI model to recognize even the most complex burn scenarios with high precision.

The system comprises multiple AI models that automatically identify wound boundaries. Within these boundaries, the AI segments burn depths and compares them against a reference image of the patient’s palm. Since the palm's surface area accounts for approximately 0.5% of total body surface area, the system can rapidly calculate the percentage of burns relative to the total body area and the proportion of deep burns within the total burn area. This provides physicians with intuitive diagnostic data, enabling faster assessment of burn severity and the development of tailored treatment plans.

Key Advantages of the Platform:

• High Accuracy: The AI model, trained on a massive dataset, achieves an accuracy rate of 99.22% in automatically identifying and quantifying burn areas.
• Rapid Diagnosis: Using only a smartphone or tablet, medical personnel can capture photos of the wounds, and the system quickly provides precise diagnostic results, accelerating clinical decision-making.

Expanding Medical Applications for a Comprehensive Healthcare System

Dr. Chang highlighted that the latest version of the platform incorporates LiDAR technology (via the Burn Evaluation Network in iOS), enabling mobile devices to automatically calculate 3D burn areas. This significantly enhances accuracy, particularly for extensive burns.

The platform's applications span multiple scenarios, including emergency triage in wartime, precision treatment within hospitals, and remote telemedicine augmented by AR technology. Moving forward, the research team plans to further refine the system by integrating it with automated fluid resuscitation systems. Their ultimate goal is to provide a more comprehensive healthcare service for burn patients worldwide.

Resource: 現場緊急檢傷！行動裝置拍照AI辨識速判燒燙傷面積