Overview
In response to clinical needs, infrared multi-spectral images are combined with traditional clinical images and other multi-modal data to build a more efficient intelligent auxiliary diagnosis system and intelligent equipment for skin health and diseases, including skin lesions automatically segmentation on skin diseases images, automatically design surgical margin and planning for skin tumor surgery.
Description
Database: Relying on the preliminary foundation, build the first standardized infrared multispectral image database of skin diseases, and further integrate other medical images and medical history texts to iterate into a large multimodal skin disease database.
Model: Design a deep learning network based on multi-scale and multi-level. The collaborative attention learning network realizes the collaborative representation of multi-modal data at the feature level, builds a multi-modal skin disease auxiliary diagnosis model, and realizes breakthroughs in algorithms. Develop the segmentation network of skin lesions and model for surgery planning, including surgical margin design and navigation of intraoperative sampling.
System: Propose an artificial intelligence system combined with the real-time augmented reality to assist dignosis and surgery for skin diseases.
Equipment: Based on the self-developed high-performance system, construct and assemble infrared multi-spectral skin disease auxiliary diagnosis equipment and multifunctional device for skin tumors surgery.
Eligibility
Inclusion Criteria:
- Informed consented.
- With a diagnosis of skin disease made by at least 3 dermatologists.
- Without life-threatening risk to intervention.
- Requires surgical treatment (For devices).
Exclusion Criteria:
- Having difficulties to follow-up.
- Poor general condition.