Region-based Annotations for the Medical Images

Abstract

Annotations facilitate the problems of visual medical information management and data mining. Most of the semantic annotations are targeted for the on-line repositories, and detached from the image data. Instead of using detached annotations, we embed them directly into the reasonably chosen area of the image and use spatial identifiers to bind them with the corresponding regions. The embedding method deals with the different quality requirements of the regions. Two-dimensional QR codes are used as the physical carrier of the annotation, providing the industrial grade for information detection, resilience and error correction capabilities. The annotation is distributed in a series of QR codes for efficient use of the available space. Embedding of annotations into the clinically irrelevant regions of the medical image shows preferable results in the field of informational capacity. The quality of annotated and compressed image is comparable to the image quality loss due to image compression.