在健康醫學網格上建立以內容為基礎的醫學影像擷取系統

Abstract

本研究基於臺北醫學大學超過 700,000 張的內視鏡及超音波影像互動式臨床影像實驗室的影像資料庫，結合開放原始碼的之以內容為基礎的影像擷取系統(GNU Image Finding Tool, GIFT)，並輔以網格技術 (Grid Technology)，利用臺灣的網際網路，在臺北醫學大學及其附設醫院、國立台灣大學之間建立起一個測試性的健康醫學網格。在此基礎之下，建置一個可用的以內容為基礎的醫學影像擷取系統。

With the increasing usage of the Internet and multimedia, a large amount of digital images are produced in the digital world right now. They are produced in many different domains, for example entertainment, commerce, education and biomedicine. The increasing usage of digital equipment in the hospital, such as CT, MRI, ultrasound and endoscope come with DICOM image formats, also produce lots of medical images per day. The volume of digital images archive is growing rapidly. Therefore, a good image retrieval system can help people to find out the images they want effectively. When it comes to medical filed, a good CBIR system of medical image can help medical education, research and even on diagnostics support. Building a CBIR system is never an easy work, especially in medical images. Thanks for the GIFT (GNU Image Finding Tool) project which already had a great CBIR system and make it open source software. It`s a nice CBIR system, which is free and open source. We use it to build a CBIR system, and try to apply it to our Clinical Interactive Image Bank of Taipei Medical University Hospital. The result is the medGIFT system. The Clinical Interactive Image Bank focus on the ultrasound images and endoscope images. However, the images inside the Image Bank are too big to fit into a single CBIR GIFT system. Therefore, we try to apply the Grid Computing technology to improve the speed of the system. The result of our experiment is fair good and the grid technology does improve the performance of medGIFT system. It made a good example of applying open source software on medical usage. The medGridGIFT system can support Evidence Based Medicine, Case Based Reasoning, and medical education to find similar case and images.