醫療影像系統(PACS)使用Hadoop儲存方法之研究

黃仲平 ； 羅壽之

《TANET2016 臺灣網際網路研討會》 2016 (2016/11) Pp. 662-667-119

https://doi.org/10.6679/TANET.2016.119

摘要

PACS 是現代醫學發展而生成的醫學影像檔案儲存及傳輸的系統，但是隨著醫學影像成像技術的快速發展與成像清晰度的大幅度提高，使得醫療影像的儲存成為大數據的問題。Hadoop HDFS的分散式儲存方式使用廉價的通用硬體進行橫向擴展，解決了傳統儲存的天花板效應。HDFS 和PACS 的結合在技術和成本上具有很大的誘因，本文從HDFS 儲存系統不能儲存小檔案的缺點，基於SequenceFile 檔案格式設計了一套符合實際應用場景的混合儲存架構。這套系統包含HDFS、SequenceFile 格式的H-DICOM 檔案和NFS 傳統儲存裝置。核心是HFHSC（H-DICOM File Hadoop SequenceFile Converter），主要用於DICOM 圖像和H-DICOM 格式的相互轉換。測試結果表明此架構可以滿足大量醫學影像資料的快速存取和處理需求。

關鍵字

醫學影像 ； 分散式檔案系統 ； Hadoop ； PACS ； DICOM

並列摘要

PACS is the basis of archival storage and transmission of medical image files under the modern development of medicine. With the rapid development of medical imaging technology, image resolutions get largely improved. These make the storage of medical image files a big data problem. Hadoop HDFS which can be expanded horizontally by using inexpensive generic hardware solves the ceiling effect caused by traditional storage devices. The combination of HDFS and PACS is attractive in technology and cost. In this paper, we start from the drawback that HDFS cannot efficiently manage small files, and then design a hybrid storage system architecture based on the format of Hadoop SequenceFile. This new architecture contains HDFS, H-DICOM format and traditional NFS storage. The core component is HFHSC (HDICOM File Hadoop SequenceFile Converter) that is used to translate DICOM images to H-DICOM formats. The test results indicate that this architecture can meet the needs of fast access and can process a vast amount of medical imaging data.

並列關鍵字

Medical Image ； Distributed File System ； Hadoop ； PACS ； DICOM