使用詮釋資料轉譯影像特徵之自動化醫學影像報告

隨著現代醫學借助醫學影像協助臨床醫生進行治療評估，撰寫醫學影像報告的需求隨之增加，醫生工作負荷量超載。病患轉診時，無規範格式的醫學影像報告，也會因為每個人書寫的習慣不同而難以作為溝通工具。再者，大多數的報告缺乏考量非醫學專業患者的需求，患者可能無法全盤理解報告內容，且報告產生往往需要漫長的等待，延遲獲取醫學影像報告可能會破壞患者參與度或醫病關係。因此，本研究從醫學數位影像及通訊(digital imaging and communications in medicine, DICOM)中自動擷取姓名、病歷號、性別、年齡、出生日期、檢查日期、檢查時間、檢查地點、檢查項目及檢查儀器等DICOM tag，依不同醫學報告需求參照報告系統(reporting and data systems, RADS)建立的結構化術語描述醫學影像的資訊，並結合腫瘤位置、腫瘤大小、腫瘤形狀、腫瘤方位、腫瘤邊緣、腫瘤內部型態、腫瘤後方呈現等人工智慧輔助診斷的影像特徵，以及影像特徵在臨床上可能對應的診斷，再由Panofsky-Shatford 圖像分析理論剖析以及Jörgensen歸納的屬性進行資訊整合，設計一款醫學影像詮釋資料模板。最終將所得資訊填入系統預先定義的醫學影像詮釋資料模板中，自動產生一份完整的醫學影像報告。醫學影像詮釋資料模板分為醫生版本及一般版本以供不同背景的使用者使用，希冀能即時協助醫生進行診斷，減去撰寫報告的時間，降低醫生工作壓力，亦有望透過清晰詳細的醫學影像報告，避免病患或其家屬不了解病況，維持醫生和患者之間的信任關係，並改善醫病之間互動與溝通。此外，為評估醫學影像詮釋資料模板報告是否能有效協助一般大眾更了解醫學影像報告，於社群平台上發放問卷，總共回收100份問卷。經探索性因素分析法分析問卷，構面因素負荷量皆大於0.5，具有建構效度，且整體問卷Cronbach's Alpha為0.916，亦具有信度。問卷分析結果顯示，不論任何性別、年齡或學歷，對於醫學影像詮釋資料模板報告態度皆無顯著的差異，都認同醫學影像詮釋資料模板報告能協助理解醫學影像的內容。

關鍵字

詮釋資料；影像特徵；自動化醫學影像報告

並列摘要

As modern medicine relies on medical imaging to assist clinicians in evaluating treatment, the need to write medical imaging reports increases, and physicians are overloaded with workloads. When patients are referred, medical imaging reports in non-standard formats will also be difficult to use as a communication tool due to the different writing habits of each person. Furthermore, most reports do not take into account the needs of non-medical patients, patients may not fully understand the content of the report, and the generation of medical imaging reports often requires to wait a long time, and thus delaying access to medical imaging reports may disrupt patient participation or the relationship between doctors and patients. Therefore, this study automatically extracts DICOM tags from digital imaging and communications in medicine (DICOM), such as patient's name, ID, sex, age, birth date, study date, study time, institution name, study description, modality, etc. According to different medical reporting requirements, this study refers to the structured terms established by reporting and data systems (RADS) to describe the information of medical images, and combine the image features of AI-assisted diagnosis such as tumor location, size, shape, orientation, margin, echo pattern, posterior features, etc., as well as the possible clinical diagnosis of the imaging features. Then, the information is integrated by the theoretical analysis of Panofsky-Shatford image analysis and the attributes summarized by Jörgensen to design a medical imaging metadata template. Finally, the obtained information is filled into the medical imaging metadata template predefined by the system, and a complete medical imaging report is automatically generated. Medical imaging metadata template are presented as physician version and general version for users of different backgrounds. It is hoped that it can immediately assist the doctors in the diagnosis, reduce the time for writing the report, and alleviate the pressure of the doctor. It is also hoped that the clear and detailed medical imaging report can be used to prevent patients or their families from ignoring the condition, maintain the trust relationship between doctors and patients, and improve the interaction and communication between doctors and patients. In addition, in order to evaluate whether the medical imaging metadata template report can effectively help the general public better understand the medical imaging report, a questionnaire was distributed on the community platform, and a total of 100 questionnaires were collected. The questionnaire was analyzed by exploratory factor analysis: the factor loadings are all greater than 0.5, which has construct validity, and the overall questionnaire Cronbach's Alpha is 0.916, which also has reliability. The results of questionnaire analysis showed that regardless of gender, age or educational background, there is nonsignificant difference in attitude towards the medical imaging metadata template report, and all agreed that it can help understand the content of medical images.

並列關鍵字

Metadata ； Image features ； Automated medical imaging reports

參考文獻

Al-Haj, A. (2015). Providing integrity, authenticity, and confidentiality for header and pixel data of DICOM images. J Digit Imaging, 28(2), 179-187. https://doi.org/10.1007/s10278-014-9734-8

Google Scholar

Amin, M. B., Greene, F. L., Edge, S. B., Compton, C. C., Gershenwald, J. E., Brookland, R. K., Meyer, L., Gress, D. M., Byrd, D. R., & Winchester, D. P. (2017). The Eighth Edition AJCC Cancer Staging Manual: Continuing to build a bridge from a population-based to a more "personalized" approach to cancer staging. CA Cancer J Clin, 67(2), 93-99. https://doi.org/10.3322/caac.21388

Google Scholar

An, J. Y., Unsdorfer, K. M. L., & Weinreb, J. C. (2019). BI-RADS, C-RADS, CAD-RADS, LI-RADS, Lung-RADS, NI-RADS, O-RADS, PI-RADS, TI-RADS: Reporting and Data Systems. RadioGraphics, 39(5), 1435-1436. https://doi.org/10.1148/rg.2019190087

Google Scholar

Ariño, A. (2003). Measures of strategic alliance performance: an analysis of construct validity. Journal of International Business Studies, 34(1), 66-79. https://doi.org/10.1057/palgrave.jibs.8400005

Google Scholar

Bernardi, R., Cakici, R., Elliott, D., Erdem, A., Erdem, E., Ikizler-Cinbis, N., Keller, F., Muscat, A., & Plank, B. (2016). Automatic description generation from images: A survey of models, datasets, and evaluation measures. Journal of Artificial Intelligence Research, 55, 409-442.

Google Scholar

主題瀏覽