肝癌病患病歷報告之辨識資訊萃取結果可信賴程度

A method for identifying confidence level of the extracted results from medical narrative reports: A case study focus on the patients with liver cancer

吳亞霖(Ya-Lin Wu)

指導教授 ： 賴飛羆

國立臺灣大學/電機資訊學院/生醫電子與資訊學研究所/碩士(2012年)

https://doi.org/10.6342/NTU.2012.00088

摘要

病歷資料擁有豐富的疾病、醫療程序和治療結果等資訊。在之前的研究裡，我們實做一資訊擷取系統提取肝癌病人文字報告裡肝癌相關資訊。資訊擷取系統提取的結果將用於建立預測肝癌復發的模型。資訊擷取後，重要的是證明這些提取結果是可靠的。但沒有經由人為檢查的方式，
在這項研究中，我們兩個團隊成員已檢查所有提取信息。根據檢查結果可得到資訊擷取系統的準確度。人為檢查所有提取結果的方式，是一個耗時耗力的工作。因此，本研究的目的在於提供一個有效率的方式去檢查提取結果。我們設計一驗證系統，用於預測每個提取結果的正確性。據驗證系統預測的結果，檢查人員可以有效地檢查那些被驗證系統預測為錯誤資訊的提取結果並且進行校正，而不需檢查所有的提取結果。透過驗證系統可以提高檢查提取結果的效率。

關鍵字

病歷資料 ； 資訊擷取系統 ； 驗證系統

並列摘要

Textual medical records constitute a rich source of information about diseases, medical procedures and treatment results. In our previous work, we implemented the information extraction (IE) system for extracting the desired information from liver cancer patients’ textual reports. These extracted results produced by IE system are used for supporting the development of recurrence predictive model. After information was extracted by the IE system, it is important to prove these extracted results are reliable. However, we are not sure about the correctness of these extracted results without checking manually by the domain experts.
In the study, two of our team members had reviewed all extracted information. According to their reviews, the precision of the IE system can be analyzed. But, checking the correctness of all extracted results manually would be a time-consuming and labor-intensive task. Therefore, the aim of this study is to provide an efficient way for facilitating the process of checking all extracted results. We designed the validation system for predicting the correctness of each extracted result. According to the prediction of the validation system, the reviewers can efficiently check the smaller part of extracted results predicted as low confidence extracted information by the validation system and correct them; instead of checking all extracted information. In this way, it can highly promote the efficiency of the future reviewing process.

並列關鍵字

Textual medical records ； Information extraction system ； Validation system

參考文獻

2. Hripcsak, G., et al., Unlocking Clinical Data from Narrative Reports: A Study of Natural Language Processing. Annals of Internal Medicine, 1995. 122(9): p. 681-688.

3. Hripcsak, G., et al., Use of Natural Language Processing to Translate Clinical Information from a Database of 889,921 Chest Radiographic Reports1. Radiology, 2002. 224(1): p. 157-163.

4. Chapman, W.W. and K.B. Cohen, Current issues in biomedical text mining and natural language processing. Journal of Biomedical Informatics, 2009. 42(5): p. 757-759.

5. Cohen, A.M. and W.R. Hersh, A survey of current work in biomedical text mining. Briefings in Bioinformatics, 2005. 6(1): p. 57-71.

11. Wimalasuriya, D. and D. Dou, Ontology-based information extraction: An introduction and a survey of current approaches. Journal of Information Science, 2010. 36(3): p. 306-323.