| 摘要: | 背景與動機：細胞學檢查評估對於許多疾病的診斷是一種安全、有效率及完善的技術。經由普遍及大規模的篩檢可以早期發現非典型增生或癌細胞，降低子宮頸癌的發病率和死亡率。傳統的細胞學診斷利用顯微鏡觀察，以敘述性的規則進行定性的診斷，此方法可能因觀察者主觀認知的差異，產生不一致的診斷結果。最近，網路學習因具有易於存取資訊及知識、支持無所不在的學習環境、提升成本效益等優點，已經普遍應用於學校和企業之教學。
研究目的：設計一個以網絡為基礎之資訊系統，有效整合影像編輯及處理、細胞分析與分類、細胞病理訓練與測試等功能，做為線上細胞影像語教學材料準備以及細胞病理學教學之平台。
研究方法：本研究利用「混搭」（mashup）技術設計一個以網頁為基礎之資訊系統。其中自動分類器利用支援向量機(SVM)進行設計，分類不同型態之子宮頸細胞以及辨別正常及異常細胞。網頁為基礎之細胞病理訓練及測試系統利用分類後之細胞來教育學生、住院醫師、資淺之病理科醫師，進行不同種類細胞之辨識。問卷採用改良式科技接受模型進行設計，除了認知有用性、認知易用性、行為意向等三個構面之外，也加入電腦自我效能、技術支援與訓練、資訊品質與整合、資訊安全等四個構面，進行線上細胞病理學習系統之評估。本研究招集中部醫院93位受測者進行實驗，其中包括17位細胞病理醫師、73位病理技術員、3位其它相關專業人員。在進行短期的訓練，訓練辨識不同種類細胞之技巧後，受測者被要求操作本系統及進行線上測驗，最後填寫問卷。
結果與評估：本研究設計一套雛型系統，包括顯微鏡、數位相機、個人電腦、影像編輯及分析模組、細胞分析與分類模組、線上細胞病理訓練及測驗系統，作為研究平台。研究結果顯示，子宮頸抹片細胞自動判讀模組對於分辨不同種類細胞及區別正常以及異常細胞的判斷正確性分別為94.43% 及98.6%。利用擴展科技接受模型（TAM）問卷評估93位醫師或醫檢師的結果顯示，本雛型系統結合細胞分類和線上細胞病理訓練及測驗系統，對於細胞病理診斷和教育有良好的效益。
結論: 大多數的使用者一致同意操作介面易於使用，並表示強烈的行為意向，希望在未來仍能採用該系統。本系統對於提高細胞診斷效率及促進學習成效預期有很大的貢獻。
Background and Motivation: Cytology evaluation is a safe, efficient, and well-established technique for the diagnoses of many diseases. Its ability to reduce the mortality and morbidity of cervical cancer is through mass screening to early detect dysplasia or pre-invasive cancer cells. Classical cytological diagnosis is based on microscopic observation of specialized cells and qualitative assessment using descriptive criteria, which may be inconsistent because of subjective variability of different observers. Recently, web-based learning is becoming prevalent in schools and enterprises around the world for its advantages of providing easy access to information and knowledge, supporting ubiquitous learning environment, and increasing cost-effectiveness for both educational institution and students.
Objective: The objective of this study is to design a web-based system to integrate various functions including image editing and processing, cell analysis and classificatione, and cytopathology training and testing for online cell image and material preparation and cytopathlogy education.
Material and Methods: Tthe "Mashup" technique was used to design the web-based system. The automatic classifiers were designed based on the support vector machine (SVM) to cluster four different types of cervical cells and to discriminate dysplasia from normal cells. The web-based cytopathology training and testing (WBCTT) system was developed based on the classified cell images to train students, resident physicians, and novice pathologists to discriminate various types of cervical cells. A questionnaire designed based on the modified technology acceptance model (TAM) was used to evaluate perceived usefulness (PU), perceived ease of use (PEU), and behavior intention (BI) of the web-based e-learning system. In addition to the 3 constructs proposed by Davis, 4 additional constructs, including computer self efficacy (CSE), technical support and training (TST), information quality and integration (IQI), and Information Privacy (IP), were also adopted to verify the model. A total of 93 persons, including 17 cytopathologists, 73 cytopathological technicians, and 3 other professionals whose tasks related to pathology, were recruited from hospitals located in middle Taiwan for this investigation. After a short tutorial of describing the rules for discriminating different types of cervical cells, the users were asked to operate the system, do online tests and finally fill the questionnaire following the learning and testing.
Results and Evaluation: A prototypic system consisting of a microscope, a digital camera, personal computer, image editing and processing program, cell analysis and classification modules, and web-based cytopathology training and testing system was designed and applied for the experiment. The experimental results demonstrate that the classification and diagnostic accuracy achieves 94.43% and 98.6%, respectively. System evaluation based on questionnaire survey of extended technology acceptance model (TAM) shows that the proposed system embedded with cell classifier and WBCTT is useful in cytopathology diagnosis and training.
Conlusion and suggestion: Most of the users agreed the operation interface of the proposed system is friendly and easy to use. They also expressed strong behavior intention to further adopt the system. It is expected to have significant contributions in increasing diagnostic efficiency and promoting learning efficiency. |
