- 學位論文
電腦輔助診斷系統在甲狀腺結節的臨床應用：統合分析與臨床試驗計畫書
The Application of Ultrasound Computer-aided Diagnosis System for Thyroid Nodules: A Meta-analysis and Protocol
摘要
背景： 超音波在甲狀腺結節篩檢中被廣泛的應用，雖然特殊的超音波影像特徵如微鈣化、低迴聲和不規則邊緣通常被認為與惡性甲狀腺疾病有關，然而，超音波檢查的主要局限性在於其操作者主觀的判斷，不同醫師判讀同一張影像會有不同的判斷。 近年來AI人工智能發展迅速，電腦輔助診斷（Computer-Aided Diagnosis）系統採用人工智能技術進行超音波圖像分析已逐漸發展成熟，本研究的目的是醫師根據電腦輔助影像分析技術提供的信息，是否能更有效地進行診斷。 方法： 統合分析匯整醫師在超音波檢查診斷甲狀腺癌診斷優勢比 (Diagnostic Odds Ratio)，與醫師使用電腦輔助診斷（CAD）系統的優勢比 (DOR)，進行比較分析。 結果： 符合篩選標準的5 篇研究進行統合分析，其醫師使用 CAD 系統的診斷表現與醫師單獨的診斷表現相當，無顯著性差異，其中因三篇研究皆為同為韓國三星發展的電腦輔助診斷（CAD）系統，在不同年份發表，排除其西元2018年及西元2019年研究，將近期2020年發表的與另兩篇研究做統合分析，其結果顯是醫師使用 CAD 系統的診斷表現優於醫師單獨的診斷表現，具顯著性差異。 結論和相關性：結果推論，應用 CAD 系統將改善臨床醫生的判讀並減少診斷的變異性。然而，此推埨需要更多的研究來探索 CAD 系統在臨床應用下是否能更有效地進行甲狀腺結節診斷，且尚未有相關研究探討超音波技術師在使用CAD系統是否能提升其甲狀腺結節良惡性的診斷率。 臨床試驗： 超音波技術師使用電腦輔助診斷系統鑑別診斷甲狀腺結節的多讀者多案例研究。
並列摘要
Importance Ultrasound (US) is extensively applied in the detection of thyroid nodules. Although the particular ultrasound features such as microcalcifications, hypo-echogenicity and irregular margins are related to malignant thyroid disease. The main limitations of US are its operator dependence and interobserver variability. Objective The purpose of this study was to analyze whether physicians had higher performance diagnostic based on the information provided by computer-aided detection. Data Sources A detailed literature search on PubMed, Embase, and Cochrane Libraries for articles published until December 2020 was carried out. Study Selection Studies on the diagnostic performance of CAD systems. Data Extraction and Synthesis Clinician performance without and with the use of CAD by pooled DOR for the diagnosis of thyroid nodules, absolute numbers of true positive (TP), true negative (TN), false positive (FP), and false negative (FN) data had been presented. Meta-analysis of studies was performed using a mixed-effects framework, including standard univariate fixed and random-effects meta-analysis and meta-regression. Main Outcome(s) and Measure(s) The diagnostic performance was assessed by pooled their diagnostic odds ratio (DOR), and was compared with that of physicians. Results Five studies with 1130 thyroid nodules and 26 Physicians were included. The diagnostic performance of the physicians was comparable to that of the physicians with CAD system (DOR 7.8 [95% CI 5.5–10.9] vs. DOR 10.9 [95% CI 8.30–14.2]). Conclusions and Relevance The results demonstrated that applying a CAD system would improve clinicians' interpretations and lessen the variability in diagnosis. However, more studies are needed to explore the use of the CAD system in an actual ultrasound diagnostic situation.