查詢結果分析
來源資料
相關文獻
- 醫學影像深度學習鑑別機制之可解釋研究
- A Maximum Likelihood Estimator and Fractal Analysis of Medical Images
- 心血管醫學影像之介紹及應用
- 腦部手術用規劃及導引系統
- A Fourier-Domain Maximum Likelihood Estimator and Its Application to Fractal Analysis of Medical Images
- 海洋性貧血之骨髓外造血之醫學影像
- 三維立體醫學影像之發展現況
- 醫療單位採用醫學影像儲傳系統關鍵因素之研究--以臺中榮民總醫院為例
- 以全球資訊網為基礎的網路門診系統
- 立體影像重建系統
頁籤選單縮合
| 題 名 | 醫學影像深度學習鑑別機制之可解釋研究=Explainable Research on Medical Image Deep Learning Identification Mechanism |
|---|---|
| 作 者 | 趙和昌; | 書刊名 | 資訊電子學刊 |
| 卷 期 | 12:1 2024.07[民113.07] |
| 頁 次 | 頁53-63 |
| 分類號 | 312.831 |
| 關鍵詞 | 可解釋深度學習; 醫學影像; Explainable deep learning; Medical image; Xception; LIME; |
| 語 文 | 中文(Chinese) |
| 中文摘要 | 本論文提出一份醫學影像深度學習鑑別機制之可解釋研究,以了解自動判別醫學影像驗證的決策模式。此機制包括四個單元,分別是資料庫單元、模型訓練單元、測試單元與可解釋單元。此機制採用無失真浮水印機制針對多個影像及浮水印產生出相對應的驗證所有權影像;再以隨機分割、擾動(Perturbations)處理,產生含有影像處理攻擊失真訊息的訓練資料集。在訓練單元裡使用Xception深度學習模型。測試單元使用相同的無失真浮水印機制。接著由訓練好的Xception模型鑑別醫學影像。可解釋模型採用LIME技術。研究成果顯示能成功地展現出自動判別醫學影像驗證的決策模式。 |
| 英文摘要 | This paper proposes an interpretable study of the deep learning identification mechanism for medical images, aiming to understand the decision-making model used in the automatic verification of medical images. This mechanism includes four units, which are database unit, model training unit, test unit and interpretable unit. In this mechanism, a lossless watermark mechanism is used to generate corresponding verification ownership images for multiple images and watermarks; then random segmentation and perturbations are used to generate a training data set containing image processing attack distortion information. Use the Xception deep learning model in the training unit. The test unit uses the same lossless watermarking mechanism. Then the medical images are identified by the trained Xception model. Explainable models employ LIME technology. The research results show that it can successfully demonstrate a decision-making model for automatic discrimination of medical image verification. |
本系統中英文摘要資訊取自各篇刊載內容。