查詢結果分析
來源資料
相關文獻
- Content-Based Mammogram Retrieval: Support Vector Machines to Learning User's Relevance Feedback
- 乳腺癌治驗
- 99mTc標記抗粘蛋白McAb乳腺癌放射免疫顯像初步臨床研究
- Investigations of Breast Masses with Technetium-99m Labeled Anti-mucin Antibody and Technetium-99m-MIBI
- 乳房X光攝影微鈣化群之電腦三度空間定位
- 兩點不露﹖--乳房X光攝影檢查
- 乳房X光攝影技術之品質控制與保證
- 乳房X光攝影在乳癌篩檢的利與弊
- 數位化乳房醫學影像資訊處理之探討
- 談因擾婦女朋友疾病--乳腺癌
頁籤選單縮合
題名 | Content-Based Mammogram Retrieval: Support Vector Machines to Learning User's Relevance Feedback=乳房X光片檢索:使用支援向量機方法學習檢索者的相關性回饋 |
---|---|
作者姓名(中文) | 魏嘉宏; 林孟文; 鄭培成; | 書刊名 | 清雲學報 |
卷期 | 31:2 2011.04[民100.04] |
頁次 | 頁103-121 |
分類號 | 410.1644 |
關鍵詞 | 支持向量機; 基於內容的影像檢索; 乳房X光攝影; 相關回饋; 乳腺癌; 醫學影像存檔與通信系統; Support vector machines; Content-based image retrieval; Mammography; Relevance feedback; Breast cancer; Picture archiving and communication systems; |
語文 | 英文(English) |
中文摘要 | 基於內容的影像檢索(CBIR)已經被醫學界提出來可用來解決醫學影像檢索的問題。在檢索的過程中,檢索者提供的相關回饋表示檢索者的查詢需求,並且可以用來彌補電腦可抽取的低階影像特徵與人類高階語義的鴻溝,以提高系統的檢索效能。本文提出了支援向量機(Support Vector machines)方法,這個方法 是利用分類模型來概括二元化影像檢索的問題,並進一步地估計每張影像隸屬於相關性影像的機率值。本研究欲於乳房X 光片檢索系統中運用相關回饋改善檢索結果,因此,我們首先使用檢索者提供的相關回饋作為訓練資料來建立此一分類模型,並藉由訓練過程獲得機率模型中最佳化的參數值,最後再展現出新的檢索 結果。實驗結果顯示,該方法經過五次的相關性回饋後,本影像檢索系統能將平均精確率從40%提昇至62%。 |
英文摘要 | Content-based image retrieval (CBIR) has been proposed to address the problem of image retrieval from the medical community. Relevance feedback, explaining the user’s query concept, can be used to bridge the semantic gap and improve the performance of CBIR systems. This paper proposes a learning method for relevance feedback, which utilizes probabilistic model to generalize the 2-class problem and provide an estimate of probability of class membership. To build the probabilistic model, relevance vector machine is proposed to classify the mammograms, and then scale them to the probability of class membership. Simultaneously, the optimal parameters of the probabilistic model can be obtained through the training session. Experimental results show that the proposed learning method can effectively improve the average precision rate from 40% to 62% through five iterations of relevance feedback rounds. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。