請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/84676
標題: | 自訓練於高維度標記細胞實例分割 Self-training with High-dimensional Markers for Cell Instance Segmentation |
作者: | Kuang-Cheng Lo 羅廣丞 |
指導教授: | 徐宏民(Winston H. Hsu) 徐宏民(Winston H. Hsu | whsu@ntu.edu.tw | ), |
關鍵字: | 機器學習,深度學習,自訓練,細胞分割,高維度影像, Cell Segmentation,Deep Learning,Self-training,Highly-multiplexed Imaging,CODEX, |
出版年 : | 2022 |
學位: | 碩士 |
摘要: | 細胞分割是許多生物學分析的先決條件。隨著多路成像技術的發展,近年來對精確分割單個細胞的需求也明顯增加。然而,當前的深度學習方法無法處理影像維度為任意順序或不同數量的染色標記。此外,在高維圖像中獲取像素級標註也非常耗時。為了解決這些問題,我們將組織病理學知識整合到我們的模型中,並提出了一個新穎的自訓練框架。具體來說,我們模擬了專家在標注細胞的過程,在訓練過程中應用空間注意力機制和最大池化操作來壓縮多通道圖像。而為了解決標注資料稀少的問題,我們除了應用自訓練來學習未標注資料外,也透過細胞核的信息來過濾偽標籤,使自訓練不受錯誤的標註影響。實驗表明,我們的方法在定性和定量結果上都優於現有方法。 Cellular segmentation is a fundamental prerequisite to many biological analyses. With the development of multiplexed imaging technologies, the need for accurately segmenting individual cells has significantly increased in recent years. However, current deep learning methods cannot deal with staining markers in an arbitrary order or different numbers. Moreover, acquiring pixel-level annotation is incredibly time-consuming in high-dimensional images. To tackle these issues, we incorporate pathology knowledge into our model and present a novel self-training framework. Concretely, we apply a serial attention mechanism and pooling operation to compress the multi-channel image during the training process. Afterward, the nuclei information guides the self-training in the pseudo-label stage. Experiments demonstrate our method is superior to the existing methods in both qualitative and quantitative results. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/84676 |
DOI: | 10.6342/NTU202203091 |
全文授權: | 同意授權(限校園內公開) |
電子全文公開日期: | 2022-09-14 |
顯示於系所單位: | 資訊網路與多媒體研究所 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
U0001-0209202201344400.pdf 授權僅限NTU校內IP使用(校園外請利用VPN校外連線服務) | 6.5 MB | Adobe PDF | 檢視/開啟 |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。