Automatic segmentation of overlapping cervical smear cells based on local distinctive features and guided shape deformation

详细信息    查看全文

• 作者：Afaf Tareef^a ; ^{atar8654@uni.sydney.edu.au}Author Vitae ; Yang Song^aAuthor Vitae ; Weidong Cai^aAuthor Vitae ; Heng Huang^bAuthor Vitae ; Hang Chang^cAuthor Vitae ; Yue Wang^dAuthor Vitae ; Michael Fulham^e ; ^fAuthor Vitae ; Dagan Feng^aAuthor Vitae ; Mei Chen^g ; ^hAuthor Vitae
• 关键词：Overlapping cervical smear cells ; Feature extraction ; Sparse coding ; Shape deformation ; Distance regularized level set
• 刊名：Neurocomputing
• 出版年：2017
• 出版时间：19 January 2017
• 年：2017
• 卷：221
• 期：Complete
• 页码：94-107
• 全文大小：1448 K
• 卷排序：221

文摘

A fully automated segmentation method is proposed for overlapping cervical cells. Our approach is based on superpixel-based features and guided shape deformation. Our shape initialization procedure is able to work with the different cell types. The practicality of our approach in segmenting highly overlapping cells is proved. Our approach outperformed existing approaches in nuclei and cytoplasm segmentation.