Current Result Document :
ÇѱÛÁ¦¸ñ(Korean Title) |
½ÉÃþ ½Ã¸àƽ ¼¼±×¸àÅ×À̼ÇÀ» È°¿ëÇÑ Àüü Á¶Á÷ ½½¶óÀ̵å À̹ÌÁöÀÇ Ç°Áú °ü¸® |
¿µ¹®Á¦¸ñ(English Title) |
Quality Control of Whole Slide Pathology Images with Deep Semantic Segmentation |
ÀúÀÚ(Author) |
ÀÌÁ¤Çö
À±¼¼¿µ
Junghyun Lee
Se-Young Yun
¸¶ÇØ ÀÚºó
ÃÖÈ£Áø
Á¤À¯Ã¤
Mahe Zabin
Ho-Jin Choi
Yuchae Jung
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 49 NO. 02 PP. 0660 ~ 0662 (2022. 12) |
Çѱ۳»¿ë (Korean Abstract) |
|
¿µ¹®³»¿ë (English Abstract) |
During routine slide preparation introduction of batch effects and artefacts are commonplace. The routine of slide preparation can result in tissue folds and unwanted artefacts. The digitization process may introduce blurring and contrast issues in brightness levels. Numerous computational tools are used for slide analysis, with a limited focus on quality issues. Semantic Segmentation is a deep learning technique used to map pixels of an image to specific labels. External artefacts present in a whole slide image introduce texture and color variations that can be generalized by deep learning architectures. The research proposes a standard UNet architecture to segment pen marks and regions where tissues appear to be folded. Experimental results show that the proposed method exhibits an average of 0.97 Intersection over Union (IOU) for the dataset. |
Å°¿öµå(Keyword) |
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|