• Àüü
  • ÀüÀÚ/Àü±â
  • Åë½Å
  • ÄÄÇ»ÅÍ
´Ý±â

»çÀÌÆ®¸Ê

Loading..

Please wait....

±¹³» ÇÐȸÁö

Ȩ Ȩ > ¿¬±¸¹®Çå > ±¹³» ÇÐȸÁö > µ¥ÀÌÅͺ£À̽º ¿¬±¸È¸Áö(SIGDB)

µ¥ÀÌÅͺ£À̽º ¿¬±¸È¸Áö(SIGDB)

Current Result Document :

ÇѱÛÁ¦¸ñ(Korean Title) À¯¹æ¾Ï ¸é¿ªÁ¶Á÷È­ÇÐ À̹ÌÁöÀÇ °¡»ó ¿°»ö º¯È¯
¿µ¹®Á¦¸ñ(English Title) Virtual Stain Transformation of Breast Cancer Immunohistochemistry Images
ÀúÀÚ(Author) ±Ç¿ÀÈì   ±Ç±â·æ   ¼ÛÇÏÁÖ   Oh-Heum Kwon   Ki-Ryong Kwon   Ha-Joo Song  
¿ø¹®¼ö·Ïó(Citation) VOL 38 NO. 03 PP. 0062 ~ 0072 (2022. 12)
Çѱ۳»¿ë
(Korean Abstract)
Á¶Á÷ÇÐÀû ¿°»öÀº ¼­·Î ´Ù¸¥ Á¶Á÷ ±¸¼º¿ä¼Ò °£ÀÇ »öä ±¸º°À» Á¦°øÇÏ¸ç ¾Ï µîÀÇ Áúº´ Áø´Ü¿¡ À־ ÇʼöÀûÀÎ °úÁ¤ÀÌ´Ù. ´ëÇ¥ÀûÀÎ ¿°»ö ¹æ¹ýÀÎ H&E ¿Ü¿¡µµ ƯÁ¤ Á¶Á÷ ±¸¼º¿ä¼Ò¸¦ ´õ Àß °­Á¶Çϱâ À§ÇØ »ç¿ëÇÏ´Â ´Ù¾çÇÑ Æ¯¼ö ¿°»öµéÀÌ ÀÖ´Ù. Ư¼ö ¿°»öÀº Á¾Á¾ ´õ Å« ºñ¿ë°ú ½Ã°£ÀÌ ÇÊ¿äÇϸç Áø´ÜÀÌ ±ä±ÞÇÑ °æ¿ì ÀÇ·á ½Ã½ºÅÛ°ú ȯÀÚ¿¡°Ô ºÎ´ãÀÌ µÉ ¼ö ÀÖ´Ù. È­ÇÐÀûÀÎ Á¶Á÷ ¿°»ö ´ë½Å ÀÌ¹Ì ¿°»öµÈ Á¶Á÷ÀÇ À̹ÌÁö¸¦ µö·¯´× ±â¼úÀ» ÀÌ¿ëÇÏ¿© ´Ù¸¥ ¿°»öÀ¸·Î º¯È¯ÇÏ´Â °ÍÀ» ¿°»ö º¯È¯(stain transformation)À̶ó°í ÇÑ´Ù. º» ³í¹®¿¡¼­´Â À¯¹æ¾Ï °Ë»ç¸¦ À§ÇØ H&E ¿°»öµÈ Á¶Á÷ À̹ÌÁö¸¦ IHC ¿°»ö À̹ÌÁö·Î º¯È¯ÇÏ´Â ¹®Á¦¸¦ ´Ù·é´Ù. º» ³í¹®¿¡¼­ Á¦¾ÈÇÏ´Â ¹æ¹ýÀº pix2pix ³×Æ®¿öÅ©¿¡ º¯È¯ ´ë»ó À̹ÌÁö¿Í Ÿ°Ù À̹ÌÁöÀÇ ÇüÅ»óÀÇ Â÷À̸¦ º¸¿ÏÇϱâ À§ÇÑ 2°¡Áö ±â¹ýÀ» Ãß°¡ Àû¿ëÇÏ´Â °ÍÀÌ´Ù. ù°´Â ´ë»ó À̹ÌÁö¿Í Ÿ°Ù À̹ÌÁöÀÇ ±ÙÀú(underlying) ±¸Á¶ÀÇ Â÷À̸¦ Æò°¡ÇÏ¿© À̸¦ ³×Æ®¿öÅ©ÀÇ ¼Õ½ÇÇÔ¼ö¿¡ ´ëÇÑ °¡ÁßÄ¡·Î »ç¿ëÇÏ°í, µÑ°·Î STNÀ» ÀÌ¿ëÇÏ¿© »ý¼ºµÈ À̹ÌÁö¿¡ ´ëÇØ ¾îÇÉ(affine) º¯È¯À» ¼öÇàÇÏ°í ±× °á°ú¸¦ Ÿ°Ù À̹ÌÁö¿Í ºñ±³ÇÑ´Ù. º» ³í¹®¿¡¼­ Á¦¾ÈÇÑ ³×Æ®¿öÅ©ÀÇ ¼º´ÉÀ» Æò°¡Çϱâ À§Çؼ­ ±¤¹üÀ§ÇÑ ½ÇÇèÀ» ¼öÇàÇÏ°í ±× °á°ú¸¦ ¼±Çà ¿¬±¸¿Í ºñ±³ÇÏ¿© ¼º´ÉÀÇ °³¼±À» È®ÀÎÇÏ¿´´Ù.
¿µ¹®³»¿ë
(English Abstract)
Histological staining provides color discrimination between different tissue components and is an essential process in the diagnosis of diseases such as cancer. In addition to H&E, which is the representative staining method, there are various special stains used to better highlight other tissue components. Special stains often require more cost and time and can be a burden to the health care system and patients when a diagnosis is urgent. Instead of chemical tissue staining, converting an already stained tissue image into another stain using deep learning technology is called stain transformation. In this paper, we address the problem of converting H&E-stained tissue images to IHC-stained images for breast cancer screening. The method proposed in this paper is to additionally apply two techniques to compensate for the difference between the source image and the target image to the pix2pix network. First, the difference between the underlying structure of the source and the target is evaluated and used as a weight for the loss function of the network. Second, affine transformation is performed on the generated image patches using STN, In order to evaluate the performance of the proposed network, extensive experiments were conducted and the performance improvement was confirmed by comparing the results with previous studies.
Å°¿öµå(Keyword) ¿°»öº¯È¯   À¯¹æ¾Ï   pix2pix ³×Æ®¿öÅ©   ÀÇ·áÀ̹ÌÁö󸮠  °ø°£º¯È¯³×Æ®¿öÅ©(STN)   ±â°è ÇнÀ   Stain Transformation   Breast Cancer   Pix2pix Network   Medical Image Processing   Spatial Transformer Network   Machine Learning  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå