¾îÅÙ¼Ç ±â¹ý ¹× ÀÇ·á ¿µ»ó¿¡ÀÇ Àû¿ë¿¡ °üÇÑ Ãֽе¿Çâ
The Latest Trends in Attention Mechanisms and Their Application in Medical Imaging
½ÅÇü¼·, ÀÌÁ¤·æ, ¾îÅÂÁØ, Àü¿äÇÑ, ±è¼¼¿ø, Ȳµµ½Ä,
¼Ò¼Ó »ó¼¼Á¤º¸
½ÅÇü¼· ( Shin Hyung-Seob ) - Yonsei University Department of Electrical and Electronic Engineering
ÀÌÁ¤·æ ( Lee Jeong-Ryong ) - Yonsei University Department of Electrical and Electronic Engineering
¾îÅÂÁØ ( Eo Tae-Joon ) - Yonsei University Department of Electrical and Electronic Engineering
Àü¿äÇÑ ( Jun Yo-Han ) - Yonsei University Department of Electrical and Electronic Engineering
±è¼¼¿ø ( Kim Se-Won ) - Yonsei University Department of Electrical and Electronic Engineering
Ȳµµ½Ä ( Hwang Do-Sik ) - Yonsei University Department of Electrical and Electronic Engineering
Abstract
µö·¯´× ±â¼úÀº ºòµ¥ÀÌÅÍ ¹× ÄÄÇ»ÆÃ ÆÄ¿ö¸¦ ±â¹ÝÀ¸·Î ÃÖ±Ù ¿µ»óÀÇÇÐ ºÐ¾ßÀÇ ¿¬±¸¿¡¼ °ý¸ñÇÒ¸¸ÇÑ ¼º°ú¸¦ ÀÌ·ç¾î ³»°í ÀÖ´Ù. ÇÏÁö¸¸ ¼º´É Çâ»óÀ» À§ÇØ µö·¯´× ³×Æ®¿öÅ©°¡ ±í¾îÁú¼ö·Ï ±× ³»ºÎÀÇ °è»ê °úÁ¤À» ÇØ¼®Çϱ⠾î·Á¿öÁ³´Âµ¥, À̴ ȯÀÚÀÇ »ý¸í°ú Á÷°áµÇ´Â ÀÇ·áºÐ¾ßÀÇ ÀÇ»ç °áÁ¤ °úÁ¤¿¡¼´Â ¸Å¿ì ½É°¢ÇÑ ¹®Á¦ÀÌ´Ù. À̸¦ ÇØ°áÇϱâ À§ÇØ ¡°¼³¸í °¡´ÉÇÑ ÀΰøÁö´É ±â¼ú¡±ÀÌ ¿¬±¸µÇ°í ÀÖÀ¸¸ç, ±×Áß Çϳª·Î °³¹ßµÈ °ÍÀÌ ¹Ù·Î ¾îÅÙ¼Ç(attention) ±â¹ýÀÌ´Ù. º» Á¾¼³¿¡¼´Â ÀÌ¹Ì ÇнÀÀÌ ¿Ï·áµÈ ³×Æ®¿öÅ©¸¦ ºÐ¼®Çϱâ À§ÇÑ Post-hoc attention°ú, ³×Æ®¿öÅ© ¼º´ÉÀÇ Ãß°¡ÀûÀÎ Çâ»óÀ» À§ÇÑ Trainable attention µÎ Á¾·ùÀÇ ±â¹ý¿¡ ´ëÇØ °¢°¢ÀÇ ¹æ¹ý ¹× ÀÇ·á ¿µ»ó ¿¬±¸¿¡ Àû¿ëµÈ »ç·Ê, ±×¸®°í ÇâÈÄ Àü¸Á µî¿¡ ´ëÇØ ÀÚ¼¼È÷ ´Ù·ç°íÀÚ ÇÑ´Ù.
Deep learning has recently achieved remarkable results in the field of medical imaging. However, as a deep learning network becomes deeper to improve its performance, it becomes more difficult to interpret the processes within. This can especially be a critical problem in medical fields where diagnostic decisions are directly related to a patient's survival. In order to solve this, explainable artificial intelligence techniques are being widely studied, and an attention mechanism was developed as part of this approach. In this paper, attention techniques are divided into two types: post hoc attention, which aims to analyze a network that has already been trained, and trainable attention, which further improves network performance. Detailed comparisons of each method, examples of applications in medical imaging, and future perspectives will be covered.
Ű¿öµå
Deep Learning; Artificial Intelligence; Medical Imaging; Attention
¿ø¹® ¹× ¸µÅ©¾Æ¿ô Á¤º¸
µîÀçÀú³Î Á¤º¸