Àá½Ã¸¸ ±â´Ù·Á ÁÖ¼¼¿ä. ·ÎµùÁßÀÔ´Ï´Ù.

79Á¾ÀÇ ÀÓÇöõÆ® ½Äº°À» À§ÇÑ µö·¯´× ¾Ë°í¸®Áò

Deep learning algorithms for identifying 79 dental implant types

±¸°­È¸º¹ÀÀ¿ë°úÇÐÁö 2022³â 38±Ç 4È£ p.196 ~ 203
°øÇöÁØ, À¯Áø¿ë, ¾ö»óÈ£, ÀÌÁØÇõ,
¼Ò¼Ó »ó¼¼Á¤º¸
°øÇöÁØ ( Kong Hyun-Jun ) - Wonkwang University College of Dentistry Department of Prosthodontics
À¯Áø¿ë ( Yoo Jin-Yong ) - HERIBio Co. Ltd.
¾ö»óÈ£ ( Eom Sang-Ho ) - HERIBio Co. Ltd.
ÀÌÁØÇõ ( Lee Jun-Hyeok ) - Korea Platform Service Technology Co. Ltd.

Abstract

¸ñÀû: º» ¿¬±¸´Â 79Á¾ÀÇ Ä¡°ú ÀÓÇöõÆ®¿¡ ´ëÇØ µö·¯´×À» ÀÌ¿ëÇÑ ½Äº° ¸ðµ¨ÀÇ Á¤È®µµ¿Í ÀÓ»óÀû À¯¿ë¼ºÀ» Æò°¡ÇÏ´Â °ÍÀ» ¸ñÀûÀ¸·Î ÇÏ¿´´Ù.

¿¬±¸ Àç·á ¹× ¹æ¹ý: 2001³âºÎÅÍ 2020³â±îÁö 30°³ Ä¡°ú¿¡¼­ ÀÓÇöõÆ® Ä¡·á¸¦ ¹ÞÀº ȯÀÚµéÀÇ Æijë¶ó¸¶ ¹æ»ç¼± »çÁø¿¡¼­ ÃÑ 45396°³ÀÇ ÀÓÇöõÆ® °íÁ¤Ã¼ À̹ÌÁö¸¦ ¼öÁýÇß´Ù. ¼öÁýµÈ ÀÓÇöõÆ® À̹ÌÁö´Â 18°³ Á¦Á¶»çÀÇ 79°³ À¯ÇüÀ̾ú´Ù. ¸ðµ¨ ÇнÀÀ» À§ÇØ EfficientNet ¹× Meta Pseudo Labels ¾Ë°í¸®ÁòÀÌ »ç¿ëµÇ¾ú´Ù. EfficientNetÀº EfficientNet-B0 ¹× Efficient-Net-B4°¡ ÇÏÀ§ ¸ðµ¨·Î »ç¿ëµÇ¾úÀ¸¸ç, Meta Pseudo Labels´Â È®Àå °è¼ö¿¡ µû¶ó µÎ °¡Áö ¸ðµ¨À» Àû¿ëÇß´Ù. EfficientNet¿¡ ´ëÇØ Top 1 Á¤È®µµ¸¦ ÃøÁ¤ÇÏ°í Meta Pseudo Labels¿¡ ´ëÇØ Top 1 ¹× Top 5 Á¤È®µµ¸¦ ÃøÁ¤ÇÏ¿´´Ù.

°á°ú: EfficientNet-B0 ¹× EfficientNet-B4´Â 89.4ÀÇ Top 1 Á¤È®µµ¸¦ º¸¿´´Ù. Meta Pseudo Labels 1Àº 87.96ÀÇ Top 1 Á¤È®µµ¸¦ º¸¿´°í, È®Àå °è¼ö°¡ Áõ°¡ÇÑ Meta Pseudo Labels 2´Â 88.35¸¦ ³ªÅ¸³Â´Ù. Top 5 Á¤È®µµ¿¡¼­ Meta Pseudo Labels 1ÀÇ Á¡¼ö´Â 97.90À¸·Î Meta Pseudo Labels 2ÀÇ 97.79º¸´Ù 0.11% ³ô¾Ò´Ù.

°á·Ð: º» ¿¬±¸¿¡¼­ ÀÓÇöõÆ® ½Äº°¿¡ »ç¿ëµÈ 4°¡Áö µö·¯´× ¾Ë°í¸®ÁòÀº ¸ðµÎ 90%¿¡ °¡±î¿î Á¤È®µµ¸¦ º¸¿´´Ù. ÀÓÇöõÆ® ½Äº°À» À§ÇÑ µö·¯´×ÀÇ ÀÓ»óÀû Àû¿ë °¡´É¼ºÀ» ³ôÀÌ·Á¸é ´õ ¸¹Àº µ¥ÀÌÅ͸¦ ¼öÁýÇÏ°í ÀÓÇöõÆ®¿¡ ÀûÇÕÇÑ ¹Ì¼¼ Á¶Á¤ ¾Ë°í¸®ÁòÀÇ °³¹ßÀÌ ÇÊ¿äÇÏ´Ù.

Purpose: This study aimed to evaluate the accuracy and clinical usability of an identification model using deep learning for 79 dental implant types.

Materials and Methods: A total of 45396 implant fixture images were collected through panoramic radiographs of patients who received implant treatment from 2001 to 2020 at 30 dental clinics. The collected implant images were 79 types from 18 manufacturers. EfficientNet and Meta Pseudo Labels algorithms were used. For EfficientNet, EfficientNet-B0 and EfficientNet-B4 were used as submodels. For Meta Pseudo Labels, two models were applied according to the widen factor. Top 1 accuracy was measured for EfficientNet and top 1 and top 5 accuracy for Meta Pseudo Labels were measured.

Results: EfficientNet-B0 and EfficientNet-B4 showed top 1 accuracy of 89.4. Meta Pseudo Labels 1 showed top 1 accuracy of 87.96, and Meta pseudo labels 2 with increased widen factor showed 88.35. In Top5 Accuracy, the score of Meta Pseudo Labels 1 was 97.90, which was 0.11% higher than 97.79 of Meta Pseudo Labels 2.

Conclusion: All four deep learning algorithms used for implant identification in this study showed close to 90% accuracy. In order to increase the clinical applicability of deep learning for implant identification, it will be necessary to collect a wider amount of data and develop a fine-tuned algorithm for implant identification.

Å°¿öµå

Ä¡°ú ÀÓÇöõÆ®; µö·¯´×; ÀΰøÁö´É; ÇÕ¼º°ö ½Å°æ¸Á
dental implants; artificial intelligence; deep learning; convolutional neural networks

¿ø¹® ¹× ¸µÅ©¾Æ¿ô Á¤º¸

µîÀçÀú³Î Á¤º¸

KCI