• ๋Œ€ํ•œ์ „๊ธฐํ•™ํšŒ
Mobile QR Code QR CODE : The Transactions of the Korean Institute of Electrical Engineers
  • COPE
  • kcse
  • ํ•œ๊ตญ๊ณผํ•™๊ธฐ์ˆ ๋‹จ์ฒด์ด์—ฐํ•ฉํšŒ
  • ํ•œ๊ตญํ•™์ˆ ์ง€์ธ์šฉ์ƒ‰์ธ
  • Scopus
  • crossref
  • orcid

  1. (Interdisciplinary Graduate Program for BIT Medical Con- vergence, Kangwon National University, Korea)
  2. (Department of Medicine, University of Ulsan College of Medicine, Korea)
  3. (Dept. of Electronic Engineering and Interdisciplinary Graduate Program for BIT Medical Con-vergence, Kangwon National University, Korea)



Dental caries, Snake algorithm, DRLSE, CADe, Active contour algorithms

1. ์„œ๋ก 

์šฐ๋ฆฌ๋‚˜๋ผ ๊ตญ๋ฏผ์˜ ๊ตฌ๊ฐ•๊ฑด๊ฐ•์„ ์œ„ํ˜‘ํ•˜๋Š” ๋Œ€ํ‘œ์ ์ธ ์งˆํ™˜์œผ๋กœ ์น˜์•„์šฐ์‹์ฆ(Dental caries)์ด ์žˆ๋‹ค. ์œ ๋ณ‘๋ฅ ์ด ์ค„์–ด๋“œ๋Š” ์ถ”์„ธ์ด๊ธด ํ•˜๋‚˜ ์•„์ง๋„ ๋‚จ๋…€ ํ‰๊ท  30%์ •๋„๊ฐ€ 1๊ฐœ ์ด์ƒ์˜ ์น˜์•„์šฐ์‹์ฆ์„ ๊ฐ€์ง€๊ณ  ์žˆ๋Š” ๋งŒํผ ์น˜์•„์šฐ์‹์ฆ์€ ๋Œ€ํ‘œ์ ์ธ ๊ตฌ๊ฐ•๋ณ‘ ์ค‘ ํ•˜๋‚˜์ด๋‹ค(1).

๋ณดํ†ต์˜ ์น˜์•„์šฐ์‹์ฆ ์ง„๋‹จ์€ ๋””์ง€ํ„ธ X-์„  ์˜์ƒ์„ ์น˜๊ณผ์˜์‚ฌ๋ฅผ ํ†ตํ•ด ์ง„๋‹จํ•œ๋‹ค. ์ด๋Š” ์น˜๊ณผ์˜์‚ฌ๊ฐ€ ์œก์•ˆ์œผ๋กœ ๊ด€์ฐฐํ•˜์—ฌ ์˜์‚ฌ์˜ ๊ฐœ์ธ์ ์ธ ์ง€์‹๊ณผ ๊ฒฝํ—˜์— ๋ฐ”ํƒ•์„ ํ•œ ์ฃผ๊ด€์ ์ธ ํŒ๋‹จ์— ์˜์กดํ•˜๊ฒŒ ๋œ๋‹ค. ๊ทธ๋ ‡๊ธฐ ๋•Œ๋ฌธ์— ์œก์•ˆ์œผ๋กœ ํ™•์ธํ•˜๊ธฐ ์–ด๋ ค์šด ํŠน์ดํ•˜๊ณ  ๋ณต์žกํ•œ ํ˜•ํƒœ์˜ ์น˜์•„์˜ ๊ฒฝ์šฐ๋Š” ๋‹ค์–‘ํ•œ ์ด์œ ๋กœ ์ž˜๋ชป๋œ ํŒ๋‹จ์„ ๋‚ด๋ฆด ์ˆ˜ ์žˆ๋‹ค. ์šฐ์‹์ฆ์„ ์ดˆ๊ธฐ์— ๋ฐœ๊ฒฌํ•˜๋ฉด ๊ฐ„๋‹จํ•œ ์ฒ˜์น˜๋งŒ์œผ๋กœ ํšŒ๋ณตํ•  ์ˆ˜ ์žˆ๊ฒ ์ง€๋งŒ ๋ฐœ๊ฒฌํ•˜์ง€ ๋ชปํ•˜์˜€์„ ๊ฒฝ์šฐ ์‹œ๊ฐ„์ด ์ง€๋‚จ์— ๋”ฐ๋ผ ์•…ํ™”๋  ๊ฐ€๋Šฅ์„ฑ์ด ์žˆ๋‹ค. ๋”ฐ๋ผ์„œ ์˜์‚ฌ์˜ ์ •ํ™•ํ•œ ์ง„๋‹จ์€ ์น˜๋ฃŒ์˜ ์˜ˆํ›„๋ฅผ ๊ฒฐ์ •ํ•˜๋Š”๋ฐ ์žˆ์–ด ์ค‘์š”ํ•œ ์—ญํ• ์„ ํ•œ๋‹ค. ๋Œ€๋ถ€๋ถ„์˜ ์น˜๊ณผ ์ง„๋ฃŒ์—์„œ๋Š” ์น˜๊ณผ์˜์‚ฌ์˜ ๊ฒฝํ—˜๊ณผ ์ง€์‹์„ ๊ธฐ๋ฐ˜ํ•œ ์œก์•ˆ๊ฒ€์‚ฌ ๋ฐ ์ด‰๊ฐ์— ์˜์กดํ•˜๊ณ  ์žˆ๋‹ค. ์ด๋Ÿฌํ•œ ๋ฐฉ๋ฒ•์ด ๋Œ€๋ถ€๋ถ„์˜ ์ถฉ์น˜๋ณ‘๋ณ€์„ ๊ฐ์ง€ ํ•  ์ˆ˜ ์žˆ์ง€๋งŒ, ์ด ์™ธ์—๋„ ์˜์‚ฌ์˜ ์น˜์•„์šฐ์‹์ฆ์„ ํŒ๋ณ„ํ•  ์ˆ˜ ์žˆ๋Š” ๋‹ค์–‘ํ•œ ๋ฐฉ๋ฒ•๋“ค์ด ์กด์žฌํ•œ๋‹ค. ์น˜์•„์กฐ์ง์˜ ์ €ํ•ญ์„ ์ธก์ •ํ•˜์—ฌ ํŒ๋ณ„ํ•˜๋Š” Electronic caries monitor (ECM), ์œก์•ˆ์œผ๋กœ ๋น›์˜ ์‚ฐ๋ž€ํ˜„์ƒ์„ ์ด์šฉํ•˜์—ฌ ํŒ๋ณ„ํ•˜๋Š” Fiber optic transilumination(FOTI)์™€ ๋””์ง€ํ„ธ ์˜์ƒ์„ ์ด์šฉํ•œ Digital imaging FOTI(DiFOTI), ํ˜•๊ด‘์„ ์ด์šฉํ•˜์—ฌ ๋ชจ๋‹ˆํ„ฐ๋งํ•  ์ˆ˜ ์žˆ๋Š” Qu antitative light-induced fluorescence(QLF)์™€ DIAGNOdent (DD)๊ฐ€ ์žˆ์œผ๋ฉฐ, ์ดˆ์ŒํŒŒ๋ฅผ ์ด์šฉํ•œ Ultrasonic Caries Detector (UCD)๊ฐ€ ์žˆ๋‹ค. FOTI์™€ DiFOTI๋Š” ๋‚ฎ์€ ๋ฏผ๊ฐ๋„๋ฅผ ๊ฐ–๊ณ , ECM, QLF ๋“ฑ์˜ ๋ฐฉ๋ฒ•์€ ์ข‹์€ ์„ฑ๋Šฅ์„ ๊ฐ€์ง€์ง€๋งŒ ๋น„์šฉ์ด ์ฆ๊ฐ€ํ•œ๋‹ค. ๊ทธ๋ ‡๊ธฐ ๋•Œ๋ฌธ์— ๋‹ค๋ฅธ ์ถ”๊ฐ€์ ์ธ ์žฅ์น˜๊ฐ€ ํ•„์š”ํ•˜์ง€ ์•Š์€ ๋””์ง€ํ„ธ X-์„  ์˜์ƒ์„ ์ด์šฉํ•œ ์น˜์•„ ์šฐ์‹ ๋ณ‘๋ณ€ ๊ฒ€์ถœ ์‹œ์Šคํ…œ์ด ๊ฐ€์žฅ ๋งŽ์ด ์—ฐ๊ตฌ๋˜๊ณ  ์žˆ๋‹ค(2-4). ์ด๋ฒˆ ์—ฐ๊ตฌ๋Š” ๊ตญ์ œ ์ถฉ์ง€ ํƒ์ง€ ํ‰๊ฐ€์‹œ์Šคํ…œ์ธ ICDAS (International Caries Detection and Assesment System)์—์„œ ์ œ์‹œํ•œ 3mm2 ์ด์ƒ์˜ ๋ฉด์ ์„ ๊ฐ€์ง€๋Š” ๋ณ‘๋ณ€์„ ๊ฒ€์ถœ ํ•˜๋Š” ๊ฒƒ์„ ๋ชฉ์ ์œผ๋กœ ๋””์ง€ํ„ธ X-์„  ์˜์ƒ์˜ ์™€๋™ ์ง„๋‹จ ๋ณด์กฐ ์‹œ์Šคํ…œ์„ ๊ฐœ๋ฐœํ•˜์˜€๋‹ค(5). X-์„  ์˜์ƒ์—์„œ์˜ ์น˜์•„ ์šฐ์‹๊ฒ€์ถœ์€ ๋ณ‘๋ณ€ ํ˜•ํƒœ๊ฐ€ ๋‹ค์–‘ํ•˜๊ธฐ ๋•Œ๋ฌธ์— ๊ฒ€์ถœํ•˜๊ธฐ ๊นŒ๋‹ค๋กญ๋‹ค. ํŠนํžˆ ์น˜์•„ ์ธ์ ‘๋ฉด์˜ ๊ฒฝ์šฐ ๊ด€์ธกํ•˜๊ธฐ ๋” ์–ด๋ ต๋‹ค. ๋ณ‘๋ณ€ ๊ฒ€์ถœ์„ ์œ„ํ•ด ์‚ฌ์šฉํ•˜๋Š” ์˜์ƒ๋ถ„ํ• ์˜ ๋Œ€ํ‘œ์ ์ธ ๋ฐฉ๋ฒ•์œผ๋กœ Active Contour model(๋™์  ์œค๊ณฝ์„  ๋ชจ๋ธ)์ด ์žˆ๋‹ค. Active Contour Model์€ Parametric model(๋งค๊ฐœ๋ณ€์ˆ˜ ๋ชจ๋ธ)๊ณผ Geometric model(๊ธฐํ•˜ํ•™์  ๋ชจ๋ธ)๋กœ ๋ถ„๋ฅ˜ํ•  ์ˆ˜ ์žˆ๋‹ค. ์Šค๋„ค์ดํฌ(Snake)๋Š” ๋Œ€ํ‘œ์ ์ธ ๋งค๊ฐœ๋ณ€์ˆ˜ ๋ชจ๋ธ๋กœ ์—๋„ˆ์ง€๊ฐ€ ์ตœ์†Œํ™” ๋˜๋„๋ก ์ •์˜๋œ ํ•จ์ˆ˜๋ฅผ ํ†ตํ•˜์—ฌ ์ดˆ๊ธฐ ๊ณก์„ ์œผ๋กœ๋ถ€ํ„ฐ ์ ์ง„์ ์œผ๋กœ ๋ณ€ํ™”์‹œ์ผœ ๋ฌผ์ฒด์˜ ์™ธ๊ฐ์„ ์ฐพ๋Š” ๋ฐฉ์‹์ด๊ณ , ๋ ˆ๋ฒจ์…‹ ๋ฐฉ๋ฒ•(Level set method)์€ ๊ธฐํ•˜ํ•™์  ๋ชจ๋ธ๋กœ ๊ด€์‹ฌ์˜์—ญ์˜ ํ˜•ํƒœ๋ณ€ํ™”์— ๋Œ€ํ•˜์—ฌ ์‰ฝ๊ฒŒ ์ ์‘ํ•  ์ˆ˜ ์žˆ๊ณ  ๋™์‹œ์— ๋‹ค์ˆ˜์˜ ๊ฐ์ฒด๋ฅผ ๊ฒ€์ถœํ•  ์ˆ˜ ์žˆ๋‹ค. ๊ธฐ์กด์˜ ์˜์ƒ์ฒ˜๋ฆฌ ๋ถ„์•ผ์—์„œ ์‚ฌ์šฉ๋˜๊ณ  ์žˆ๋Š” Active Contour model ์ค‘ ์Šค๋„ค์ดํฌ ์•Œ๊ณ ๋ฆฌ์ฆ˜๊ณผ ๋ ˆ๋ฒจ์…‹ ๋ฐฉ๋ฒ•์—์„œ ๊ฑฐ๋ฆฌ ์ •๊ทœํ™”ํ•ญ์„ ์ถ”๊ฐ€ํ•œ DRLSE(Direct Regularized Level Set Evolution) ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ๊ธฐ๋ฐ˜์œผ๋กœ ํ•œ ์น˜์•„ ์™€๋™ ๊ฒ€์ถœ ํ”„๋กœ๊ทธ๋žจ์„ ๋น„๊ตํ•˜์˜€๋‹ค.

2. ์ด๋ก  ๋ฐ ๋ฐฉ๋ฒ•

2.1 ์ „์ฒ˜๋ฆฌ ๊ณผ์ •(Pre-processing)

์น˜์•„์˜ X-์„  ์˜์ƒ์€ ์™€๋™๊ณผ ์ฃผ๋ณ€๊ณผ์˜ ๊ฒฝ๊ณ„๊ฐ€ ๋ชจํ˜ธํ•˜๋ฉฐ ์กฐ์ง ๋‚ด์—์„œ๋„ ์น˜์•„์˜ ๋‘๊ป˜์ฐจ์ด์™€ ๊ฐ™์€ ๋‹ค๋ฆ„์ด ์ƒ๊ธฐ๊ธฐ ๋•Œ๋ฌธ์— ์™€๋™์˜ ์œค๊ณฝ์„ ์ •ํ™•ํ•˜๊ฒŒ ์ฐพ๋Š” ๊ฒƒ์€ ์–ด๋ ค์šด ์ผ์ด๋‹ค. ๋˜ํ•œ X-์„  ์ดฌ์˜ ์‹œ์— ์ƒ๊ธฐ๋Š” ๋…ธ์ด์ฆˆ์™€ ๋ฐ๊ธฐ์ฐจ์ด๊ฐ€ ์ƒ๊ธด๋‹ค. ์ „์ฒ˜๋ฆฌ ๊ณผ์ •์€ ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ์ ์šฉํ•˜๊ธฐ ์ ํ•ฉํ•œ ์˜์ƒ์„ ๋งŒ๋“ค๋„๋ก ํ•œ๋‹ค. ์ „์ฒ˜๋ฆฌ๋กœ๋Š” ๋ช…์•” ๋Œ€๋น„ ์ŠคํŠธ๋ ˆ์นญ, ์–‘๋ฐฉํ–ฅ ํ•„ํ„ฐ, ์บ๋‹ˆ ์—ฃ์ง€๊ฐ€ ์‚ฌ์šฉ๋˜์—ˆ๋‹ค.

2.1.1 ๋ช…์•” ๋Œ€๋น„ ์ŠคํŠธ๋ ˆ์นญ(Contrast Stretching)

X-์„  ์˜์ƒ ํš๋“ ์‹œ ์˜์ƒ์˜ ๋ฐ๊ธฐ ์ฐจ์ด๊ฐ€ ๋ฐœ์ƒํ•œ๋‹ค. ์ด๋ฅผ ๋ณด์ •ํ•˜๊ธฐ ์œ„ํ•ด์„œ ๋ช…์•” ๋Œ€๋น„ ์ŠคํŠธ๋ ˆ์นญ ๋ณ€ํ™˜์„ ์ ์šฉํ•˜์˜€๋‹ค. ํžˆ์Šคํ† ๊ทธ๋žจ์ด๋ž€ ์–ด๋–ค ๋ฐ์ดํ„ฐ๊ฐ€ ์–ผ๋งˆ๋‚˜ ๋งŽ์€์ง€ ๋‚˜ํƒ€๋‚ด๋Š” ๋„์ˆ˜ ๋ถ„ํฌํ‘œ๋ฅผ ๊ทธ๋ž˜ํ”„๋กœ ๋‚˜ํƒ€๋‚ธ ๊ฒƒ์ด๋‹ค. ์˜์ƒ์ด ๋ณด๊ธฐ์— ์„ ๋ช…ํ•˜๊ณ  ๊นจ๋—ํ•˜๊ฒŒ ๋ณด์ด๋ ค๋ฉด ์–ด๋‘์šด ๋ถ€๋ถ„์—์„œ ๋ฐ์€ ๋ถ€๋ถ„๊นŒ์ง€ ํžˆ์Šคํ† ๊ทธ๋žจ์ด ๊ณ ๋ฃจ ๋ถ„ํฌ๋˜์–ด ์žˆ์–ด์•ผํ•œ๋‹ค. ๋ช…์•” ๋Œ€๋น„ ์ŠคํŠธ๋ ˆ์นญ์ด๋ž€ ๋ช…์•” ๋ถ„ํฌ๊ฐ€ ์ข์€ ํžˆ์Šคํ† ๊ทธ๋žจ์„ ์ขŒ์šฐ๋กœ ์žก์•„๋‹น๊ฒจ ๊ณ ๋ฅธ ๋ช…์•” ๋ถ„ํฌ๋ฅผ ๊ฐ€์ง„ ํžˆ์Šคํ† ๊ทธ๋žจ์ด ๋˜๊ฒŒ ํ•˜๋Š” ๊ฒƒ์ด๋‹ค. ํžˆ์Šคํ† ๊ทธ๋žจ์˜ ๋ถ„ํฌ๊ฐ€ ๋ฐ”๋€๋‹ค๋Š” ๊ฒƒ์€ ์˜์ƒ ๋‚ด์˜ ํ™”์†Œ ๊ฐ’์˜ ๋ถ„ํฌ๊ฐ€ ๋ฐ”๋€๋‹ค๋Š” ๊ฒƒ์„ ์˜๋ฏธํ•œ๋‹ค. ๊ฐ ํ”ฝ์…€์˜ ๊ฐ’์„ ์‹(1)์— ์ ์šฉํ•˜๋ฉด ์ƒˆ๋กœ์šด ๊ฐ’์„ ์ •ํ•  ์ˆ˜ ์žˆ๋‹ค. ๋”ฐ๋ผ์„œ ์˜์ƒ์˜ ํ™”์งˆ์ด ๋ณ€๊ฒฝ๋˜๊ณ  ๊ฐœ์„ ๋  ์ˆ˜ ์žˆ๋Š” ๊ฒƒ์ด๋‹ค.

(1)
n e w  pixel  = ( p i x e l - l o w ) * 255 ฮ” x ,     ฮ” x = h i g h - l o w

2.1.2 ์–‘๋ฐฉํ–ฅ ํ•„ํ„ฐ(Bilateral filter)

์–‘๋ฐฉํ–ฅ ํ•„ํ„ฐ๋Š” ๊ฒฝ๊ณ„๋ฅผ ๋ณด์กดํ•˜๋ฉด์„œ ๋…ธ์ด์ฆˆ๋ฅผ ๊ฐ์†Œ์‹œํ‚ค๋Š” ํ•„ํ„ฐ๋กœ, ๋‘ ์ข…๋ฅ˜์˜ ํ•„ํ„ฐ๊ฐ€ ํ•ฉ์ณ์„œ ์–‘๋ฐฉํ–ฅ ํ•„ํ„ฐ๋ผ ๋ถˆ๋ฆฐ๋‹ค. ์–‘๋ฐฉํ–ฅ ํ•„ํ„ฐ๋Š” ๋ธ”๋ก๋‚ด์˜ ํ™”์†Œ์˜ ์œ„์น˜์ฐจ์ด์— ๋”ฐ๋ผ ์˜์—ญํ•„ํ„ฐ๋ฅผ ์ง€์ •ํ•˜๊ณ  ํ™”์†Œ ๊ฐ’์˜ ์ฐจ์ด์— ๋”ฐ๋ผ ๋ฒ”์œ„ ํ•„ํ„ฐ๋ฅผ ์ง€์ •ํ•˜์—ฌ ์ด ๋‘˜์˜ ๊ฐ๊ฐ ๊ฐ€์ค‘์น˜์— ๋”ฐ๋ผ ์ ์šฉํ•˜๋Š” ๋ฐฉ์‹์ด๋‹ค. ์ด๋ฅผ ์ˆ˜์‹์œผ๋กœ ํ‘œํ˜„ํ•˜๋ฉด ๋‹ค์Œ๊ณผ ๊ฐ™๋‹ค.

(2)
I ~ ( x ) = 1 C โˆ‘ e y โˆˆ ฮฉ - โ€– y - x โ€– 2 2 ฯƒ d 2 e - | I ( y ) - I ( x ) | 2 2 ฯƒ r 2    I ( y )

ฯƒd๋Š” ํ™”์†Œ ์œ„์น˜ ์ฐจ์ด์— ๋Œ€ํ•œ ๊ฐ€์ค‘์น˜, ฯƒr๋Š” ํ™”์†Œ ๊ฐ’์˜ ์ฐจ์ด์— ๋Œ€ํ•œ ๊ฐ€์ค‘์น˜๋ฅผ ๋‚˜ํƒ€๋‚ธ๋‹ค. x๋Š” ํ˜„์žฌ ํ™”์†Œ์˜ ์ขŒํ‘œ, Ix๋Š” x์˜ ํ™”์†Œ ๊ฐ’, y๋Š” ํ•„ํ„ฐ๋ง ๋ฒ”์œ„๋กœ ์„ค์ •๋œ โ„ฆ๋‚ด์˜ ์ฃผ๋ณ€ ํ™”์†Œ์˜ ์ขŒํ‘œ, โ„ฆ๋Š” ํ•„ํ„ฐ๋งํ•  ์œˆ๋„์šฐ์˜ ๋ฒ”์œ„, I(y)๋Š” y์˜ ํ™”์†Œ ๊ฐ’์„ ๋‚˜ํƒ€๋‚ธ๋‹ค. C๋Š” ํ‘œ์ค€ํ™” ์ƒ์ˆ˜๋กœ ๋‹ค์Œ๊ณผ ๊ฐ™๋‹ค.

(3)
C = โˆ‘ y โˆˆ ฮฉ e - โ€– y - x โ€– 2 2 ฯƒ d 2 e - | I ( y ) - I ( x ) | 2 2 ฯƒ r 2

์–‘๋ฐฉํ–ฅ ํ•„ํ„ฐ๋Š” ๋‹ค๋ฅธ ํ•„ํ„ฐ์™€ ๋‹ฌ๋ฆฌ ฯƒd์™€ ฯƒr ๋‘ ๊ฐ€์ง€ ํŒŒ๋ผ๋ฏธํ„ฐ๋ฅผ ๋ชจ๋‘ ๊ณ ๋ คํ•˜๊ธฐ ๋•Œ๋ฌธ์— ๋…ธ์ด์ฆˆ๋ฅผ ๊ฐ์†Œ์‹œํ‚ค๊ณ  ์—ฃ์ง€๋ถ€๋ถ„์€ ๋ณด์กดํ•  ์ˆ˜ ์žˆ๋‹ค.

2.1.3 ์บ๋‹ˆ ์—ฃ์ง€(Canny edge)

์—ฃ์ง€๋Š” ์ด๋ฏธ์ง€์˜ ๊ฒฝ๊ณ„๋ฅผ ํŠน์ง•์ง“๊ธฐ ๋•Œ๋ฌธ์— ์ด๋ฏธ์ง€ ์ฒ˜๋ฆฌ์— ์žˆ์–ด ์ค‘์š”ํ•œ ์š”์†Œ์ด๋‹ค. ์—ฃ์ง€๋ฅผ ์ฐพ๋Š” ํ”„๋กœ์„ธ์Šค๋Š” ์ฒ˜๋ฆฌํ•  ๋ฐ์ดํ„ฐ์˜ ์–‘์„ ํฌ๊ฒŒ ์ค„์ž„๊ณผ ๋™์‹œ์— ๊ฐ์ฒด ๊ฒฝ๊ณ„์— ๋Œ€ํ•œ ์œ ์šฉํ•œ ๊ตฌ์กฐ ์ •๋ณด๋ฅผ ๋ณด์กดํ•˜์—ฌ ์ด๋ฏธ์ง€ ๋ถ„์„์„ ๋‹จ์ˆœํ•˜๊ฒŒ ํ•œ๋‹ค. ์บ๋‹ˆ ์—ฃ์ง€ ๊ฒ€์ถœ๊ธฐ๋Š” ์ตœ์ ์˜ ์—์ง€ ๊ฒ€์ถœ๊ธฐ๋กœ ์•Œ๋ ค์ ธ ์ž‡๋‹ค. ์บ๋‹ˆ ์—ฃ์ง€๋Š” 1986๋…„ John F. Canny ์— ์˜ํ•ด ๊ฐœ๋ฐœ๋œ ์•Œ๊ณ ๋ฆฌ์ฆ˜์ด๋‹ค(6). ์บ๋‹ˆ ์—ฃ์ง€๋Š” ๋‚ฎ์€ ์—๋Ÿฌ์œจ, ์ •ํ™•ํ•œ ์—ฃ์ง€ ์ ๋“ค์˜ ์œ„์น˜, ๊ฒน์น˜์ง€ ์•Š๋Š” ๋‹จ์ผ ์—ฃ์ง€ ์ ์˜ ์‘๋‹ต ์ด๋ ‡๊ฒŒ ์„ธ ๊ฐ€์ง€์˜ ์žฅ์ ์„ ๋ณด์ธ๋‹ค. ์—ฐ๊ตฌ์— ์‚ฌ์šฉ๋œ ์บ๋‹ˆ ์—ฃ์ง€๋Š” ๋‹ค์Œ๊ณผ ๊ฐ™์ด 5๋‹จ๊ณ„๋กœ ์ ์šฉํ•˜์˜€๋‹ค(7).

1. ์žก์Œ์„ ์ œ๊ฑฐํ•˜๊ธฐ ์œ„ํ•œ ์Šค๋ฌด๋”ฉ ๋‹จ๊ณ„

2. ์ด๋ฏธ์ง€์˜ ๊ฒฝ์‚ฌ๋„(๋ฏธ๋ถ„)๊ฐ’์„ ํ†ตํ•ด ์—ฃ์ง€๋ฅผ ์ฐพ๋Š” ๋‹จ๊ณ„

3. ์†Œ๋ฒจ ๋งˆ์Šคํ‚น์œผ๋กœ ์—ฃ์ง€์˜ ๊ฐ•๋„์™€ ๋ฐฉํ–ฅ์„ ์ถ”์ถœํ•˜๋Š” ๋‹จ๊ณ„

4. ๋น„์–ต์ œ(Non maximum suppression)์ ์šฉ๋‹จ๊ณ„

5. Double thresholding ์ ์šฉ ๋‹จ๊ณ„

์ด๋ ‡๊ฒŒ ์ ์šฉํ•œ ์ „์ฒ˜๋ฆฌ ๊ณผ์ •๋“ค์€ ์Šค๋„ค์ดํฌ์™€ DRLSE๊ฐ€ ๋” ์œค๊ณฝ์„ ์„ ์ž˜ ์ฐพ๋„๋ก ๋„์›€์„ ์ค€๋‹ค.

2.2 ์Šค๋„ค์ดํฌ ์•Œ๊ณ ๋ฆฌ์ฆ˜

Kass(8)์— ์˜ํ•ด ์ฒ˜์Œ ์†Œ๊ฐœ๋œ ์Šค๋„ค์ดํฌ ์•Œ๊ณ ๋ฆฌ์ฆ˜์€ ์ •์˜๋œ ์—๋„ˆ์ง€ ํ•จ์ˆ˜๊ฐ€ ์ตœ์†Œํ™” ์—ฐ์‚ฐ์„ ๋ฐ˜๋ณตํ•˜์—ฌ ์Šค๋„ค์ดํฌ ํฌ์ธํŠธ๋ฅผ ์ด๋™ํ•œ๋‹ค. ์ด๋Š” ๊ฐ์ฒด ๋ฐฉํ–ฅ์œผ๋กœ ์ด๋™ํ•˜์—ฌ ๊ฐ์ฒด ์œค๊ณฝ์„ ์ฐพ๋Š” ๋ฐฉ๋ฒ•์ด๋‹ค. ์Šค๋„ค์ดํฌ ์•Œ๊ณ ๋ฆฌ์ฆ˜์˜ ๊ฐœ๋…์„ ๊ทธ๋ฆผ. 1๊ณผ ๊ฐ™์ด ๋‚˜ํƒ€๋‚ด์—ˆ๋‹ค. ์Šค๋„ค์ดํฌ ๊ฐ ํฌ์ธํŠธ๋ฅผ Vs,t,i=(xs,t,i,ys,t,i)๋ผ ํ–ˆ์„ ๋•Œ s๋Š” ์Šค๋„ค์ดํฌ, t๋Š” ์‹œ๊ฐ„, i๋Š” i๋ฒˆ์งธ ์Šค๋„ค์ดํฌ ํฌ์ธํŠธ๋ฅผ ๋‚˜ํƒ€๋‚ด๊ณ , x์™€ y๋Š” ๊ฐ ์Šค๋„ค์ดํฌ ํฌ์ธํŠธ์˜ ์ขŒํ‘œ๋ฅผ ๋‚˜ํƒ€๋‚ธ๋‹ค. ๊ทธ๋ฆผ1์™€ ๊ฐ™์ด ์Šค๋„ค์ดํฌ๋Š” ์ดˆ๊ธฐ ํฌ์ธํŠธ์—์„œ ์—๋„ˆ์ง€ ํ•จ์ˆ˜๋ฅผ ์ตœ์†Œํ™”ํ•˜๋„๋ก ์—ฐ์‚ฐ์„ ๋ฐ˜๋ณตํ•˜์—ฌ ์Šค๋„ค์ดํฌ ํฌ์ธํŠธ ๊ฐ„๊ฒฉ๋“ค์ด ์ ์ง„์ ์œผ๋กœ ๊ฐ™์€ ๊ฐ„๊ฒฉ์ด ๋˜๋ฉด์„œ ์Šค๋„ค์ดํฌ ํฌ์ธํŠธ๋ฅผ ์—ฐ๊ฒฐํ•œ ์œค๊ณฝ์„ ์€ ์›๋ชจ์–‘์„ ๋ˆ๋‹ค. ์Šค๋„ค์ดํฌ ํฌ์ธํŠธ๋“ค์˜ ์›€์ง์ž„์€ ๋ฌผ์ฒด์˜ ์œค๊ณฝ์˜ ๋„๋‹ฌํ•˜๊ฑฐ๋‚˜ ์—๋„ˆ์ง€ ํฌํ™”ํ˜„์ƒ์œผ๋กœ ๋” ์ด์ƒ ์—ฐ์‚ฐํ•  ์ˆ˜ ์—†์„ ๋•Œ ๋ฉˆ์ถ˜๋‹ค.

๊ทธ๋ฆผ. 1. ์Šค๋„ค์ดํฌ ๋ฐฉ๋ฒ•์˜ ํ๋ฆ„๋„

Fig. 1. Snake algorithm process

../../Resources/kiee/KIEE.2018.67.12.1678/fig1.png

์Šค๋„ค์ดํฌ ์•Œ๊ณ ๋ฆฌ์ฆ˜์˜ ์›€์ง์ž„์€ ๋‹ค์Œ๊ณผ ๊ฐ™์ด ์ •์˜๋œ ์—๋„ˆ์ง€ ํ•จ์ˆ˜์— ๊ธฐ๋ฐ˜ํ•˜์—ฌ ์›€์ง์ธ๋‹ค.

(4)
E s n a k e * = โˆซ 0 1 E s n a k e ( V ( s ) ) d s = โˆซ 0 1 E i n t ( V ( s ) ) + E e x t ( V ( s ) ) d s

(5)
E i n t = ฮฑ ( s ) | V s ( s ) | 2 + ฮฒ | V s s ( s ) | 2

(6)
E e x p ( V ( s ) ) = - ฮณ | โˆ‡ f ( V ( s ) ) | 2

์Šค๋„ค์ดํฌ ํ•จ์ˆ˜๋Š” ๋‚ด๋ถ€์—๋„ˆ์ง€(Eint)์™€ ์™ธ๋ถ€์—๋„ˆ์ง€(Eext)๋กœ ๋‚˜๋ˆŒ ์ˆ˜ ์žˆ๋Š”๋ฐ ๋‹ค์‹œ ๋‚ด๋ถ€์—๋„ˆ์ง€๋Š” ์‹(5)๊ณผ ๊ฐ™์ด ๋‚˜ํƒ€๋‚ผ ์ˆ˜ ์žˆ๋‹ค. ฮฑ, ฮฒ๋Š” ๊ฐ€์ค‘์น˜๋ฅผ ์˜๋ฏธํ•˜๊ณ  Vs๋Š” 1์ฐจ ๋ฏธ๋ถ„ Vss๋Š” 2์ฐจ ๋ฏธ๋ถ„์„ ๋œปํ•œ๋‹ค. ๋‚ด๋ถ€์—๋„ˆ์ง€๋Š” ์Šค๋„ค์ดํฌ ํฌ์ธํŠธ๋“ค์˜ ๊ฐ„๊ฒฉ์„ ์กฐ์ ˆํ•˜๊ณ  ๊ณก์„ ์˜ ๋ชจ์–‘์„ ๊ฒฐ์ •ํ•œ๋‹ค. ์™ธ๋ถ€์—๋„ˆ์ง€๋Š” ์‹(6)๊ณผ ๊ฐ™์ด ํ‘œํ˜„ํ•  ์ˆ˜ ์žˆ๋Š”๋ฐ, โˆ‡๋Š” Gradient, f๋Š” ์Šค๋„ค์ดํฌ ํฌ์ธํŠธ์˜ ๋ช…์•” ๊ฐ’, ฮณ๋Š” ๊ฐ€์ค‘์น˜๋ฅผ ๋‚˜ํƒ€๋‚ธ๋‹ค. ์™ธ๋ถ€์—๋„ˆ์ง€๋Š” ๊ฒฝ์‚ฌ๋„๊ฐ€ ํฐ ๊ฐ’์—์„œ ์ ์ง„์ ์œผ๋กœ ์ž‘์€ ๊ฐ’์„ ๊ฐ–๊ฒŒ ํ•˜์—ฌ ์Šค๋„ค์ดํฌ ํฌ์ธํŠธ๊ฐ€ ๊ฐœ์ฒด ์œค๊ณฝ์—์„œ ์ •์ง€๋˜๋„๋ก ํ•œ๋‹ค. ์•ž์„œ ๋งํ–ˆ๋“ฏ์ด ๊ฐ™์ด ์ดˆ๊ธฐ ์Šค๋„ค์ดํฌ ํฌ์ธํŠธ์—์„œ ์—ฐ์‚ฐ์„ ๋ฐ˜๋ณตํ•˜์—ฌ ์Šค๋„ค์ดํฌ ํฌ์ธํŠธ๋“ค์„ ๋ฌผ์ฒด๋ฐฉํ–ฅ์œผ๋กœ ์ด๋™์‹œ์ผœ ์œค๊ณฝ์„ ์ถ”์ถœํ•œ๋‹ค. ์ด๋•Œ ๋‚ด๋ถ€์—๋„ˆ์ง€ Eint์™€ ์™ธ๋ถ€์—๋„ˆ์ง€ Eext์˜ ๊ฐ€์ค‘์น˜๋ฅผ ์กฐ์ ˆํ•˜์—ฌ ์ ์ ˆํ•œ ์กฐํ•ฉ์„ ์ฐพ๋Š”๋‹ค. ์Šค๋„ค์ดํฌ ๋ฐฉ๋ฒ•์€ ๋‹ค๋ฅธ ๋ฐฉ๋ฒ•๋“ค์— ๋น„ํ•ด ๊ณ„์‚ฐ์ด ๊ฐ„๋‹จํ•˜์—ฌ ๋น ๋ฅด๊ฒŒ ์œค๊ณฝ์„ ์ฐพ์„ ์ˆ˜ ์žˆ๋‹ค. ๊ทธ๋Ÿฌ๋‚˜ ์Šค๋„ค์ดํฌ๋Š” ์›๋ชจ์–‘์„ ๊ธฐ๋ฐ˜์œผ๋กœ ํ•˜๊ธฐ ๋•Œ๋ฌธ์— ์›์ด๋‚˜ ํƒ€์› ๋ชจ์–‘์ด ์•„๋‹Œ ๋‹ค๋ฅธ ๋ชจ์–‘์— ์žˆ์–ด์„œ ์œค๊ณฝ์„ ์ œ๋Œ€๋กœ ์ถ”์ถœํ•˜์ง€ ๋ชปํ•˜๋Š” ๋‹จ์ ์ด ์žˆ๋‹ค.

2.3 DRLSE ์•Œ๊ณ ๋ฆฌ์ฆ˜

์Šค๋„ค์ดํฌ๋Š” ๊ณ„์‚ฐ๋Ÿ‰์ด ์ ์–ด ๋น ๋ฅด๊ฒŒ ์œค๊ณฝ์„ ์ฐพ์„ ์ˆ˜ ์žˆ์ง€๋งŒ, ๊ฐœ์ฒด๊ฐ€ ์› ๋ชจ์–‘์ด ์•„๋‹Œ ๋ณต์žกํ•œ ๊ฒฝ์šฐ์—๋Š” ์ œ๋Œ€๋กœ ๊ฒฝ๊ณ„์— ๋„๋‹ฌํ•˜์ง€ ๋ชปํ•˜๋Š” ํ•œ๊ณ„๊ฐ€ ์žˆ๋‹ค. ํ•˜์ง€๋งŒ ์ด๋Ÿฌํ•œ ๋ณต์žกํ•œ ๊ฒฝ์šฐ์— ์˜์ƒ์„ ๋ถ„ํ• ํ•˜๋Š” ๋ฐฉ๋ฒ•์œผ๋กœ ๋ ˆ๋ฒจ์…‹๋ฐฉ๋ฒ•๊ณผ ์ด๋ฅผ ๋ณด์™„ํ•œ DRLSE (Direct Regularized Level Set Evolution) ์•Œ๊ณ ๋ฆฌ์ฆ˜์ด ์•Œ๋ ค์ ธ ์žˆ๋‹ค. Osher์™€ Shehian์— ์˜ํ•ด ์ฒ˜์Œ ์†Œ๊ฐœ๋œ ๋ ˆ๋ฒจ์…‹ ๋ฐฉ๋ฒ•(9)์€ ๊ณก์„ ์˜ ์•”์‹œ์  ํ‘œํ˜„ ๋ฐฉ๋ฒ•์„ ์ด์šฉํ•œ ๋ฐฉ์‹์œผ๋กœ ์ฒ˜์Œ์— ์„ค์ •ํ•œ ๊ณก์„ ์˜ ์ดˆ๊ธฐ ์œค๊ณฝ์„ ๋ช‡ ๋ฒˆ์˜ ์—…๋ฐ์ดํŠธ๋ฅผ ํ†ตํ•ด ๊ณก์„ ์„ ์ „๊ฐœํ•ด๋‚˜๊ฐ€๋ฉด์„œ ์ฐพ๊ณ ์ž ํ•˜๋Š” ๋ฌผ์ฒด์˜ ๊ฒฝ๊ณ„๋กœ ์œค๊ณฝ์„ ์ˆ˜๋ ดํ•ด ๋‚ด๊ฐ€๋Š” ๋ฐฉ์‹์ด๋‹ค. DRLSE๋Š” ๊ฑฐ๋ฆฌ ์ •๊ทœํ™”ํ•ญ์„ ๊ธฐ์กด ๋ ˆ๋ฒจ์…‹ ์—๋„ˆ์ง€ ํ•จ์ˆ˜์— ์ถ”๊ฐ€์‹œํ‚จ ๋ฐฉ๋ฒ•์ด๋‹ค.

๋ณธ ๋…ผ๋ฌธ์—์„œ๋Š” ๋ ˆ๋ฒจ์…‹(level set)ํ•จ์ˆ˜๋ฅผ ์ด์šฉํ•˜์—ฌ ์œค๊ณฝ์„ ์„ ๊ตฌ์„ฑํ•˜๋Š” Chan-Vese๋ชจ๋ธ(10)์„ ์ ์šฉํ•˜๊ณ , ์—๋„ˆ์ง€ ํ•จ์ˆ˜๋Š” DRLSE๋ฅผ ์‚ฌ์šฉํ•˜์˜€๋‹ค. DRLSE๋Š” ๊ธฐ์กด์˜ ์—๋„ˆ์ง€ํ•จ์ˆ˜์— ๊ฑฐ๋ฆฌ์ •๊ทœํ™”ํ•ญ์„ ํฌํ•จ์‹œ์ผœ ์ˆ˜ํ˜์˜ ์•ˆ์ •์„ฑ๊ณผ ์‹œ๊ฐ„์„ ํ–ฅ์ƒ์‹œํ‚จ๋‹ค.

ร˜: โ„ฆโ†’R ๋ฅผ ๋„๋ฉ”์ธ โ„ฆ์— ์ •์˜๋œ ๋ ˆ๋ฒจ์…‹ ํ•จ์ˆ˜๋ผ๊ณ  ํ•œ๋‹ค. ๋ ˆ๋ฒจ์…‹ ํ•จ์ˆ˜๋Š” ๊ทธ๋ฆผ. 2์™€ ๊ฐ™์ด ํ‘œํ˜„ํ•  ์ˆ˜ ์žˆ๋‹ค. ์œค๊ณฝ์„  ๊ธฐ์ค€ ์•ˆ์ชฝ๋ถ€๋ถ„์€ ร˜ > 0, ๋ฐ”๊นฅ์ชฝ ๋ถ€๋ถ„์€ ร˜ < 0, ์œค๊ณฝ์„ ์€ ร˜=0์ด๋ผ๊ณ  ์ •์˜ํ•œ๋‹ค. ๋ ˆ๋ฒจ์…‹ ํ•จ์ˆ˜๋ฅผ ์ด์šฉํ•˜์—ฌ ร˜=0์ธ ์ง€์ ์„ ์ฐพ์•„ ์œค๊ณฝ์„ ์„ ์ฐพ๋Š”๋‹ค. ์ด๋•Œ ๋ ˆ๋ฒจ์…‹ ํ•จ์ˆ˜์— ์˜ํ•œ ์—๋„ˆ์ง€ํ•จ์ˆ˜(E(ร˜))๋ฅผ ๋‹ค์Œ๊ณผ ๊ฐ™์ด ์ •์˜ํ•œ๋‹ค.

๊ทธ๋ฆผ. 2. ๋ ˆ๋ฒจ์…‹ ํ•จ์ˆ˜

Fig. 2. Level set function

../../Resources/kiee/KIEE.2018.67.12.1678/fig2.png

(7)
E ( ฯ• ) = ฮผ R p ( ฯ• ) + E e x t ( ฯ• )

์—ฌ๊ธฐ์„œ ฮผ๋Š” 0๋ณด๋‹ค ํฐ ์‹ค์ˆ˜, Rp(ร˜)๋Š” ๊ฑฐ๋ฆฌ ์ •๊ทœํ™”ํ•ญ์œผ๋กœ ๋ ˆ๋ฒจ์…‹ํ•จ์ˆ˜ ร˜๊ฐ€ ๊ฑฐ๋ฆฌ ํ•จ์ˆ˜์™€ ์–ผ๋งˆ๋‚˜ ๊ฐ€๊นŒ์šด์ง€ ๋‚˜ํƒ€๋‚ด๋Š” ํ•ญ์œผ๋กœ ์—๋„ˆ์ง€ ๋ฐ€๋„ ํ•จ์ˆ˜ p๋ฅผ ์ด์šฉํ•˜์—ฌ ๋‹ค์Œ๊ณผ ๊ฐ™์ด ์ •์˜๋œ๋‹ค.

(8)
R p ( ฯ• ) = 1 2 โˆซ ฮฉ p ( | โˆ‡ ฯ• | ) 2 d x

Eext๋Š” ์ด๋ฏธ์ง€์— ๋Œ€ํ•œ ์™ธ๋ถ€์—๋„ˆ์ง€๋กœ ์‹(9)๊ณผ ๊ฐ™์ด ์ •์˜๋œ๋‹ค.

(9)
E e x t ( ฯ• ) = โˆซ ฮฉ | u 0 - c 1 | 2 H ( ฯ• ) d x + โˆซ ฮฉ | u 0 - c 2 | 2 ( 1 - H ( ฯ• ) ) d x

ํ•จ์ˆ˜ H๋Š” Heavisideํ•จ์ˆ˜์ด๊ณ , ์ƒ์ˆ˜ c1, c2๋Š” ๊ทธ ์˜์—ญ์—์„œ์˜ ํ‰๊ท ๊ฐ’์„ ์˜๋ฏธํ•œ๋‹ค. ์˜์ƒ์˜ ์ด๋ฏธ์ง€๋ถ„ํ• ์˜ ๊ฒฝ์šฐ ๋” ์ข‹์€ ํšจ๊ณผ๋ฅผ ๋‚ด๊ธฐ ์œ„ํ•ด ์—ฃ์ง€(edge)๊ธฐ๋ฐ˜ ์ •๋ณด๋ฅผ ์‚ฌ์šฉํ•˜์˜€๋‹ค. ๊ทธ๋Ÿฌ๊ธฐ ์œ„ํ•ด ์„ ํ–‰์ž‘์—…์œผ๋กœ ์—ฃ์ง€(edge)์— ๋Œ€ํ•œ ์ตœ์ ํ™” ์ž‘์—…์ด ํ•„์š”ํ•˜๋‹ค. ์—ฃ์ง€ ํ•จ์ˆ˜ ๋Š” ๋‹ค์Œ๊ณผ ๊ฐ™์ด ์ •์˜ํ•œ๋‹ค.

(10)
g = 1 1 + โˆ‡ G ฯƒ * u 0 2

Gฯƒ๋Š” ํ‘œ์ค€ ํŽธ์ฐจ ฯƒ๋ฅผ ๊ฐ–๋Š” ๊ฐ€์šฐ์‹œ์•ˆ ์ปค๋„๋กœ ์‚ฌ์šฉ๋˜๋Š” ์ปจ๋ณผ๋ฃจ์…˜์€ ๋…ธ์ด์ฆˆ๋ฅผ ์ค„์ด๊ธฐ ์œ„ํ•ด ์ด๋ฏธ์ง€๋ฅผ ๋ถ€๋“œ๋Ÿฝ๊ฒŒ ํ•˜๋Š”๋ฐ ์‚ฌ์šฉ๋œ๋‹ค. ์—ฃ์ง€ ํ•จ์ˆ˜ g๋Š” ์—ฃ์ง€์—์„œ 0์— ๊ฐ€๊นŒ์šด ํ•จ์ˆ˜ ๊ฐ’์„ ๊ฐ–๋Š”๋‹ค. ๋ ˆ๋ฒจ์…‹ ํ•จ์ˆ˜๋ฅผ ๋‹ค์‹œ ์žฌ์ •์˜๋ฅผ ํ•˜๋ฉด ๋‹ค์Œ๊ณผ ๊ฐ™๋‹ค(11).

(11)
E ( ฯ• ) = ฮผ R p ( ฯ• ) + ฮป L g ( ฯ• ) + ฮฑ A g ( ฯ• )

(12)
L g ( ฯ• ) = โˆซ ฮฉ g ฮด ( ฯ• ) | โˆ‡ ฯ• | d x

(13)
A g ( ฯ• ) = โˆซ ฮฉ g H ( - ฯ• ) d x

์Šค๋„ค์ดํฌ ์•Œ๊ณ ๋ฆฌ์ฆ˜๊ณผ DRLSE์„ ์ด์šฉํ•˜์—ฌ ๊ฐ๊ฐ ์ •์˜๋œ ์—๋„ˆ์ง€ํ•จ์ˆ˜๊ฐ€ ๊ฐ€์žฅ ์ตœ์†Œํ™”๋˜๋Š” ์œค๊ณฝ์„ ์„ ์ฐพ์•„๋‚ธ ๊ฒƒ์ด ๋ชฉ์ ์ด๋‹ค.

3. ๊ฒฐ ๊ณผ

๋ณธ ๋…ผ๋ฌธ์—์„œ๋Š” ์„œ๋กœ ๋‹ค๋ฅธ ํ˜•ํƒœ์™€ ๋ชจ์–‘์„ ๊ฐ€์ง„ ํŒฌํ…€์น˜์•„์˜ X์„  ์˜์ƒ์—์„œ ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ์ ์šฉํ•˜์—ฌ ์น˜์•„์™€๋™์„ ๊ฒ€์ถœํ•˜์˜€๋‹ค. ํŒฌํ…€์น˜์•„๋Š” ์ „์น˜ 2๊ฐœ, ๊ฒฌ์น˜ 1๊ฐœ, ์†Œ๊ตฌ์น˜ 2๊ฐœ, ๋Œ€๊ตฌ์น˜ 2๊ฐœ ์ด 7๊ฐœ์˜ ์น˜์•„๋กœ ๊ตฌ์„ฑ๋˜์–ด ์žˆ๋‹ค. ํŒฌํ…€์น˜์•„๋Š” ์ œ 3๋Œ€๊ตฌ์น˜๋ฅผ ์ œ์™ธํ•œ ์„ฑ์ธ ์น˜์•„ 1๋ฒˆ ์ค‘์ ˆ์น˜๋ถ€ํ„ฐ 7๋ฒˆ ์ œ2๋Œ€๊ตฌ์น˜๊นŒ์ง€ 7์ข…์˜ ์ž์—ฐ์น˜์•„๋ฅผ ํ™•๋ณดํ•œ ํ›„์—, ์ธ์œ„์ ์œผ๋กœ ๋‹ค๋ฅธ ํ˜•ํƒœ์˜ ๊ตฌ๋ฉ์„ ๋งŒ๋“ค์—ˆ๋‹ค. X-์„  ์˜์ƒํš๋“์„ ์œ„ํ•ด ๊ตฌ๋‚ด X-์„  ๋ฐœ์ƒ์žฅ์น˜(๋ชจ๋ธ๋ช…: CS 2100, ์ œ์กฐํšŒ์‚ฌ: Carestream Dental LLC, USA)๋ฅผ ์ด์šฉํ•˜์˜€๊ณ , X-์„  ์กฐ์‚ฌ ์กฐ๊ฑด์€ ๊ด€์ „์•• 60kV, ๊ด€์ „๋ฅ˜ 7mA, ๊ทธ๋ฆฌ๊ณ  ๋…ธ์ถœ์‹œ๊ฐ„ 0.125s์œผ๋กœ ์กฐ์ •ํ•˜์˜€๋‹ค(3). ์น˜์•„๋งˆ๋‹ค 2๊ฐœ์˜ ์™€๋™์ด ์กด์žฌํ•˜๋ฉฐ ์ด 14๊ฐœ์˜ ์™€๋™์— ๋Œ€ํ•ด ์Šค๋„ค์ดํฌ ์•Œ๊ณ ๋ฆฌ์ฆ˜๊ณผ DRLSE ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ๊ฐ๊ฐ ์ ์šฉํ•˜์˜€๋‹ค. ์•Œ๊ณ ๋ฆฌ์ฆ˜์€ Intelยฎ Xeonยฎ Silver 4110 Processor, 128GB DDR4 RAM, NVIDIA Titan XP์˜ ํ™˜๊ฒฝ์—์„œ ์ง„ํ–‰ํ•˜์˜€๋‹ค. ์Šค๋„ค์ดํฌ๋Š” ์ดˆ๋ก์ƒ‰, DRLSE๋Š” ๋นจ๊ฐ„์ƒ‰์œผ๋กœ ๋‚˜ํƒ€๋‚ด์—ˆ๋‹ค. ๊ทธ๋ฆผ. 3๋ฅผ ๋ณด๋ฉด ์ด ์‹คํ—˜์˜ ์ „์ฒด์ ์ธ ์‹œ์Šคํ…œ์˜ ํ๋ฆ„์„ ํ™•์ธํ•  ์ˆ˜ ์žˆ๋‹ค. ์‚ฌ์šฉ์ž์— ์˜ํ•ด ROI (Region of Interest)๋ฅผ ์„ค์ •ํ•˜๋ฉด, ์ „์ฒ˜๋ฆฌ ๊ณผ์ •์„ ๊ฑฐ์ณ ์Šคํ…Œ์ดํฌ ์•Œ๊ณ ๋ฆฌ์ฆ˜๊ณผ DRLSE ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ์ ์šฉํ•œ๋‹ค. ์ดˆ๊ธฐ ์œค๊ณฝ์„  ์„ค์ •์€ ์œค๊ณฝ์„ ์˜ ์ถ”์ถœ๊ฒฐ๊ณผ์— ๋งŽ์€ ์˜ํ–ฅ์„ ์ค€๋‹ค. ๋ณธ ๋…ผ๋ฌธ์—์„œ๋Š” ๋ณ‘๋ณ€์„ ์ถฉ๋ถ„ํžˆ ํฌํ•จํ•˜๋Š” ์‚ฌ๊ฐํ˜•์„ ์„ค์ •ํ•˜๊ณ , ๋‘ ์•Œ๊ณ ๋ฆฌ์ฆ˜์— ๋™์ผํ•˜๊ฒŒ ์ ์šฉํ•˜์˜€๋‹ค. ์•Œ๊ณ ๋ฆฌ์ฆ˜๋“ค์€ ์ดˆ๊ธฐ ์œค๊ณฝ์„ ์œผ๋กœ๋ถ€ํ„ฐ ์ตœ์†Œํ™” ์—ฐ์‚ฐ์„ ๋ฐ˜๋ณตํ•˜์—ฌ ์˜ˆ์ƒ๋˜๋Š” ๋ณ‘๋ณ€์˜ ์œค๊ณฝ์„ ์„ ์ฐพ์„ ์ˆ˜ ์žˆ๋‹ค. ์Šค๋„ค์ดํฌ์˜ ๊ฐ€์ค‘์น˜๋Š” ์—ฐ์†์„ฑ์—๋„ˆ์ง€, ๊ณก๋ฅ  ์—๋„ˆ์ง€, ์ด๋ฏธ์ง€ ์—๋„ˆ์ง€๋กœ ์ด๋ฃจ์–ด์ ธ ์žˆ๋Š”๋ฐ ๊ฐ๊ฐ 0.45, 0.26, 0.28 ๋กœ ์„ค์ •ํ•˜์˜€๋‹ค. DRLSE์—์„œ๋Š” ์—๋„ˆ์ง€ ํ•จ์ˆ˜์˜ ํŒŒ๋ผ๋ฏธํ„ฐ ฮผ๋Š” 0.2/3, ฮป๋Š” 5, ฮฑ๋Š” 1.1, ์—ฃ์ง€ ํ•จ์ˆ˜ g์˜ ๊ฐ€์šฐ์‹œ์•ˆ ์ปค๋„ ํŒŒ๋ผ๋ฏธํ„ฐ ฯƒ๋ฅผ 20์œผ๋กœ ์„ค์ •ํ•˜์˜€๋‹ค. ์‹œ์Šคํ…œ์€ ๊ทธ๋ฆผ. 4๋Š” ํŒฌํ…€์น˜์•„ X์„  ์˜์ƒ์— ์Šค๋„ค์ดํฌ์™€ DRLSE ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ๊ฐ๊ฐ ์ ์šฉํ•œ ๊ฒฐ๊ณผ๋ฅผ ๋ณด์—ฌ์ค€๋‹ค.

๊ทธ๋ฆผ. 3. ์ „์ฒด ์‹œ์Šคํ…œ ํ๋ฆ„๋„

Fig. 3. Overview of the proposed system

../../Resources/kiee/KIEE.2018.67.12.1678/fig3.png

๊ทธ๋ฆผ. 4. ์Šค๋„ค์ดํฌ DRLSE ์•Œ๊ณ ๋ฆฌ์ฆ˜ ์ ์šฉ๊ฒฐ๊ณผ

Fig. 4. Detection result of Snake, DRLSE algorithm

../../Resources/kiee/KIEE.2018.67.12.1678/fig4.png

๊ทธ๋ฆผ. 4๋Š” ์Šค๋„ค์ดํฌ์™€ DRLSE ์•Œ๊ณ ๋ฆฌ์ฆ˜ ์ ์šฉ ๊ฒฐ๊ณผ ์˜์ƒ์ด๋‹ค. ๊ฐ ์น˜์•„์˜ ๋งจ ์™ผ์ชฝ ์˜์ƒ์€ ์น˜๊ณผ ์ „๋ฌธ์˜๊ฐ€ ํ‘œ์‹œํ•œ ์™€๋™์˜์—ญ(ground truth)์„ ๋‚˜ํƒ€๋‚ด๋ฉฐ ์ดํ›„ ์˜ค๋ฅธ์ชฝ์˜ ์˜์ƒ๋“ค์€ ์ฐจ๋ก€๋Œ€๋กœ ์Šค๋„ค์ดํฌ, DRLSE๋ฅผ ์ ์šฉํ•œ ์˜์ƒ์ด๋‹ค. ์Šค๋„ค์ดํฌ์™€ DRLSE์˜ ์ด๋ฏธ์ง€ ๋ถ„ํ• ํ˜•ํƒœ์˜ ๊ฐ€์žฅ ํฐ ํŠน์ง•์€ ์œค๊ณฝ์„ ์ด๋ฃจ๋Š” ํ˜•ํƒœ์ด๋‹ค. ๊ธฐ๋ณธ์ ์œผ๋กœ ์Šค๋„ค์ดํฌ๋Š” ์Šค๋„ค์ดํฌ ์—๋„ˆ์ง€ ํ•จ์ˆ˜์— ๊ธฐ๋ฐ˜ํ•˜์—ฌ ์›ํ˜•์˜ ํ˜•ํƒœ๋ฅผ ์ด๋ฃฌ๋‹ค. ์Šค๋„ค์ดํฌ ํฌ์ธํŠธ๋ฅผ ๊ธฐ์ค€์œผ๋กœ ํฌ์ธํŠธ ๊ฐ„์— ๊ฐ™์€ ๊ฐ„๊ฒฉ์ด ๋˜๋ฉด์„œ ์›ํ˜•์˜ ํ˜•ํƒœ๋ฅผ ์ด๋ฃจ๊ณ  ๊ฐ™์€ ๊ฐ„๊ฒฉ์ด ๋˜๊ฑฐ๋‚˜ ๊ฒฝ๊ณ„์— ๋จธ๋ฌผ๋ €์„ ๋•Œ ๋” ์ด์ƒ ์ˆ˜๋ ดํ•˜์ง€ ์•Š๊ณ  ๋ฉˆ์ถ”๋Š” ํŠน์ง•์ด ์žˆ๋‹ค. ์ด๋Ÿฌํ•œ ํŠน์ง•์€ ์ฐพ๊ณ ์ž ํ•˜๋Š” ๋ณ‘๋ณ€์ด ์›ํ˜•์ด๋‚˜ ํƒ€์›ํ˜•์ด ์•„๋‹ ๊ฒฝ์šฐ ์ œ๋Œ€๋กœ ์ฐพ์ง€ ๋ชปํ•œ๋‹ค. ๊ทธ๋ฆผ. 4์˜ 1๋ฒˆ ์ „์น˜์™€ 6๋ฒˆ ๋Œ€๊ตฌ์น˜์˜ ๊ฒฝ์šฐ ์™€๋™์˜ ํ˜•ํƒœ๊ฐ€ ๋น„๊ต์  ์›ํ˜•์„ ๋„์–ด ์ •ํ™•ํ•˜๊ฒŒ ์ฐพ์•˜๋‹ค. ํ•˜์ง€๋งŒ 5๋ฒˆ ์†Œ๊ตฌ์น˜์˜ ์™ผ์ชฝ ์™€๋™์˜ ์‚ฌ์ง„์„ ๋ณด๋ฉด ์™€๋™ ๊ฒฝ๊ณ„์— ์ œ๋Œ€๋กœ ์ˆ˜๋ ดํ•˜์ง€ ๋ชปํ•˜๊ณ  ์›ํ˜•์œผ๋กœ ์ˆ˜๋ ดํ•œ ๊ฒƒ์„ ๋ณผ ์ˆ˜ ์žˆ๋‹ค. DRLSE๋Š” ์Šค๋„ค์ดํฌ์™€ ๋‹ฌ๋ฆฌ ํ˜•ํƒœ๊ฐ€ ๊ณ ์ •๋˜์ง€ ์•Š๊ณ  ์—๋„ˆ์ง€ ํฌํ™”ํ˜„์ƒ์ด ์—†๊ธฐ ๋•Œ๋ฌธ์— ๋” ์„ธ๋ฐ€ํ•˜๊ฒŒ ๋ณ‘๋ณ€์„ ์ฐพ์„ ์ˆ˜ ์žˆ์—ˆ๋‹ค. ๋Œ€๋ถ€๋ถ„์˜ ์น˜์•„์—์„œ DRLSE๊ฐ€ ์Šค๋„ค์ดํฌ์— ๋น„ํ•ด ๋ชจ์„œ๋ฆฌ ๋ถ€๋ถ„๊นŒ์ง€ ๋ณ‘๋ณ€ ํƒ์ง€ ํ•˜๋Š” ๊ฒƒ์„ ํ™•์ธ ํ•  ์ˆ˜ ์žˆ๋‹ค. 4๋ฒˆ ์†Œ๊ตฌ์น˜์˜ ๊ฒฝ์šฐ๋„ ์–‘์ชฝ ์™€๋™ ๋ชจ๋‘ ์น˜์•„ ๋‚ด๋ถ€ ์œค๊ณฝ์—์„œ ๋” ํŒŒ๊ณ ๋“ฆ์„ ํ™•์ธํ•  ์ˆ˜๊ฐ€ ์žˆ๋‹ค. ๋‘ ์•Œ๊ณ ๋ฆฌ์ฆ˜ ๋ชจ๋‘ ๋ฐฐ๊ฒฝ๊ณผ ๊ฐ์ฒด์˜ ๊ฒฝ๊ณ„์— ์žˆ์–ด์„œ๋Š” ์ž˜ ๊ตฌ๋ณ„์€ ํ•ด๋‚ด์—ˆ๋‹ค. ํ•˜์ง€๋งŒ ์ผ๋ถ€ ์น˜์•„์˜ ๋‚ด๋ถ€๊ฒฝ๊ณ„์— ์žˆ์–ด ์ž˜ ์ฐพ์ง€ ๋ชปํ•œ ๊ฒฝ์šฐ๊ฐ€ ์žˆ์—ˆ๋‹ค. ๊ทธ ์ด์œ ๋Š” ์น˜์•„์™€ ์™ธ๋ถ€ ๊ฒฝ๊ณ„๋Š” ๋ช…์•”์— ์ฐจ์ด๊ฐ€ ๋‘๋“œ๋Ÿฌ์ง€๊ฒŒ ์ฐจ์ด๊ฐ€ ๋‚˜์ง€๋งŒ ๋‚ด๋ถ€๊ฒฝ๊ณ„์— ์žˆ์–ด์„œ๋Š” ์™€๋™๊ณผ ์น˜์•„ ๋ช…์•”์ฐจ์ด๊ฐ€ ๋ฏธ๋น„ํ•˜์—ฌ ์œค๊ณฝ์„ ์ •ํ™•ํ•˜๊ฒŒ ์ฐพ์ง€ ๋ชปํ•˜์˜€๋‹ค. ๋˜ํ•œ ์™€๋™์ด ํด ๊ฒฝ์šฐ ์ผ๋ถ€ ์œค๊ณฝ์—์„œ ๋” ์ด์ƒ ์ค„์ง€ ์•Š๊ณ  ๋ฉˆ์ถ”๋Š” ์˜ค๋ฅ˜๊ฐ€ ์žˆ์—ˆ๋‹ค.

ํ‘œ 1์€ ์•Œ๊ณ ๋ฆฌ์ฆ˜ ์‹คํ–‰์‹œ๊ฐ„์„ ์ธก์ •ํ•œ ํ‘œ์ด๋‹ค. ์Šค๋„ค์ดํฌ๋Š” DRLSE์— ๋น„ํ•ด ๊ณ„์‚ฐ๋Ÿ‰์ด ์ ์–ด ์‹คํ–‰์‹œ๊ฐ„์ด ๋น ๋ฅด๋‹ค. ํ‘œ 1์„ ๋ณด๋ฉด ์Šค๋„ค์ดํฌ๋Š” ํ‰๊ท  7.43์ดˆ DRLSE๋Š” 76.42์ดˆ๋กœ ์•ฝ 10๋ฐฐ ์ •๋„ ๋น ๋ฅด๋‹ค.

ํ‘œ 1. ์•Œ๊ณ ๋ฆฌ์ฆ˜ ์‹คํ–‰์‹œ๊ฐ„ ๋น„๊ต

Table 1. Runtime comparison for two algorithms

์†Œ์š”์‹œ๊ฐ„(sec)

Snake

DRLSE

์น˜์•„์ข…๋ฅ˜

์™ผ์ชฝ

์˜ค๋ฅธ์ชฝ

์™ผ์ชฝ

์˜ค๋ฅธ์ชฝ

1.์ „์น˜

9

7

86

90

2.์ „์น˜

7

8

64

67

3.๊ฒฌ์น˜

9

6

84

81

4.์†Œ๊ตฌ์น˜

7

7

68

68

5.์†Œ๊ตฌ์น˜

8

6

67

65

6.๋Œ€๊ตฌ์น˜

7

8

92

86

7.๋Œ€๊ตฌ์น˜

7

8

76

76

ํ‘œ 2๋Š” ์น˜๊ณผ ์ „๋ฌธ์˜๊ฐ€ ํ‘œ์‹œํ•œ ์™€๋™์˜์—ญ๊ณผ ์•Œ๊ณ ๋ฆฌ์ฆ˜์˜ ์„ฑ๋Šฅ์„ ๋น„๊ตํ•œ ํ‘œ์ด๋‹ค. ์น˜๊ณผ ์ „๋ฌธ์˜์˜ ์™€๋™์˜์—ญ์„ ๋ณด์ •ํ•˜๊ธฐ ์œ„ํ•ด ํŒŒ๋ผ๋ฏธํ„ฐ ฯ„๋ฅผ ์„ค์ •ํ•˜์—ฌ ์น˜๊ณผ ์ „๋ฌธ์˜๊ฐ€ ํ‘œ์‹œํ•œ ์™€๋™์˜์—ญ์„ ๊ธฐ์ค€์œผ๋กœ ์–‘์ชฝ 2ฯ„๋งŒํผ ๋ฒ”์œ„๋ฅผ ์„ค์ •ํ•ด์ฃผ์—ˆ๋‹ค. ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ฑ๋Šฅ์„ ์ •๋Ÿ‰์ ์œผ๋กœ ๋น„๊ตํ•˜๊ธฐ ์œ„ํ•ด ์‹(14)๊ณผ ๊ฐ™์ด ์ •ํ™•๋„๋ฅผ ์ •์˜ํ•˜์˜€๋‹ค.

ํ‘œ 2. ์•Œ๊ณ ๋ฆฌ์ฆ˜ ์„ฑ๋Šฅ ๋น„๊ต

Table 2. Accuracy comparison for two algorithms

Accuracy(ฯ„=3)

Snake

DRLSE

์น˜์•„์ข…๋ฅ˜

์™ผ์ชฝ

์˜ค๋ฅธ์ชฝ

์™ผ์ชฝ

์˜ค๋ฅธ์ชฝ

1.์ „์น˜

0.2648

0.5048

0.4564

0.4065

2.์ „์น˜

0.6437

0.8333

0.4812

0.8097

3.๊ฒฌ์น˜

0.6485

0.5906

0.6795

0.8417

4.์†Œ๊ตฌ์น˜

0.7595

0.6302

0.9222

0.5967

5.์†Œ๊ตฌ์น˜

0.2721

0.5714

0.3713

0.4309

6.๋Œ€๊ตฌ์น˜

0.8626

0.6997

0.668

0.9333

7.๋Œ€๊ตฌ์น˜

0.404

0.3612

0.0361

0.398

ํ‰๊ท 

0.5748

0.5944

๋Š” ์•Œ๊ณ ๋ฆฌ์ฆ˜์ด ์ฐพ์€ ํ”ฝ์…€ ์ˆ˜, ๋Š” ์•Œ๊ณ ๋ฆฌ์ฆ˜์ด ์ฐพ์€ ์œค๊ณฝ์„ ์˜ ํ”ฝ์…€ ์ค‘ ฯ„ ๋ฒ”์œ„ ๋‚ด์— ์†ํ•˜๋Š” ํ”ฝ์…€ ์ˆ˜์ด๋‹ค. ์‹คํ—˜๊ฒฐ๊ณผ ์ตœ์ข…์ ์œผ๋กœ ์‹คํ—˜์— ์ ์šฉํ•œ ฯ„๋Š” 3์ด๋‹ค. ฯ„๊ฐ€ 2์ผ ๋•Œ ์Šค๋„ค์ดํฌ ํ‰๊ท  ์ •ํ™•๋„๋Š” 0.4343, DRLSE 0.4469๋กœ ๋น„๊ต์  ๋‚ฎ์œผ๋ฉฐ, ฯ„๊ฐ€ 4์ผ ๋•Œ ์Šค๋„ค์ดํฌ๋Š” 0.6557, DRLSE๋Š” 0.6812์ด์˜€๋‹ค.

(14)
A c c u r a c y = P i x e l t r u e P i x e l d e t

4. ๊ฒฐ ๋ก 

๋ณธ ๋…ผ๋ฌธ์—์„œ๋Š” ์น˜์•„์™€๋™์„ ๊ฒ€์ถœํ•˜๊ธฐ ์œ„ํ•ด ์Šค๋„ค์ดํฌ, DRLSE ๋‘ ๊ฐ€์ง€์˜ ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ์ ์šฉํ•˜๊ณ  ๊ฒฐ๊ณผ๋ฅผ ๋น„๊ตํ•˜์˜€๋‹ค.

์ „์ฒ˜๋ฆฌ ๊ณผ์ •์€ ์ด๋ฏธ์ง€ ๋ถ„ํ•  ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ์ ์šฉ ์ „ ์žก์Œ์„ ์ œ๊ฑฐํ•˜๊ณ  ์˜์ƒ์˜ ๊ฒฝ๊ณ„๋ฅผ ๊ฐ•์กฐํ•˜๋Š” ์—ญํ• ์„ ํ•œ๋‹ค. ์ „์ฒ˜๋ฆฌ๊ณผ์ •์œผ๋กœ ๋ช…์•” ๋Œ€๋น„ ์ŠคํŠธ๋ ˆ์นญ, ์–‘๋ฐฉํ–ฅ ํ•„ํ„ฐ, ์บ๋‹ˆ ์—ฃ์ง€๊ฐ€ ์‚ฌ์šฉ๋˜์—ˆ๋‹ค. ๋ช…์•” ๋Œ€๋น„์ŠคํŠธ๋ ˆ์นญ์€ ์˜์ƒ์˜ ํžˆ์Šคํ† ๊ทธ๋žจ ๋ถ„ํฌ๋ฅผ ๊ณ ๋ฅด๊ฒŒ ํ•˜์—ฌ ์˜์ƒ์ด ๋” ์ž˜ ๋ณด์ด๋„๋ก ํ•˜์˜€์œผ๋ฉฐ, ์–‘๋ฐฉํ–ฅ ํ•„ํ„ฐ๋ฅผ ์‚ฌ์šฉํ•˜์—ฌ ์žก์Œ์ œ๊ฑฐ์™€ ์—ฃ์ง€๋ฅผ ๋” ๋„๋“œ๋ผ์ง€๊ฒŒ ํ•˜์˜€๋‹ค. ์บ๋‹ˆ ์—ฃ์ง€๋Š” ์—ฃ์ง€๋งŒ์„ ์ถ”์ถœํ•˜์—ฌ ์•Œ๊ณ ๋ฆฌ์ฆ˜์ด ์—ฃ์ง€๋ฅผ ์ฐพ์„ ์ˆ˜ ์žˆ๊ฒŒ ํ•˜์˜€๋‹ค. ์ดํ›„ ์ด๋ฏธ์ง€ ๋ถ„ํ• ์—๋Š” ์Šค๋„ค์ดํฌ ์•Œ๊ณ ๋ฆฌ์ฆ˜๊ณผ DRLSE ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ์ด์šฉํ•˜์˜€์œผ๋ฉฐ, ๊ฐ ์•Œ๊ณ ๋ฆฌ์ฆ˜๋“ค์˜ ์ •์˜๋œ ์—๋„ˆ์ง€ํ•จ์ˆ˜๊ฐ€ ๊ฐ€์žฅ ์ตœ์†Œํ™”๋˜๋Š” ์œค๊ณฝ์„ ์„ ์ฐพ์•„๋‚ด์—ˆ๋‹ค. ๊ทธ ๊ฒฐ๊ณผ ์Šค๋„ค์ดํฌ๋Š” ์—๋„ˆ์ง€ ํ•จ์ˆ˜์— ๊ธฐ๋ฐ˜ํ•˜์—ฌ ์ดˆ๊ธฐ ์Šค๋„ค์ดํฌ ํฌ์ธํŠธ๋“ค๊ฐ„ ๋“ฑ๊ฐ„๊ฒฉ์ด ๋˜์–ด ๊ฒฐ๊ตญ ์›ํ˜•์œผ๋กœ ์ˆ˜๋ ดํ•˜๊ฒŒ ๋œ๋‹ค. ์ด๋Ÿฌํ•œ ํŠน์„ฑ ๋•Œ๋ฌธ์— ์›ํ˜•์ด ์•„๋‹Œ ๋‹ค๋ฅธ ํ˜•ํƒœ์˜ ์™€๋™์„ ์ฐพ๋Š”๋ฐ ํ•œ๊ณ„๊ฐ€ ์žˆ์—ˆ๋‹ค. ๋ ˆ๋ฒจ์…‹ ๋ฐฉ๋ฒ•์„ ๊ธฐ๋ฐ˜ํ•œ DRLSE๋Š” ๊ด€์‹ฌ์˜์—ญ์˜ ํ˜•ํƒœ ๋ณ€ํ™”์— ๋Œ€ํ•˜์—ฌ ์‰ฝ๊ฒŒ ์ ์‘ํ•  ์ˆ˜ ์žˆ๋Š” ์žฅ์ ์„ ๊ฐ€์ง€๊ณ  ์žˆ๋‹ค. ์Šค๋„ค์ดํฌ ๋ณด๋‹ค ๊ณ„์‚ฐ๋Ÿ‰์ด ๋งŽ์•„ ์‹œ๊ฐ„์ด ๊ฑธ๋ฆฌ์ง€๋งŒ ๋” ์„ธ๋ฐ€ํ•˜๊ฒŒ ์™€๋™์„ ์ฐพ์„ ์ˆ˜ ์žˆ์—ˆ๋‹ค.

ํ–ฅํ›„ ํŒฌํ…€์น˜์•„๊ฐ€ ์•„๋‹Œ ์‹ค์ œ์น˜์•„ X-์„  ์˜์ƒ์— ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ์ ์šฉํ•˜์—ฌ ์‹ค์ œ ํ™˜๊ฒฝ์—์„œ๋Š” ๋‘ ๊ฐ€์ง€ ์•Œ๊ณ ๋ฆฌ์ฆ˜์ด ์–ด๋–ป๊ฒŒ ๋‹ค๋ฅธ์ง€์— ๋Œ€ํ•œ ๋ถ„์„ ์—ฐ๊ตฌ๊ฐ€ ํ•„์š”ํ•˜๋‹ค. ๋˜ํ•œ ฯ„์— ๋”ฐ๋ผ ์ •ํ™•๋„๊ฐ€ ๋‹ฌ๋ผ์ง€๊ธฐ ๋•Œ๋ฌธ์— ์ตœ์ ์˜ ๊ฐ’์„ ๊ฒฐ์ •ํ•˜๋Š” ๊ฒƒ์— ๋Œ€ํ•œ ์—ฐ๊ตฌ๊ฐ€ ํ•„์š”ํ•˜๋‹ค.

๊ฐ์‚ฌ์˜ ๊ธ€

This work was supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT) (No. 2017R1E1A1A03070297). This research was supported by the MSIT(Ministry of Science and ICT), Korea, under the ITRC(Information Technology Research Center) support program(IITP-2018-0-01433) supervised by the IITP(Institute for Information & communications Technology Promotion).

References

1 
Ministry of Health and Welfare , 2016, Korea Centers for Disease Control and Prevention, Korea Health Statistics 2015Google Search
2 
Pretty I. A., Oh PK, 2006, Review-Caries detection and diagnosis : Novel technologies, in Journal Of Dentistry โ€“ Els evier, Vol. 34, pp. 727-739DOI
3 
Huh J., Nam H., Kim J., Park J., Shin S., Lee R., 2015, Studies of Automatic Dental Cavity Detection System as an Auxiliary Tool for Diagnosis of Dental Caries in Digital X-ray Image, Prog Med Phys, Vol. 26, pp. 52-58DOI
4 
Heo Chang-hoe, Kim Min-jeong, Cho Hyun-chong, 2016, A Study of Computer-aided Detection System for Dental Cavity on Digital X-ray Image, The transactions of The Korean Institute of Electrical Engineers, Vol. 65, No. 8, pp. 1424-1429Google Search
5 
Pitts N. B., Ekstrand K. R., The ICDAS Foundation , 2013, International Caries Detection and Assessment System (ICDAS) and its International Caries Classification and Management System (ICCMS) โ€“ Methods for staging of the caries process and enabling dentists to manage caries, ommunity Dent. Oral. Epidemiol, Vol. 41, pp. 41-52DOI
6 
Canny J., 1987, A computational approach to edge detection, in Readings in Computer Vision, ed: Elsevier, pp. 184-203DOI
7 
Green B., 2002, Canny edge detection tutorial, Retrieved: March, Vol. 6Google Search
8 
Kass M., Witkin A., Terzopoulos D., 1988, Snakes: Active contour models, International journal of computer vision, Vol. 1, pp. 321-331DOI
9 
OSHER Stanley, SETHIAN James A., 1988, Fronts propagating with curvature-dependent speed: algorithms based on Hamilton-Jacobi formulations, Journal of computational physics, Vol. 79, No. 1, pp. 12-49DOI
10 
Chan T. F., Vese L. A., 2001, Active contours without edges, IEEE Transactions on image processing, Vol. 10, pp. 266-277Google Search
11 
Li C., Xu C., Gui C., Fox M. D., 2010, Distance regularized level set evolution and its application to image segmentation, IEEE transactions on image processing, Vol. 19, pp. 3243-3254DOI

์ €์ž์†Œ๊ฐœ

๊น€ ๋Œ€ ํ•œ (Dae-han Kim)
../../Resources/kiee/KIEE.2018.67.12.1678/au1.png

1992๋…„ 06์›” 26์ผ์ƒ

2017๋…„๋„ ๊ฐ•์›๋Œ€ํ•™๊ต ์ „์ž๊ณตํ•™์ „๊ณต ์กธ์—…

2017๋…„~ํ˜„์žฌ ๊ฐ•์›๋Œ€ํ•™๊ต BIT ์˜๋ฃŒ์œตํ•ฉํ•™ ์„์‚ฌ๊ณผ์ •

E-mail : daehan@kangwon.ac.kr

ํ—ˆ ์ฐฝ ํšŒ (Chang-hoe Heo)
../../Resources/kiee/KIEE.2018.67.12.1678/au2.png

1990๋…„ 11์›” 12์ผ์ƒ

2015๋…„๋„ ๊ฐ•์›๋Œ€ํ•™๊ต ์ „์ž๊ณตํ•™์ „๊ณต ์กธ์—…

2017 ๊ฐ•์›๋Œ€ํ•™๊ต BIT ์˜๋ฃŒ์œตํ•ฉํ•™ ์กธ์—…(์„์‚ฌ)

2017~ํ˜„์žฌ ์šธ์‚ฐ๋Œ€ํ•™๊ต ์˜๊ณผ๋Œ€ํ•™ ์˜ํ•™๊ณผ ์—ฐ๊ตฌ์›

E-mail : gobonk07@gmail.com

์กฐ ํ˜„ ์ข… (Hyun-chong Cho)
../../Resources/kiee/KIEE.2018.67.12.1678/au3.png

2009๋…„ ๋ฏธ๊ตญ ํ”Œ๋กœ๋ฆฌ๋‹ค ๋Œ€ํ•™๊ต ์ „๊ธฐ์ปดํ“จํ„ฐ๊ณตํ•™๊ณผ ์กธ์—… (์„์‚ฌ, ๋ฐ•์‚ฌ)

2013๋…„~ํ˜„์žฌ ๊ฐ•์›๋Œ€ํ•™๊ต IT๋Œ€ํ•™ ์ „์ž๊ณตํ•™๊ณผ ๋ฐ BIT ์˜๋ฃŒ์œตํ•ฉํ•™ ์กฐ๊ต์ˆ˜