DETECTION OF ORAL MUCOSAL LESIONS AND MEASUREMENT OF THEIR AREAS FROM INTRAORAL PHOTOGRAPHS WITH DEEP LEARNING AND IMAGE PROCESSING TECHNIQUES

Project no. 32243 · Status: Kapanmış (Yayın bekleniyor) · 2024

Summary

<p> The&nbsp;aim&nbsp;of&nbsp;this&nbsp;study&nbsp;is&nbsp;to&nbsp;detect&nbsp;and&nbsp;mask&nbsp;oral&nbsp;mucosal&nbsp;lesions&nbsp;from&nbsp;intraoral&nbsp;images&nbsp;and&nbsp;to&nbsp;measure&nbsp;the&nbsp;lesion&nbsp;area&nbsp;with&nbsp;deep&nbsp;learning&nbsp;and&nbsp;image&nbsp;processing&nbsp;techniques,&nbsp;which&nbsp;are&nbsp;sub-branches&nbsp;of&nbsp;artificial&nbsp;intelligence.&nbsp;By&nbsp;measuring&nbsp;the&nbsp;lesion&nbsp;area,&nbsp;it&nbsp;is&nbsp;aimed&nbsp;to&nbsp;determine&nbsp;the&nbsp;change&nbsp;in&nbsp;the&nbsp;lesion&nbsp;over&nbsp;time&nbsp;and&nbsp;to&nbsp;help&nbsp;specialist&nbsp;physicians&nbsp;for&nbsp;treatment&nbsp;planning.&nbsp;In&nbsp;the&nbsp;literature,&nbsp;it&nbsp;is&nbsp;seen&nbsp;that&nbsp;there&nbsp;is&nbsp;no&nbsp;comprehensive&nbsp;study&nbsp;on&nbsp;intraoral&nbsp;lesion&nbsp;detection&nbsp;and&nbsp;segmentation,&nbsp;especially&nbsp;in&nbsp;disease&nbsp;detection&nbsp;studies,&nbsp;and&nbsp;there&nbsp;is&nbsp;no&nbsp;study&nbsp;that&nbsp;measures&nbsp;the&nbsp;area&nbsp;of&nbsp;the&nbsp;diseased&nbsp;area&nbsp;after&nbsp;detection.&nbsp;The&nbsp;contribution&nbsp;of&nbsp;this&nbsp;study&nbsp;to&nbsp;the&nbsp;literature&nbsp;is&nbsp;the&nbsp;detection&nbsp;of&nbsp;oral&nbsp;mucosal&nbsp;lesions&nbsp;from&nbsp;intraoral&nbsp;images&nbsp;and&nbsp;measurement&nbsp;of&nbsp;the&nbsp;lesion&nbsp;area.&nbsp;For&nbsp;this&nbsp;purpose,&nbsp;a&nbsp;unique&nbsp;data&nbsp;set&nbsp;was&nbsp;created&nbsp;with&nbsp;the&nbsp;contributions&nbsp;of&nbsp;expert&nbsp;physicians.&nbsp;In&nbsp;the&nbsp;study,&nbsp;lesion&nbsp;detection&nbsp;is&nbsp;performed&nbsp;with&nbsp;deep&nbsp;learning&nbsp;and&nbsp;the&nbsp;area&nbsp;of&nbsp;the&nbsp;lesion&nbsp;whose&nbsp;boundaries&nbsp;are&nbsp;determined&nbsp;can&nbsp;be&nbsp;measured.&nbsp;The&nbsp;measurements&nbsp;are&nbsp;compared&nbsp;with&nbsp;the&nbsp;measurements&nbsp;of&nbsp;ImageJ&nbsp;program.The&nbsp;Mask-RCNN&nbsp;algorithm&nbsp;is&nbsp;used&nbsp;in&nbsp;this&nbsp;study,&nbsp;and&nbsp;after&nbsp;training&nbsp;with&nbsp;intraoral&nbsp;image&nbsp;dataset,&nbsp;it&nbsp;is&nbsp;able&nbsp;to&nbsp;detect&nbsp;images&nbsp;with&nbsp;new&nbsp;lesions&nbsp;and&nbsp;indicate&nbsp;their&nbsp;boundaries&nbsp;with&nbsp;a&nbsp;mask.&nbsp;The&nbsp;Mask-RCNN&nbsp;algorithm&nbsp;was&nbsp;evaluated&nbsp;by&nbsp;considering&nbsp;the&nbsp;performance&nbsp;metrics&nbsp;commonly&nbsp;used&nbsp;in&nbsp;the&nbsp;literature.&nbsp;When&nbsp;these&nbsp;values&nbsp;are&nbsp;compared&nbsp;with&nbsp;the&nbsp;studies&nbsp;in&nbsp;the&nbsp;literature,&nbsp;it&nbsp;is&nbsp;seen&nbsp;that&nbsp;the&nbsp;Mask-RCNN&nbsp;algorithm&nbsp;used&nbsp;and&nbsp;developed&nbsp;in&nbsp;the&nbsp;study&nbsp;shows&nbsp;a&nbsp;good&nbsp;performance.</p>

View details