Style harmonization of panoramic radiography using deep learning.

IF 1.6 3区医学 Q3 DENTISTRY, ORAL SURGERY & MEDICINE

Oral Radiology Pub Date : 2024-10-29 DOI:10.1007/s11282-024-00782-2

Hak-Sun Kim, Jaejung Seol, Ji-Yun Lee, Sang-Sun Han, Jaejun Yoo, Chena Lee

{"title":"Style harmonization of panoramic radiography using deep learning.","authors":"Hak-Sun Kim, Jaejung Seol, Ji-Yun Lee, Sang-Sun Han, Jaejun Yoo, Chena Lee","doi":"10.1007/s11282-024-00782-2","DOIUrl":null,"url":null,"abstract":"Objectives: This study aimed to harmonize panoramic radiograph images from different equipment in a single institution to display similar styles.Methods: A total of 15,624 panoramic images were acquired using two different equipment: 8079 images from Rayscan Alpha Plus (R-unit) and 7545 images from Pax-i plus (P-unit). Among these, 222 image pairs (444 images) from the same patients comprised the test dataset to harmonize the P-unit images with the R-unit image style using CycleGAN. Objective evaluations included Frechet Inception Distance (FID) and Learned Perceptual Image Patch Similarity (LPIPS) assessments. Additionally, expert evaluation was conducted by two oral and maxillofacial radiologists on transformed P-unit and R-unit images. The statistical analysis of LPIPS employed a Student's t-test.Results: The FID and mean LPIPS values of the transformed P-unit images (7.362, 0.488) were lower than those of the original P-unit images (8.380, 0.519), with a significant difference in LPIPS (p < 0.05). The experts evaluated 43.3-46.7% of the transformed P-unit images as R-unit images, 20.0-28.3% as P-units, and 28.3-33.3% as undetermined images.Conclusions: CycleGAN has the potential to harmonize panoramic radiograph image styles. Enhancement of the model is anticipated for the application of images produced by additional units.","PeriodicalId":56103,"journal":{"name":"Oral Radiology","volume":null,"pages":null},"PeriodicalIF":1.6000,"publicationDate":"2024-10-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Oral Radiology","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.1007/s11282-024-00782-2","RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"DENTISTRY, ORAL SURGERY & MEDICINE","Score":null,"Total":0}

引用次数: 0

Abstract

Objectives: This study aimed to harmonize panoramic radiograph images from different equipment in a single institution to display similar styles.

Methods: A total of 15,624 panoramic images were acquired using two different equipment: 8079 images from Rayscan Alpha Plus (R-unit) and 7545 images from Pax-i plus (P-unit). Among these, 222 image pairs (444 images) from the same patients comprised the test dataset to harmonize the P-unit images with the R-unit image style using CycleGAN. Objective evaluations included Frechet Inception Distance (FID) and Learned Perceptual Image Patch Similarity (LPIPS) assessments. Additionally, expert evaluation was conducted by two oral and maxillofacial radiologists on transformed P-unit and R-unit images. The statistical analysis of LPIPS employed a Student's t-test.

Results: The FID and mean LPIPS values of the transformed P-unit images (7.362, 0.488) were lower than those of the original P-unit images (8.380, 0.519), with a significant difference in LPIPS (p < 0.05). The experts evaluated 43.3-46.7% of the transformed P-unit images as R-unit images, 20.0-28.3% as P-units, and 28.3-33.3% as undetermined images.

Conclusions: CycleGAN has the potential to harmonize panoramic radiograph image styles. Enhancement of the model is anticipated for the application of images produced by additional units.

查看原文本刊更多论文

利用深度学习协调全景放射摄影的风格。

目的：本研究旨在统一来自同一机构不同设备的全景X光图像：本研究旨在统一一个机构中来自不同设备的全景放射影像，以显示相似的风格：使用两种不同的设备共采集了 15624 张全景图像：8079 张来自 Rayscan Alpha Plus（R-unit），7545 张来自 Pax-i plus（P-unit）。其中，来自同一患者的 222 对图像（444 幅图像）组成了测试数据集，以便使用 CycleGAN 协调 P-unit 图像与 R-unit 图像风格。客观评估包括弗雷谢特起始距离（FID）和学习感知图像补丁相似性（LPIPS）评估。此外，两位口腔颌面放射科医生还对转换后的 P 单位和 R 单位图像进行了专家评估。对 LPIPS 的统计分析采用了学生 t 检验：结果：转换后的 P 单位图像的 FID 值和 LPIPS 平均值（7.362，0.488）均低于原始 P 单位图像（8.380，0.519），且 LPIPS 差异显著（p 结论：CycleGAN 有潜力协调 P 单位图像和 R 单位图像的显示效果：CycleGAN 具有协调全景放射影像风格的潜力。预计该模型将在应用更多单位制作的图像时得到改进。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Oral Radiology DENTISTRY, ORAL SURGERY & MEDICINE-

CiteScore

4.20

自引率

13.60%

发文量

审稿时长

>12 weeks

期刊介绍： As the official English-language journal of the Japanese Society for Oral and Maxillofacial Radiology and the Asian Academy of Oral and Maxillofacial Radiology, Oral Radiology is intended to be a forum for international collaboration in head and neck diagnostic imaging and all related fields. Oral Radiology features cutting-edge research papers, review articles, case reports, and technical notes from both the clinical and experimental fields. As membership in the Society is not a prerequisite, contributions are welcome from researchers and clinicians worldwide.