Cross-vendor robustness of a hybrid deep learning framework for 4-stage periodontitis classification on panoramic radiographs

人工智能 射线照相术 稳健性(进化) 深度学习 分割 牙周炎 计算机科学 模式识别(心理学) 医学 Sørensen–骰子系数 人工神经网络 掷骰子 组内相关 口腔正畸科 计算机视觉 重复性 医学影像学 深层神经网络 全景片 牙科 均值漂移 领域(数学分析) 图像分割 机器学习 域适应 口腔颌面外科 卷积神经网络 管道(软件)
作者
Sang-Jeong Lee,Ji Yong Han,Kyung‐Hoe Huh,Min-Suk Heo,Sam-Sun Lee,Jo‐Eun Kim,Su Yang,Won-Jin Yi
出处
期刊:Dentomaxillofacial Radiology [Oxford University Press]
卷期号:: twag055-
标识
DOI:10.1093/dmfr/twag055
摘要

OBJECTIVES: Although deep learning for periodontitis diagnosis on panoramic radiographs has advanced rapidly, most studies use single-vendor data and cross-vendor generalization is rarely evaluated. Therefore, we evaluate the cross-vendor robustness of a hybrid CNN-CAD framework for automated four-stage periodontitis classification. METHODS: Five hundred panoramic radiographs were retrospectively collected from three vendors (Instrumentarium, n=400; Vatech, n=50; PointNix, n=50). The framework extends a previously published hybrid pipeline by adding a YOLO-based CNN for missing-teeth quantification, enabling four-stage classification according to the 2017 World Workshop criteria. Three dataset configurations were evaluated: pooled multi-vendor training and two leave-one-device-out (LODO) splits. We compared segmentation backbones (U-Net, Dense U-Net, SegNet, Mask R-CNN) and YOLO detector variants. Agreement with three oral and maxillofacial radiologists (3, 5 and 10 years of experience) was assessed using mean absolute difference (MAD), Pearson and intraclass correlation, Bland-Altman, and Passing-Bablok analyses. RESULTS: Under pooled multi-vendor training, Mask R-CNN achieved Dice coefficients of 0.96, 0.92 and 0.94 for periodontal bone level, cemento-enamel junction level and teeth/implants respectively; CNNv4-tiny reached a mean AP of 0.86 for missing teeth. The MAD between automated and expert staging was 0.31, overall image-level ICC was 0.93 (95% CI 0.86-0.97; p<0.01), and Bland-Altman bias against the most experienced radiologist was 0.007. Under LODO, Dice coefficient dropped to 0.75-0.86 (all p<0.001 versus pooled), quantifying substantial vendor-induced domain shift. CONCLUSIONS: The hybrid framework achieves expert-level agreement under pooled multi-vendor training but degrades in held-out vendors, providing a quantitative reference for vendor-induced domain shift in panoramic radiograph AI and motivating vendor-aware training or domain adaptation in future clinical deployments. ADVANCES IN KNOWLEDGE: This work provides, to our knowledge, the first cross-vendor benchmark for deep-learning-based periodontitis staging on panoramic radiographs and quantifies vendor-induced domain shift directly addressing the external-validation gap recently identified in this journal.
最长约 10秒,即可获得该文献文件

科研通智能强力驱动
Strongly Powered by AbleSci AI
科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
wuwui发布了新的文献求助10
1秒前
2秒前
2秒前
丁香鱼完成签到,获得积分10
3秒前
jzt12138完成签到,获得积分10
3秒前
yyz完成签到,获得积分10
3秒前
葡萄的应助被bigpluto采纳,获得10
4秒前
5秒前
cc发布了新的文献求助10
5秒前
6秒前
7秒前
AEGUO完成签到,获得积分10
8秒前
秋风的应助被百褶裙采纳,获得10
8秒前
9秒前
素愫完成签到,获得积分10
10秒前
tongtong发布了新的文献求助10
11秒前
科研通AI6.2的应助被u806686采纳,获得10
11秒前
无辜小花卷完成签到,获得积分10
12秒前
尘南浔发布了新的文献求助10
12秒前
12秒前
kuaikuailele完成签到,获得积分10
12秒前
12秒前
13秒前
13秒前
luyjabc完成签到,获得积分10
14秒前
钱念波完成签到 ,获得积分10
14秒前
14秒前
狂野迎海完成签到 ,获得积分10
15秒前
15秒前
鳗鱼元正发布了新的文献求助10
15秒前
15秒前
ZJ完成签到,获得积分0
15秒前
15秒前
小罗同学完成签到,获得积分10
16秒前
星辰大海的应助被快乐的忆安采纳,获得10
16秒前
xyb完成签到,获得积分20
16秒前
lilyvan完成签到 ,获得积分10
16秒前
17秒前
111122223333发布了新的文献求助10
18秒前
李健的小迷弟的应助被Tangerine采纳,获得10
18秒前
高分求助中
(应助此贴封号)【重要!!请各用户(尤其是新用户)详细阅读】【科研通的精品贴汇总】 10000
拉伸高分子熔体中缠结行为的模拟和小角中子散射研究 2000
Handbook of Sport Coaching Research 1000
Orthognathic Surgery - 2 Volume Set 500
高分子基功能界面水蒸发材料与应用研究 500
外场增强的水汽发电研究 500
基于结构调控的功能化水诱导发电材料与器件研究 500
热门求助领域 (近24小时)
化学 材料科学 医学 生物 纳米技术 计算机科学 工程类 有机化学 化学工程 生物化学 物理 复合材料 内科学 催化作用 光电子学 人工智能 细胞生物学 无机化学 物理化学 数学
热门帖子
关注 科研通微信公众号,转发送积分 8193605
求助须知:如何正确求助?哪些是违规求助? 9579751
关于积分的说明 21867369
捐赠科研通 7793242
什么是DOI,文献DOI怎么找? 3437325
关于科研通互助平台的介绍 2560761
邀请新用户注册赠送积分活动 2484193