相关实验视频
Updated: Jan 31, 2026

Smartphone Fundus Photography
Published on: July 6, 2017
基于对学龄儿童查底部摄影数据的轴长度预测模型
Zixun Wang1, Hua Rong1, Jingtao Yu1
1Tianjin Key Laboratory of Retinal Functions and Diseases, Tianjin Branch of National Clinical Research Center for Ocular Disease, Eye Institute and School of Optometry, Tianjin Medical University Eye Hospital, Tianjin, China.
深度学习模型使用眼睛图像准确地预测儿童的轴长. 整合年龄和折射误差改善了预测,而性别信息降低了准确性.
科学领域:
- 眼科医生 眼科 眼科
- 人工智能的人工智能
- 儿科健康 儿科健康
背景情况:
- 轴长 (AL) 是儿科眼睛发育和折射误差进展的一个关键参数.
- 准确的AL预测对于监测和管理儿童近视至关重要.
- 目前用于AL测量的方法可能是侵入性的或需要专门的设备.
研究的目的:
- 开发和评估深度学习 (DL) 模型,用于预测学龄儿童的轴长 (AL).
- 评估将临床数据 (年龄,二极管球,性别) 与彩色底部照片 (CFPs) 整合用于AL预测的影响.
- 用热图来解释DL模型预测,以了解特征的重要性.
主要方法:
- 利用了2,779张6至10岁儿童的微小异常色底照片 (CFP).
- 采用ResNet101架构用于DL模型开发,将临床参数集成到完全连接的层中.
- 将数据分为培训 (70%),验证 (20%) 和测试 (10%) 集,以进行可靠的模型评估.
- 使用Grad-CAM热图来实现模型的解释性.
主要成果:
- 一个仅用于CFP的DL模型实现了高预测精度 (R2 = 0.70).
- 将年龄和二极管球 (DS) 与CFP集成,进一步提高了AL预测的准确性 (R2 = 0.75).
- 将性别信息与CFP,年龄和DS一起纳入,显著降低了预测效率 (R2 = 0.41).
- 热图表明,视网膜血管和周血管组织是预测AL的关键特征.
结论:
- 深度学习模型可以有效地使用彩色基底照片预测儿科轴长.
- 年龄和折射误差 (DS) 增强了基于DL的AL预测,而性别等分类变量可能会降低性能.
- DL模型利用细微的 fundus 血管变化来预测AL,为折射发育提供了洞察力.
更多相关视频
10:57Determining Soil-transmitted Helminth Infection Status and Physical Fitness of School-aged Children
Published on: August 22, 2012
08:05Measuring Statistical Learning Across Modalities and Domains in School-Aged Children Via an Online Platform and Neuroimaging Techniques
Published on: June 30, 2020
相关概念视频
Axial and Appendicular Muscles
Axial Muscles
Axial muscles, situated along the body's midline, are intricately connected to the axial skeleton, which includes the skull, spine, ribs, and sternum. These muscles facilitate facial expressions and...
Predicting Molecular Geometry
Overview of the Axial Skeleton
The axial skeleton of the...
General Case of Eccentric Axial Loading
Consider a member subjected to equal and opposite forces that are applied along a line that does not coincide with the member's neutral axis. In unsymmetrical...
Normal Strain under Axial Loading
Model Approaches for Pharmacokinetic Data: Physiological Models