首页|基于深度学习下低视力患者辅具智能化适配研究

基于深度学习下低视力患者辅具智能化适配研究

扫码查看
目的:初步探讨构建移动出诊方式的农村地区低视力患者辅具智能化验配的神经网络模型.方法:选取2019年5月—2023年5月福建医科大学附属第二医院视障康复指导中心的728例患者为研究对象,结合描述性研究方法,构建基于神经网络算法的智能化辅具适配模型并进行验证.结果:关于低视力辅具的使用,71.15%的患者验配了不超过2种辅具,而28.85%的患者验配了3种及以上.在视力损害程度上,轻度视力损害占10.44%,中度视力损害占43.27%,严重视力损害占26.51%,失明占19.78%.经过综合分析,选取了准确度作为模型性能的主要评价标准,F1值作为辅助评价标准,在模型阈值为0.4时,准确度约为80%,F1值约为0.31,可作为模型的分类判断的阈值.结论:中国农村地区低视力患者辅具适配与视功能及生存质量、康复需求密切相关.构建农村地区低视力患者移动出诊方式的辅具智能化适配神经网络模型,具有一定临床应用价值.
Research on Intelligent Adaptation of Assistive Devices for Patients with Low Vision Based on Deep Learning
Objective:To establish a neural network model of intelligent assistive test for patients with low vision in rural areas with mobile home visit.Methods:A total of 728 patients from the Rehabilitation Guidance Center for the visually Impaired in the Second Affiliated Hospital of Fujian Medical University from May 2019 to May 2023 were selected as the research objects.Combined with descriptive research methods,an intelligent assistive device adaptation model based on neural network algorithm was constructed and verified.Results:Regarding the use of low vision aids,71.15%of patients had no more than two aids,while 28.85%of patients had three or more aids.In the degree of visual impairment,mild visual impairment accounted for 10.44%,moderate visual impairment accounted for 43.27%,severe visual impairment accounted for 26.51%,and blindness accounted for 19.78%.After comprehensive analysis,the accuracy was selected as the main evaluation standard of the model performance,and the F1 value was used as the auxiliary evaluation standard.When the model threshold was 0.4,the accuracy was about 80%,and the F1 value was about 0.31,which could be used as the threshold for classification and judgment of the model.Conclusion:The adaptation of assistive devices in patients with low vision in rural areas of China is closely related to visual function,quality of life and rehabilitation needs.It is of certain clinical application value to construct an intelligent adaptive neural network model of assistive devices for mobile out-patient mode of low vision patients in rural areas.

Low visionDeep learningNeural network modelAssistive device adaptation

戴炳发、阚遵琪、裴鹏鹏、叶文文、黄丽娟

展开 >

福建医科大学附属第二医院/视障辅助技术福建省高校工程研究中心,福建 泉州 362000

中国中医科学院广安门医院,北京 100053

长治医学院附属和济医院,山西 长治 046000

低视力 深度学习 神经网络模型 辅具适配

福建省卫生健康委员会青年科研基金

2020QNA061

2024

中国药物滥用防治杂志
中国药物滥用防治协会 军事医学科学院毒物药物研究所

中国药物滥用防治杂志

影响因子:0.584
ISSN:1006-902X
年,卷(期):2024.30(9)
  • 11