利用人工智能技术和图像分析早期检测帕金森病
Early Detection of Parkinson's Disease Using AI Techniques and Image Analysis.
作者信息
Ianculescu Marilena, Petean Corina, Sandulescu Virginia, Alexandru Adriana, Vasilevschi Ana-Mihaela
机构信息
National Institute for Research and Development in Informatics, 011455 Bucharest, Romania.
Faculty of Electrical Engineering, Electronics and Information Technology, Valahia University of Targoviste, 130004 Targoviste, Romania.
出版信息
Diagnostics (Basel). 2024 Nov 21;14(23):2615. doi: 10.3390/diagnostics14232615.
BACKGROUND
Parkinson's disease (PD) diagnosis benefits significantly from advancements in artificial intelligence (AI) and image processing techniques. This paper explores various approaches for processing hand-drawn Archimedean spirals in order to detect signs of PD.
METHODS
The best approach is selected to be integrated in a neurodegenerative disease management platform called NeuroPredict. The most innovative aspects of the presented approaches are related to the employed feature extraction techniques that convert hand-drawn spirals into a frequency spectra, so that frequency features may be extracted and utilized as inputs for various classification algorithms. A second category of extracted features contains information related to the thickness and pressure of drawings.
RESULTS
The selected approach achieves an overall accuracy of 95.24% and allows acquiring new test data using only a pencil and paper, without requiring a specialized device like a graphic tablet or a digital pen.
CONCLUSIONS
This study underscores the clinical relevance of AI in enhancing diagnostic precision for neurodegenerative diseases.
背景
帕金森病（PD）的诊断因人工智能（AI）和图像处理技术的进步而受益匪浅。本文探索了处理手绘阿基米德螺旋线的各种方法，以检测帕金森病的迹象。
方法
选择最佳方法集成到一个名为NeuroPredict的神经退行性疾病管理平台中。所提出方法的最具创新性的方面与所采用的特征提取技术有关，该技术将手绘螺旋线转换为频谱，以便提取频率特征并将其用作各种分类算法的输入。提取的第二类特征包含与绘图的厚度和压力相关的信息。
结果
所选方法的总体准确率达到95.24%，并且仅使用铅笔和纸就能获取新的测试数据，无需绘图板或数字笔等专门设备。
结论
本研究强调了人工智能在提高神经退行性疾病诊断精度方面的临床相关性。
Zotero 插件