Pereira Clayton R, Pereira Danilo R, Silva Francisco A, Masieiro João P, Weber Silke A T, Hook Christian, Papa João P
Department of Computing, Federal University of São Carlos, Brazil.
University of Western São Paulo, Brazil.
Comput Methods Programs Biomed. 2016 Nov;136:79-88. doi: 10.1016/j.cmpb.2016.08.005. Epub 2016 Aug 26.
Even today, pointing out an exam that can diagnose a patient with Parkinson's disease (PD) accurately enough is not an easy task. Although a number of techniques have been used in search for a more precise method, detecting such illness and measuring its level of severity early enough to postpone its side effects are not straightforward. In this work, after reviewing a considerable number of works, we conclude that only a few techniques address the problem of PD recognition by means of micrography using computer vision techniques. Therefore, we consider the problem of aiding automatic PD diagnosis by means of spirals and meanders filled out in forms, which are then compared with the template for feature extraction.
In our work, both the template and the drawings are identified and separated automatically using image processing techniques, thus needing no user intervention. Since we have no registered images, the idea is to obtain a suitable representation of both template and drawings using the very same approach for all images in a fast and accurate approach.
The results have shown that we can obtain very reasonable recognition rates (around ≈67%), with the most accurate class being the one represented by the patients, which outnumbered the control individuals in the proposed dataset.
The proposed approach seemed to be suitable for aiding in automatic PD diagnosis by means of computer vision and machine learning techniques. Also, meander images play an important role, leading to higher accuracies than spiral images. We also observed that the main problem in detecting PD is the patients in the early stages, who can draw near-perfect objects, which are very similar to the ones made by control patients.
即便在当今，指出一项能够足够准确地诊断帕金森病（PD）患者的检查并非易事。尽管已经运用了多种技术来探寻更精确的方法，但要足够早地检测出这种疾病并测量其严重程度以延缓其副作用并非易事。在这项工作中，在回顾了大量研究后，我们得出结论，只有少数技术通过使用计算机视觉技术的显微图像来解决PD识别问题。因此，我们考虑借助表格中填充的螺旋线和曲折线来辅助自动PD诊断的问题，然后将其与用于特征提取的模板进行比较。
在我们的工作中，使用图像处理技术自动识别并分离模板和图形，因此无需用户干预。由于我们没有注册图像，想法是以快速且准确的方式对所有图像使用相同的方法来获得模板和图形的合适表示。
结果表明我们能够获得非常合理的识别率（约67%），在所提出的数据集中，最准确的类别是由患者代表的类别，其数量超过了对照个体。
所提出的方法似乎适用于借助计算机视觉和机器学习技术辅助自动PD诊断。此外，曲折线图像起着重要作用，导致比螺旋线图像更高的准确率。我们还观察到，检测PD的主要问题在于早期阶段的患者，他们能够画出与对照患者绘制的非常相似的近乎完美的图形。
