Jose J Prabin, Jaffino G, Al Awadh Mohammed, Rao Koppula Srinivas, Yafang Yan, Sivalingam Krishna Moorthy
School of Electronics Engineering, Vellore Institute of Technology , Vellore, India.
School of Electronics Engineering, Vellore Institute of Technology, Vellore, India.
Sci Rep. 2025 Jul 29;15(1):27584. doi: 10.1038/s41598-025-10199-4.
The timely detection of abnormal health conditions is crucial in achieving successful medical intervention and enhancing patient outcomes. Despite advances in health monitoring, existing methods often struggle with achieving high accuracy, sensitivity, and specificity in real-time detection. This work addresses the need for improved performance in health monitoring systems in real time sensor data. In this work, real-time health monitoring data is obtained through the utilization of MAX 30102 and LM35 sensors, which capture the physiological features such as heart rate, blood oxygen levels and body temperature. The acquired data from these sensors is then transmitted to ThingSpeak, a cloud-based platform developed for the Internet of Things (IoT), where the data are analysed. In order to ensure consistency, the sensed features are subjected to a standardization process, ensuring they are scaled uniformly. In this work joint process estimator rider optimization algorithm (JPEROA) for Deep stack auto-encoder is proposed to perform the classification task. In JPEROA algorithm line coefficients and delay coefficients parameters are estimated to improve the performance of the system. The performance of the proposed method is compared with other five machine learning algorithms, including Support Vector Machine, Random Forest, Gradient Boosting, Naive Bayes, and Multilayer Perceptron neural networks. The proposed method also evaluated using PTB Diagnostic dataset signals. The performance of the algorithms is assessed using multiple performance metrics such as accuracy, sensitivity and specificity. The proposed method provides a maximum accuracy of 0.9625 and maximum sensitivity of 0.975 and specificity of 0.95.
及时检测异常健康状况对于实现成功的医疗干预和改善患者预后至关重要。尽管健康监测取得了进展，但现有方法在实时检测中往往难以实现高精度、高灵敏度和高特异性。这项工作满足了实时传感器数据健康监测系统中提高性能的需求。在这项工作中，通过使用MAX 30102和LM35传感器获取实时健康监测数据，这些传感器可捕捉心率、血氧水平和体温等生理特征。然后将从这些传感器获取的数据传输到ThingSpeak，这是一个为物联网（IoT）开发的基于云的平台，在该平台上对数据进行分析。为确保一致性，对感测到的特征进行标准化处理，确保它们均匀缩放。在这项工作中，提出了用于深度堆栈自动编码器的联合过程估计器骑手优化算法（JPEROA）来执行分类任务。在JPEROA算法中，估计线路系数和延迟系数参数以提高系统性能。将所提出方法的性能与其他五种机器学习算法进行比较，包括支持向量机、随机森林、梯度提升、朴素贝叶斯和多层感知器神经网络。还使用PTB诊断数据集信号对所提出的方法进行评估。使用诸如准确率、灵敏度和特异性等多个性能指标评估算法的性能。所提出的方法提供了0.9625的最大准确率、0.975的最大灵敏度和0.95的特异性。
