Parkinson's disease (PD) diagnosis based on speech data has been proved to be an effective way in recent years. There are still some problems on preprocessing samples, ensemble learning, and so on. The problems can further cause misleading of classifiers, unsatisfactory classification accuracy and stability. This paper proposed a new diagnosis algorithm of PD by combining multi-edit sample selection method and random forest. At the end of it, this paper presents a group of experiments carried out with the newest public datasets. Experimental results showed that this proposed algorithm realized the classification of the samples and the subjects of PD. Furthermore, it achieved average classification accuracy of 100% and obtained improvement of up to 29.44% compared to those provided by the subjects. This paper proposes a new speech diagnosis algorithm for PD based on instance selection; and the method algorithm has a higher and more stable classification accuracy, compared with the other algorithms.
In this paper, the research has been conducted by the Microsoft kinect for windows v2 for obtaining the walking trajectory data from hemiplegic patients, based on which we achieved automatic identification of the hemiplegic gait and sorted the significance of identified features. First of all, the experimental group and two control groups were set up in the study. The three groups of subjects respectively completed the prescribed standard movements according to the requirements. The walking track data of the subjects were obtained straightaway by Kinect, from which the gait identification features were extracted: the moving range of pace, stride and center of mass (up and down/left and right). Then, the bayesian classification algorithm was utilized to classify the sample set of these features so as to automatically recognize the hemiplegia gait. Finally, the random forest algorithm was used to identify the significance of each feature, providing references for the diagnose of disease by ranking the importance of each feature. This thesis states that the accuracy of classification approach based on bayesian algorithm reaches 96%; the sequence of significance based on the random forest algorithm is step speed, stride, left-right moving distance of the center of mass, and up-down moving distance of the center of mass. The combination of step speed and stride, and the combination of step speed and center of mass moving distance are important reference for analyzing and diagnosing of the hemiplegia gait. The results may provide creative mind and new references for the intelligent diagnosis of hemiplegia gait.
With the rapid improvement of the perception and computing capacity of mobile devices such as smart phones, human activity recognition using mobile devices as the carrier has been a new research hot-spot. The inertial information collected by the acceleration sensor in the smart mobile device is used for human activity recognition. Compared with the common computer vision recognition, it has the following advantages: convenience, low cost, and better reflection of the essence of human motion. Based on the WISDM data set collected by smart phones, the inertial navigation information and the deep learning algorithm-convolutional neural network (CNN) were adopted to build a human activity recognition model in this paper. The K nearest neighbor algorithm (KNN) and the random forest algorithm were compared with the CNN network in the recognition accuracy to evaluate the performance of the CNN network. The classification accuracy of CNN model reached 92.73%, which was much higher than KNN and random forest. Experimental results show that the CNN algorithm model can achieve more accurate human activity recognition and has broad application prospects in predicting and promoting human health.
Objective To investigate the key risk factors for low anterior resection syndrome (LARS) within 6 months after rectal cancer surgery and to construct a risk prediction model based on the random forest algorithm, providing a reference for early clinical intervention. Methods A retrospective study was conducted on patients who underwent rectal cancer surgery at the West China Hospital of Sichuan University from January 2020 to August 2021. A prediction model for the occurrence of LARS within 6 months after rectal cancer surgery was constructed using the random forest algorithm. The dataset was divided into a training set and a test set in an 8∶2 ratio. The model performance was evaluated by accuracy, sensitivity, specificity, area under the receiver operating characteristic curve (AUC), Brier score, and decision curve analysis (DCA). Results A total of 394 patients were enrolled. Among the 394 patients, 106 developed LARS within 6 months after surgery, with an incidence rate of 26.9%. According to the importance ranking in the random forest algorithm, the key predictive factors were: distance from the inferior tumor margin to the dentate line, body mass index (BMI), tumor size, time to first postoperative flatus, operation time, age, neoadjuvant therapy, and TNM stage. The prediction model constructed using these key factors achieved the accuracy of 73.4%, sensitivity of 75.0%, specificity of 72.7%, AUC (95% confidence interval) of 0.801 (0.685, 0.916), and the Brier score of 0.198. DCA showed that the model provided favorable clinical benefit when the threshold probability was between 25% and 64%. Conclusions The results of this study suggest that patients with a shorter distance from the tumor to the dentate line, higher BMI, and larger tumor size are at higher risk of developing LARS. The risk prediction model constructed in this study demonstrates a good predictive performance and may provide a useful reference for early identification of high-risk patients after rectal cancer surgery.