Recovery Gestures Classification Using KNN and LDA Models.

Hand and joint mobility recovery involve performing a set of exercises. Gestures are often used in the hand mobility recovery process. This paper discusses the selection and the use of 2 models of neural networks for the classification of data that describe Leap Motion gestures. The gestures are: the hand opening and closing gesture and the palm rotation gesture. The purpose is the optimal selection of the neural network model to be used in the classification of the data describing the recovery gestures. The models chosen for the classification of the two gestures were: Linear Discriminant Analysis (LDA) and K-neighbors Classifier (KNN). The accuracies achieved in the classification of the gestures for each model are: 0.91 - LDA and 0.98 - KNN.

Application for Screening to Monitor and Assess Pregnancy Potential Risks.

Research in the field of maternal-fetal medicine brings a new approach, by involving several fields: genetics, informatics, teratology, imaging, obstetric diagnosis, maternal-fetal physiology, endocrinology, and aims to determine the relationships that appear between the maternal medical pathology and the fetal one. In this article, we present an application for monitoring and calculating risk in Trisomy-21 for pregnant women. To calculate the risk, we used 2 methods, one mathematical and one using neural networks to investigate which one offers higher precision. Following the experimental results, due to the use of several variables that increase the risk for Trisomy-21, the conclusion is that the method using neural networks is better, having an accuracy of 95%.

Comparison of Data Classification Results for Leap Motion Recovery Gestures.

Static and dynamic gestures are frequently used in activities supporting learning, recovery healthcare, engineering, and 3D games to increase the interactivity between man and machine. The gestures are detected via hardware devices and data is processed using different software methods. This paper presents the manner of detection and interpretation of two gestures, a hand rotation gesture and a palm closing and opening gesture, using the Leap Motion device. These two dynamic gestures are very often used in hand recovery exercises. For comparing the two gestures we use data classification methods, Support Vector Machine (SVM) and Multilayer Perceptron (MLP). The data for the gesture classification were 80% training data and 20% testing data. The metrics for comparison are precision, recall, F1-score, and the total number of testing cases (support). The SVM classifier gives an accuracy of 99.4% and the MLP classifier a 96.2%. We built two confusion matrices for better visualizing the results.

Gesture Classification for a Hand Controller Device Using Neural Networks.

This paper presents a convolutional neural network-based classification of the hand flexion and extension gestures used in wrist recovery after injury. The hand gesture recognition device used in our study is the Leap Motion controller. The Leap Motion device's inability to accurately differentiate the left hand from the right hand when performing hand rotation gestures was eliminated by introducing hand and thumb direction vectors into the database used to train the neural network. A 3D environment was created for the introduction of the data describing the gesture into the database. A classification accuracy of 95% was achieved for the hand flexion and extension gesture divided into three levels for each hand. The populated database may also be used to classify other gestures involving hand rotation.

Improving the Prescription Process Information Support with Structured Medical Prospectuses Using Neural Networks.

To provide the best treatment, a physician needs information about both the patient and the medicines matching the patient status and improving it. In this article, we present three methods for structuring the sections of medical prospectuses using neural networks. To structure the information from a medical prospectus we use 3 web sources with structured data from sections (with names sections from prospectuses and with uniformized names of sections) to train as input for neural networks. The tests were conducted on Romanian prospectuses. After running the three algorithms, the prospectuses were compared in terms of accuracy and execution time for each source. It was concluded that the accuracy is higher in convolutional networks and in the case of uniform name sections. The output data is used in applications with decision support for the treatment, matching best treatment with the patient's status.

Named Entity Recognition and Classification for Medical Prospectuses.

Structuring and processing natural language is a growing challenge in the medical field. Researchers are looking for new ways to extract knowledge to create databases and applications to help doctors treat patients and minimize medical errors. A very important part in treating a patient is to provide a fair and effective treatment for diseases. In this article we present a method of extracting important information from medical prospectuses, such as a drug-treated condition, a medicine name, a drug type, etc. To extract these entities, we use Stanford NER Tagger trained for prospectuses in Romanian language. The model was trained and tested with 3 types of medication. For each test, the accuracy of the extracted data was calculated. The extracted medical information is used to create databases with structured information that are useful for decision-support applications to check for or find suggestions for the best treatments.

Enhancing Precision in Gesture Detection for Hand Recovery After Injury Using Leap Motion and Machine Learning.

This paper presents an improved solution for detecting gestures with a better precision using the Leap Motion sensor and Machine Learning support. A neural network is trained to recognize a hand rotation gesture expressing the grade of recovery, with a supination and pronation exercise. The supination-pronation movement is divided into 4 levels because the users are not usually able to perform a complete rotation gesture in hand recovery after injury. The neural network is trained with data representing the hand rotation angle measurements on the x, y and z axes. The Neural Network training is based on the Tensorflow library. 3 tests were carried out to test the network and eventually a 96% gesture-detection accuracy was achieved.