Use of machine learning to identify autism with natural social gaze behavior (Preprint)

BACKGROUND Abnormal gaze behavior is a prominent feature of the autism spectrum disorder (ASD). Previous eye tracking studies had participants watch images (i.e., picture, video and webpage), and the application of machine learning (ML) on these data showed promising results in identify ASD individuals. Given the fact that gaze behavior differs in face-to-face interaction from image viewing tasks, no study has investigated whether natural social gaze behavior could accurately identify ASD. OBJECTIVE The objective of this study was to examine whether and what area of interest (AOI)-based features extracted from the natural social gaze behavior could identify ASD. METHODS Both children with ASD and typical development (TD) were eye-tracked when they were engaged in a face-to-face conversation with an interviewer. Four ML classifiers (support vector machine, SVM; linear discriminant analysis, LDA; decision tree, DT; and random forest, RF) were used to determine the maximum classification accuracy and the corresponding features. RESULTS A maximum classification accuracy of 84.62% were achieved with three classifiers (LDA, DT and RF). Results showed that the mouth, but not the eyes AOI, was a powerful feature in detecting ASD. CONCLUSIONS Natural gaze behavior could be leveraged to identify ASD, suggesting that ASD might be objectively screened with eye tracking technology in everyday social interaction. In addition, the comparison between our and previous findings suggests that eye tracking features that could identify ASD might be culture dependent and context sensitive.

Download Full-text

Evaluation of an eye tracking setup for studying visual attention in face-to-face conversations

Scientific Reports ◽

10.1038/s41598-021-81987-x ◽

2021 ◽

Vol 11 (1) ◽

Author(s):

Antonia Vehlen ◽

Ines Spenthof ◽

Daniel Tönsing ◽

Markus Heinrichs ◽

Gregor Domes

Keyword(s):

Eye Tracking ◽

Data Quality ◽

Autism Spectrum ◽

Independent Study ◽

Facial Features ◽

Gaze Behavior ◽

Tracking Accuracy ◽

Attentional Processing ◽

Face To Face ◽

Gaze Patterns

AbstractMany eye tracking studies use facial stimuli presented on a display to investigate attentional processing of social stimuli. To introduce a more realistic approach that allows interaction between two real people, we evaluated a new eye tracking setup in three independent studies in terms of data quality, short-term reliability and feasibility. Study 1 measured the robustness, precision and accuracy for calibration stimuli compared to a classical display-based setup. Study 2 used the identical measures with an independent study sample to compare the data quality for a photograph of a face (2D) and the face of the real person (3D). Study 3 evaluated data quality over the course of a real face-to-face conversation and examined the gaze behavior on the facial features of the conversation partner. Study 1 provides evidence that quality indices for the scene-based setup were comparable to those of a classical display-based setup. Average accuracy was better than 0.4° visual angle. Study 2 demonstrates that eye tracking quality is sufficient for 3D stimuli and robust against short interruptions without re-calibration. Study 3 confirms the long-term stability of tracking accuracy during a face-to-face interaction and demonstrates typical gaze patterns for facial features. Thus, the eye tracking setup presented here seems feasible for studying gaze behavior in dyadic face-to-face interactions. Eye tracking data obtained with this setup achieves an accuracy that is sufficient for investigating behavior such as eye contact in social interactions in a range of populations including clinical conditions, such as autism spectrum and social phobia.

Download Full-text

Machine Learning-Based Electroencephalographic Phenotypes of Schizophrenia and Major Depressive Disorder

Frontiers in Psychiatry ◽

10.3389/fpsyt.2021.745458 ◽

2021 ◽

Vol 12 ◽

Author(s):

Kuk-In Jang ◽

Sungkean Kim ◽

Soo Young Kim ◽

Chany Lee ◽

Jeong-Ho Chae

Keyword(s):

Machine Learning ◽

Major Depressive Disorder ◽

Depressive Disorder ◽

Classification Accuracy ◽

Negative Symptoms ◽

Support Vector ◽

Svm Classifier ◽

Major Depressive ◽

Linear Discriminant ◽

Cortical Sources

Background: Psychiatric diagnosis is formulated by symptomatic classification; disease-specific neurophysiological phenotyping could help with its fundamental treatment. Here, we investigated brain phenotyping in patients with schizophrenia (SZ) and major depressive disorder (MDD) by using electroencephalography (EEG) and conducted machine-learning-based classification of the two diseases by using EEG components.Materials and Methods: We enrolled healthy controls (HCs) (n = 30) and patients with SZ (n = 34) and MDD (n = 33). An auditory P300 (AP300) task was performed, and the N1 and P3 components were extracted. Two-group classification was conducted using linear discriminant analysis (LDA) and support vector machine (SVM) classifiers. Positive and negative symptoms and depression and/or anxiety symptoms were evaluated.Results: Considering both the results of statistical comparisons and machine learning-based classifications, patients and HCs showed significant differences in AP300, with SZ and MDD showing lower N1 and P3 than HCs. In the sum of amplitudes and cortical sources, the findings for LDA with classification accuracy (SZ vs. HCs: 71.31%, MDD vs. HCs: 74.55%), sensitivity (SZ vs. HCs: 77.67%, MDD vs. HCs: 79.00%), and specificity (SZ vs. HCs: 64.00%, MDD vs. HCs: 69.67%) supported these results. The SVM classifier showed reasonable scores between SZ and HCs and/or MDD and HCs. The comparison between SZ and MDD showed low classification accuracy (59.71%), sensitivity (65.08%), and specificity (54.83%).Conclusions: Patients with SZ and MDD showed deficiencies in N1 and P3 components in the sum of amplitudes and cortical sources, indicating attentional dysfunction in both early and late sensory/cognitive gating input. The LDA and SVM classifiers in the AP300 are useful to distinguish patients with SZ and HCs and/or MDD and HCs.

Download Full-text

Classification of Black Plastics Waste Using Fluorescence Imaging and Machine Learning

Recycling ◽

10.3390/recycling4040040 ◽

2019 ◽

Vol 4 (4) ◽

pp. 40 ◽

Cited By ~ 1

Author(s):

Florian Gruber ◽

Wulf Grählert ◽

Philipp Wollmann ◽

Stefan Kaskel

Keyword(s):

Machine Learning ◽

Classification Accuracy ◽

Near Infrared ◽

Search Algorithm ◽

Random Search ◽

Support Vector ◽

Classification Algorithms ◽

Linear Discriminant ◽

Black Plastic

This work contributes to the recycling of technical black plastic particles, for example from the automotive or electronics industries. These plastics cannot yet be sorted with sufficient purity (up to 99.9%), which often makes economical recycling impossible. As a solution to this problem, imaging fluorescence spectroscopy with additional illumination in the near infrared spectral range in combination with classification by machine learning or deep learning classification algorithms is here investigated. The algorithms used are linear discriminant analysis (LDA), k-nearest neighbour classification (kNN), support vector machines (SVM), ensemble models with decision trees (ENSEMBLE), and convolutional neural networks (CNNs). The CNNs in particular attempt to increase overall classification accuracy by taking into account the shape of the plastic particles. In addition, the automatic optimization of the hyperparameters of the classification algorithms by the random search algorithm was investigated. The aim was to increase the accuracy of the classification models. About 400 particles each of 14 plastics from 12 plastic classes were examined. An attempt was made to train an overall model for the classification of all 12 plastics. The CNNs achieved the highest overall classification accuracy with 93.5%. Another attempt was made to classify 41 mixtures of industrially relevant plastics with a maximum of three plastic classes per mixture. The same average classification accuracy of 99.0% was achieved for the ENSEMBLE, SVM, and CNN algorithms. The target overall classification accuracy of 99.9% was achieved for 18 of the 41 compounds. The results show that the method presented is a promising approach for sorting black technical plastic waste.

Download Full-text

Improving Motor Imagery EEG Signals Classification Accuracy with CSP by Available Machine Learning Approach

Journal of Engineering Science ◽

10.3329/jes.v12i2.54632 ◽

2021 ◽

Vol 12 (2) ◽

pp. 67-77

Author(s):

Umme Farhana ◽

Mst Jannatul Ferdous

Keyword(s):

Machine Learning ◽

Feature Extraction ◽

Motor Imagery ◽

Classification Accuracy ◽

Support Vector ◽

Band Pass Filter ◽

K Nearest Neighbor ◽

Eeg Signals ◽

Pass Filter ◽

Linear Discriminant

In brain computer interface (BCI) systems, the electroencephalography (EEG) signals give a pathway to a motor disabled person to communicate outside using the brain signal and a computer. EEG signals of different motor imagery (MI) movements can be differentiated using an effective classification technique to aid a motor disabled patient. The purpose of this paper is to classify two different types of MI movement tasks, movement of the left hand and movement of the right foot EEG signals accurately. For this purpose we have used a publicly available dataset. Since the feature extraction for classification is an important task, so we have used popular common spatial pattern (CSP) method for spatial feature extraction. Two different machine learning classifiers named support vector machine (SVM) and K-nearest neighbor (KNN) have been used to verify the proposed method. We got the highest average results 95.55%, 98.73% and 92.38% in case of SVM and 93.5%, 98.73% and 90.15% in case of KNN for classification accuracy, sensitivity, and specificity, respectively when a Butterworth band-pass filter passed through [10–30] Hz. On the other hand accuracy came to 89.4% in [10-30] Hz when applying CSP for feature extraction and fisher linear discriminant analysis (FLDA) for classification on this dataset earlier. Journal of Engineering Science 12(2), 2021, 67-77

Download Full-text

Application of Machine Learning Approaches for the Design and Study of Anticancer Drugs

Current Drug Targets ◽

10.2174/1389450119666180809122244 ◽

2019 ◽

Vol 20 (5) ◽

pp. 488-500 ◽

Cited By ~ 6

Author(s):

Yan Hu ◽

Yi Lu ◽

Shuo Wang ◽

Mengying Zhang ◽

Xiaosheng Qu ◽

...

Keyword(s):

Machine Learning ◽

Drug Design ◽

Anticancer Drugs ◽

Nearest Neighbor ◽

Cost Effective ◽

Support Vector ◽

Learning Approaches ◽

K Nearest Neighbor ◽

Activity Prediction ◽

Linear Discriminant

Background: Globally the number of cancer patients and deaths are continuing to increase yearly, and cancer has, therefore, become one of the world's highest causes of morbidity and mortality. In recent years, the study of anticancer drugs has become one of the most popular medical topics. Objective: In this review, in order to study the application of machine learning in predicting anticancer drugs activity, some machine learning approaches such as Linear Discriminant Analysis (LDA), Principal components analysis (PCA), Support Vector Machine (SVM), Random forest (RF), k-Nearest Neighbor (kNN), and Naïve Bayes (NB) were selected, and the examples of their applications in anticancer drugs design are listed. Results: Machine learning contributes a lot to anticancer drugs design and helps researchers by saving time and is cost effective. However, it can only be an assisting tool for drug design. Conclusion: This paper introduces the application of machine learning approaches in anticancer drug design. Many examples of success in identification and prediction in the area of anticancer drugs activity prediction are discussed, and the anticancer drugs research is still in active progress. Moreover, the merits of some web servers related to anticancer drugs are mentioned.

Download Full-text

Internet of Things-Based Intelligent Smart Home Control System

Security and Communication Networks ◽

10.1155/2021/9928254 ◽

2021 ◽

Vol 2021 ◽

pp. 1-17

Author(s):

Olutosin Taiwo ◽

Absalom E. Ezugwu

Keyword(s):

Machine Learning ◽

Mobile Application ◽

Smart Home ◽

Human Life ◽

Machine Learning Algorithms ◽

Support Vector ◽

Automation System ◽

Home Automation ◽

Area Of Interest ◽

Home Automation System

The smart home is now an established area of interest and research that contributes to comfort in modern homes. With the Internet being an essential part of broad communication in modern life, IoT has allowed homes to go beyond building to interactive abodes. In many spheres of human life, the IoT has grown exponentially, including monitoring ecological factors, controlling the home and its appliances, and storing data generated by devices in the house in the cloud. Smart home includes multiple components, technologies, and devices that generate valuable data for predicting home and environment activities. This work presents the design and development of a ubiquitous, cloud-based intelligent home automation system. The system controls, monitors, and oversees the security of a home and its environment via an Android mobile application. One module controls and monitors electrical appliances and environmental factors, while another module oversees the home’s security by detecting motion and capturing images. Our work uses a camera to capture images of objects triggered by their motion being detected. To avoid false alarms, we used the concept of machine learning to differentiate between images of regular home occupants and those of an intruder. The support vector machine algorithm is proposed in this study to classify the features of the image captured and determine if it is that of a regular home occupant or an intruder before sending an alarm to the user. The design of the mobile application allows a graphical display of the activities in the house. Our work proves that machine learning algorithms can improve home automation system functionality and enhance home security. The work’s prototype was implemented using an ESP8266 board, an ESP32-CAM board, a 5 V four-channel relay module, and sensors.

Download Full-text

ADOS-Eye-Tracking: The Archimedean Point of View and Its Absence in Autism Spectrum Conditions

Frontiers in Psychology ◽

10.3389/fpsyg.2021.584537 ◽

2021 ◽

Vol 12 ◽

Author(s):

Ulrich Max Schaller ◽

Monica Biscaldi ◽

Anna Burkhardt ◽

Christian Fleischhaker ◽

Michael Herbert ◽

...

Keyword(s):

Social Interaction ◽

Eye Tracking ◽

Children And Adolescents ◽

Face Perception ◽

Autism Spectrum ◽

Autism Spectrum Conditions ◽

Gaze Behavior ◽

Face Region ◽

Mobile Eye Tracking ◽

Standardized Diagnostic

Face perception and emotion categorization are widely investigated under laboratory conditions that are devoid of real social interaction. Using mobile eye-tracking glasses in a standardized diagnostic setting while applying the Autism Diagnostic Observation Schedule (ADOS-2), we had the opportunity to record gaze behavior of children and adolescents with and without Autism Spectrum Conditions (ASCs) during social interaction. The objective was to investigate differences in eye-gaze behavior between three groups of children and adolescents either (1) with ASC or (2) with unconfirmed diagnosis of ASC or (3) with neurotypical development (NTD) during social interaction with an adult interviewer in a diagnostic standard situation using the ADOS-2. In a case control study, we used mobile eye-tracking glasses in an ecologically valid and highly standardized diagnostic interview to investigate suspected cases of ASC. After completion of the ASC diagnostic gold standard including the ADOS-2, the participants were assigned to two groups based on their diagnosis (ASC vs. non-ASC) and compared with a matched group of neurotypically developed controls. The primary outcome measure is the percentage of total dwell times assessed for different areas of interest (AOI) with regard to the face and body of a diagnostic interviewer and the surrounding space. Overall, 65 children and adolescents within an age range of 8.3–17.9 years were included in the study. The data revealed significant group differences, especially in the central-face area. Previous investigations under laboratory conditions gave preferential attention to the eye region during face perception to describe differences between ASC and NTD. In this study – using an ecologically valid setting within a standard diagnostic procedure – the results indicate that neurotypically developed controls seem to process faces and facial expressions in a holistic manner originating from the central-face region. Conversely, participants on the Autism Spectrum (tAS) seem to avoid the central-face region and show unsystematic gaze behavior, not using the preferred landing position in the central-face region as the Archimedean point of face perception. This study uses a new approach, and it will be important to replicate these preliminary findings in future research.

Download Full-text

Traditional versus Neural Network Classification Methods for Facial Emotion Recognition

Current Directions in Biomedical Engineering ◽

10.1515/cdbme-2021-2052 ◽

2021 ◽

Vol 7 (2) ◽

pp. 203-206

Author(s):

Herag Arabian ◽

Verena Wagner-Hartl ◽

Knut Moeller

Keyword(s):

Neural Network ◽

Machine Learning ◽

Emotion Recognition ◽

Network Architecture ◽

Autism Spectrum ◽

Facial Emotion Recognition ◽

Facial Emotion ◽

Machine Learning Techniques ◽

Support Vector ◽

Machine Model

Abstract Facial emotion recognition (FER) is a topic that has gained interest over the years for its role in bridging the gap between Human and Machine interactions. This study explores the potential of real time FER modelling, to be integrated in a closed loop system, to help in treatment of children suffering from Autism Spectrum Disorder (ASD). The aim of this study is to show the differences between implementing Traditional machine learning and Deep learning approaches for FER modelling. Two classification approaches were taken, the first approach was based on classic machine learning techniques using Histogram of Oriented Gradients (HOG) for feature extraction, with a k-Nearest Neighbor and a Support Vector Machine model as classifiers. The second approach uses Transfer Learning based on the popular “Alex Net” Neural Network architecture. The performance of the approaches was based on the accuracy of randomly selected validation sets after training on random training sets of the Oulu-CASIA database. The data analyzed shows that traditional machine learning methods are as effective as deep neural net models and are a good compromise between accuracy, extracted features, computational speed and costs.

Download Full-text

Cardiovascular Disease Prediction from Electrocardiogram by using Machine Learning Method: A Snapshot from the Subjects of the Malaysian Cohort

10.21203/rs.2.22561/v1 ◽

2020 ◽

Author(s):

Nazrul Anuar Nayan ◽

Hafifah Ab Hamid ◽

Mohd Zubir Suboh ◽

Noraidatulakma Abdullah ◽

Rosmina Jaafar ◽

...

Keyword(s):

Machine Learning ◽

Cardiovascular Disease ◽

Nearest Neighbor ◽

Total Cholesterol Level ◽

Population Based ◽

Support Vector ◽

K Nearest Neighbor ◽

Cvd Risk ◽

Linear Discriminant ◽

Artificial Neural Network Ann

Abstract Background: Cardiovascular disease (CVD) is the leading cause of deaths worldwide. In 2017, CVD contributed to 13,503 deaths in Malaysia. The current approaches for CVD prediction are usually invasive and costly. Machine learning (ML) techniques allow an accurate prediction by utilizing the complex interactions among relevant risk factors. Results: This study presents a case–control study involving 60 participants from The Malaysian Cohort, which is a prospective population-based project. Five parameters, namely, the R–R interval and root mean square of successive differences extracted from electrocardiogram (ECG), systolic and diastolic blood pressures, and total cholesterol level, were statistically significant in predicting CVD. Six ML algorithms, namely, linear discriminant analysis, linear and quadratic support vector machines, decision tree, k-nearest neighbor, and artificial neural network (ANN), were evaluated to determine the most accurate classifier in predicting CVD risk. ANN, which achieved 90% specificity, 90% sensitivity, and 90% accuracy, demonstrated the highest prediction performance among the six algorithms. Conclusions: In summary, by utilizing ML techniques, ECG data can serve as a good parameter for CVD prediction among the Malaysian multiethnic population.

Download Full-text

Introduction and Implementation of Machine Learning Algorithms in R

Advances in Business Information Systems and Analytics - Sentiment Analysis and Knowledge Discovery in Contemporary Business ◽

10.4018/978-1-5225-4999-4.ch008 ◽

2019 ◽

pp. 126-147

Author(s):

S. R. Mani Sekhar ◽

G. M. Siddesh

Keyword(s):

Machine Learning ◽

Neural Networks ◽

Support Vector Machine ◽

Discriminant Analysis ◽

Computer Science ◽

Learning Algorithms ◽

Machine Learning Algorithms ◽

Support Vector ◽

Linear Discriminant ◽

The Given

Machine learning is one of the important areas in the field of computer science. It helps to provide an optimized solution for the real-world problems by using past knowledge or previous experience data. There are different types of machine learning algorithms present in computer science. This chapter provides the overview of some selected machine learning algorithms such as linear regression, linear discriminant analysis, support vector machine, naive Bayes classifier, neural networks, and decision trees. Each of these methods is illustrated in detail with an example and R code, which in turn assists the reader to generate their own solutions for the given problems.

Download Full-text