Use of machine learning methods in prediction of short-term outcome in autism spectrum disorders

Abstract

OBJECTIVE: Studies show partial improvements in some core symptoms of Autism Spectrum Disorders (ASD) in time. However, the predictive factors (e.g. pretreatment IQ, comorbid psychiatric disorders, adaptive, and language skills, etc.) for a better the outcome was not studied with machine learning methods. We aimed to examine the predictors of outcome with machine learning methods, which are novel computational methods including statistical estimation, information theories and mathematical learning automatically discovering useful patterns in large amounts of data. METHOD: The study the group comprised 433 children (mean age: 72.3 ± 45.9 months) with ASD diagnosis. The ASD symptoms were assessed by the Autism Behavior Checklist, Aberrant Behavior Checklist, Clinical Global Impression scales at baseline (T0) and 12th (T1), 24th (T2), and 36th (T3) months. We tested the performance of for machine learning algorithms (Naive Bayes, Generalized Linear Model, Logistic Regression, Decision Tree) on our data, including the 254 items in the baseline forms. Patients with ≤2 CGI points in ASD symptoms at in 36 months were accepted as the group who has "better outcome" as the prediction class. RESULTS: The significant proportion of the cases showed significant improvement in ASD symptoms (39.7% in T1, 60.7% in T2; 77.8% in T3). Our machine learning model in T3 showed that diagnosis group affected the prognosis. In the autism group, older father and mother age; in PDD-NOS group, MR comorbidity, less birth weight and older age at diagnosis have a worse outcome. In Asperger's Disorder age at diagnosis, age at first evaluation and developmental cornerstones has affected prognosis. CONCLUSION: In accordance with other studies we found early age diagnosis, early start rehabilitation, the severity of ASD symptoms at baseline assessment predicted outcome. Also, we found comorbid psychiatric diagnoses are affecting the outcome of ASD symptoms in clinical observation. The machine learning models reveal several others are more significant (e.g. parental age, birth weight, sociodemographic variables, etc.) in terms of prognostic information and also planning treatment of children with ASD.

A Systematic Literature Review on the Application of Machine-Learning Models in Behavioral Assessment of Autism Spectrum Disorder

“…The trial-error approach involves repetitive evaluation of the ML models using a varying combination of the features; the most influential combination achieves superior results with fewer input parameters. Specifically, the studies utilized various feature selection techniques, including trial-error [ 13 , 34 , 35 , 39 , 45 ], Variable Analysis (Va) [ 46 , 47 ], information gain (IG) and chi-square testing (CHI) [ 48 ], sequential feature selection (SFS) [ 49 ], correlation-based feature selection (CFS) and minimum redundancy maximum relevance (mRMR) [ 12 ]. Additionally, ML-based feature selection techniques employed include recursive feature selection [ 40 ], sparsity/parsimony enforcing regularization techniques [ 50 ], stepwise backward feature selection [ 37 ], and forward feature selection [ 36 ].…”

“…As shown in Table 2 , the commonly implemented ML algorithms are Random Forest (RF) [ 12 , 43 , 47 , 51 ], Support Vector Machines (SVM) [ 37 , 38 , 40 , 49 , 50 ], Alternative Decision Tree (ADTree) [ 34 , 35 , 39 , 45 ], and Logistic Regression (LR) [ 13 , 37 , 48 ]. To achieve comparative results, most of the studies employed several algorithms, such as Adaboost, Artificial Neural Network (ANN), Linear Discriminant Analysis (LDA), Naïve Bayes, and K-Nearest Neighbor (KNN).…”

“…The most utilized data collection instruments are AQ-10 [ 11 , 13 , 43 , 44 , 46 , 47 , 48 , 49 , 51 , 52 ], Q-CHAT-10 [ 11 , 44 , 46 , 52 ], ADOS [ 34 , 37 , 39 , 40 , 42 , 50 ], ADI-R [ 35 , 38 , 42 ], and Social Responsiveness Scale (SRS) [ 36 , 38 , 53 ]. Others include Autism Behavior Checklist, Aberrant Behavior Checklist, Clinical Global Impression [ 45 ], and MCHAT-based Pictorial Autism Assessment Schedule (PASS) [ 12 ]. Thus, the need for improving the reliability of these assessment instruments and ascertaining their relevance in ML modelling remains.…”

“…The most prominent sources of data utilized in the studies include Boston Autism Consortium (AC), Autism Genetic Resource Exchange (AGRE), Simons Simplex Collection (SSC) [ 34 , 35 , 36 , 37 , 39 , 50 , 53 ], National Database for Autism Research (NDAR) [ 37 , 39 ], and Simons Variation In Individuals Project (SVIP) [ 37 , 39 , 50 ]. Other studies utilized data sets from ASDTest: Kaggle and UCI ML repository [ 11 , 13 , 43 , 44 , 46 , 47 , 48 , 49 , 51 , 52 ], Association of Parents and Friends for the Support and Defense of the rights of people with Autism (APADA) [ 42 ], PASS app [ 12 ], Ondokuz Mayis University Samsun [ 45 ], and ASD outpatient clinics in Germany [ 40 ]. To achieve standardized comparative results, there is a need for standardized ASD data repositories for machine learning studies [ 25 ].…”

Use of Empirical Mode Decomposition in ERP Analysis to Classify Familial Risk and Diagnostic Outcomes for Autism Spectrum Disorder

“…These approaches can aid the interpretation of physiological data and be used by clinicians to consider disease diagnosis based on multiple sources of information (e.g., EEG, electrocardiography, electronic medical records, and magnetic resonance imaging). There has been growing interest in investigating ASD diagnostic utility through neuroimaging data using machine learning [ 12 , 13 , 14 , 15 , 16 ].…”