Autism Spectrum Disorder (ASD) is a neuro-developmental disorder that affects social and cognitive skills, causing repetitive behaviors, restricted interests, communication problems, and difficulty in social interaction. Early diagnosis of ASD can prevent its severity and prolonged effects.

Machine learning (ML) can be used to screen and diagnose ASD more quickly and accurately. ML techniques can train ASD models using clinical approaches, which are time-consuming diagnostic procedures that are not frequently carried out unless the predictive risk of ASD development is high. ML models can help streamline the diagnostic process, assisting families in getting to critical therapies more quickly.

Various ML classification models can be used for early prediction of autism to prevent its prolonged effects in adults and children. For instance, a Convolutional Neural Network (CNN) based prediction model has been found to work better for predicting ASD in adults, children, and adolescents, with higher accuracy rates of 99.53%, 98.30%, and 96.88%, respectively.

Federated Learning (FL) is an advanced approach of ML that ensures data security by keeping data secure with the owner organization. FL trains a small-sized local ML-based classifier onsite without moving data over the network. FL is beneficial in resolving data privacy, data protection, and data security issues, as well as network latency, communication delay, and data theft issues.

FL has been applied for the detection of multiple neurological disorders, including ASD. For instance, two different ML models, including SVM and LR, have been trained locally using four different ASD datasets, demonstrating the potential of FL in detecting ASD in both children and adults.

In summary, ML and FL can help screen and diagnose ASD more quickly and accurately, ensuring data security and privacy while streamlining the diagnostic process.

Source code