Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
126 tokens/sec
GPT-4o
47 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
47 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

ASD-DiagNet: A hybrid learning approach for detection of Autism Spectrum Disorder using fMRI data (1904.07577v1)

Published 16 Apr 2019 in cs.LG, eess.IV, and stat.ML

Abstract: Mental disorders such as Autism Spectrum Disorders (ASD) are heterogeneous disorders that are notoriously difficult to diagnose, especially in children. The current psychiatric diagnostic process is based purely on the behavioural observation of symptomology (DSM-5/ICD-10) and may be prone to over-prescribing of drugs due to misdiagnosis. In order to move the field towards more quantitative fashion, we need advanced and scalable machine learning infrastructure that will allow us to identify reliable biomarkers of mental health disorders. In this paper, we propose a framework called ASD-DiagNet for classifying subjects with ASD from healthy subjects by using only fMRI data. We designed and implemented a joint learning procedure using an autoencoder and a single layer perceptron which results in improved quality of extracted features and optimized parameters for the model. Further, we designed and implemented a data augmentation strategy, based on linear interpolation on available feature vectors, that allows us to produce synthetic datasets needed for training of machine learning models. The proposed approach is evaluated on a public dataset provided by Autism Brain Imaging Data Exchange including 1035 subjects coming from 17 different brain imaging centers. Our machine learning model outperforms other state of the art methods from 13 imaging centers with increase in classification accuracy up to 20% with maximum accuracy of 80%. The machine learning technique presented in this paper, in addition to yielding better quality, gives enormous advantages in terms of execution time (40 minutes vs. 6 hours on other methods). The implemented code is available as GPL license on GitHub portal of our lab (https://github.com/pcdslab/ASD-DiagNet).

Citations (249)

Summary

  • The paper presents a novel hybrid learning model integrating autoencoders and a single-layer perceptron to boost ASD diagnostic accuracy using fMRI data.
  • It employs a linear interpolation-based data augmentation strategy that expands the training dataset and improves model performance.
  • The method outperforms existing approaches by achieving up to 80% accuracy and reducing processing time from 6 hours to 40 minutes.

Evaluation of ASD-DiagNet Framework for Autism Spectrum Disorders Diagnosis Using fMRI Data

The paper "Bare Demo of IEEEtran.cls for IEEE Journals" presents a machine learning framework called ASD-DiagNet, established to assist in the diagnosis of Autism Spectrum Disorders (ASD) using functional magnetic resonance imaging (fMRI) data. Given the complexity and heterogeneity of ASD, particularly in pediatric populations, traditional diagnostic methodologies, which rely on behavioral observations described in DSM-5/ICD-10, pose significant challenges. There exists a substantial risk of misdiagnosis and consequent over-prescription of pharmacological treatments.

Methodology

ASD-DiagNet employs an innovative approach by integrating an autoencoder with a single-layer perceptron to enhance the quality of extracted features from fMRI data. This joint learning process optimizes the parameterization of the model, thus improving diagnostic accuracy. Additionally, the authors introduce a data-augmentation strategy based on the linear interpolation of available feature vectors, facilitating the creation of synthetic datasets. Such augmentation techniques expand the available data pool, which is particularly valuable in enhancing machine learning model training given the typical scarcity of medical imaging datasets.

Evaluation and Results

The ASD-DiagNet framework was evaluated using the publicly available Autism Brain Imaging Data Exchange (ABIDE) dataset, comprising 1035 subjects from 17 different brain imaging centers. The model demonstrated robust performance, surpassing state-of-the-art methods from 13 imaging centers with a classification accuracy increase of up to 20%, achieving a maximum accuracy of 80%. A noteworthy aspect of this framework is its computational efficiency, markedly reducing processing time from 6 hours to 40 minutes compared to existing methods, which has significant implications for scaling diagnostic processes in clinical settings.

Implications and Future Directions

The promising results of ASD-DiagNet underscore its potential to transform the diagnostic landscape of mental health disorders, particularly ASD. By introducing a data-driven, objective methodology, this framework challenges the reliance on purely subjective measures of diagnosis and opens new avenues for integrating neuroimaging biomarkers into clinical practice. Future research could focus on further refining these machine learning models, exploring additional imaging modalities, and expanding the dataset to include a more diverse population sample to generalize findings broadly. Moreover, the development of real-time processing capabilities and seamless integration with existing diagnostic infrastructures would be instrumental in enhancing the clinical utility of this framework.

The paper's contributions extend both practical and theoretical aspects of machine learning applications in mental health diagnostics. This work demonstrates the importance of interdisciplinary approaches that combine advanced computational techniques with clinical insights, setting a precedent for future artificial intelligence tools in neuropsychiatry. The openly available code on the GitHub repository offers a resource for researchers seeking to replicate or extend the described methodologies, fostering a collaborative push towards improved diagnostic accuracy and operational efficiency in ASD interventions.

Github Logo Streamline Icon: https://streamlinehq.com

GitHub