EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Hybrid detection model for autism spectrum disorder using convolutional neural network and autoencoder

Muhammed Aminu, Mukhtar Danlami Danlami

Aminu Kano Academic Scholars Association Multidisciplinary Journal · 29.04.2026 · 4 (1) · 89-110

DOI: 10.64726/npc4m236 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Aminu Kano Academic Scholars Association Multidisciplinary Journal, online erschienen am 29.04.2026. Verfasst von Muhammed Aminu und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Autism Spectrum Disorder (ASD) is a developmental condition that influences how individuals communicate, interact socially, and behave. Detecting ASD early and accurately is very important, as it allows for timely support and intervention. However, conventional diagnostic methods are often slow, subjective, and heavily dependent on specialist assessments, which creates a need for faster and more reliable automated solutions. In this study, a hybrid deep learning model is developed by combining a Convolutional Neural Network (CNN) with an Autoencoder (AE) to detect ASD from facial images. The CNN is used to capture important spatial features from the images, while the autoencoder learns compact and meaningful representations of the data. These features are then merged and passed through fully connected layers to perform classification. To improve the model’s learning process, a hybrid loss function that integrates cross-entropy loss with reconstruction loss is applied. The model’s performance is evaluated using key metrics such as accuracy, precision, recall, F1-score, confusion matrix, and the Receiver Operating Characteristic (ROC) curve. In addition, k-fold cross-validation is used to ensure the model is robust and performs consistently across different data splits. The experimental results show that the proposed CNN–AE hybrid model achieves an impressive accuracy of 98.4%, with strong consistency and very low misclassification rates. These findings suggest that combining supervised and unsupervised learning techniques enhances both feature extraction and classification performance. Overall, the model offers a reliable and efficient approach for ASD detection and shows strong potential to assist clinicians in making faster and more accurate diagnostic decisions in real world healthcare settings.

Zitieren

APA

Aminu, M., & Danlami, M. D. (2026). Hybrid detection model for autism spectrum disorder using convolutional neural network and autoencoder. Aminu Kano Academic Scholars Association Multidisciplinary Journal, 4(1), 89-110. https://doi.org/10.64726/npc4m236

Vancouver

Aminu M, Danlami MD. Hybrid detection model for autism spectrum disorder using convolutional neural network and autoencoder. Aminu Kano Academic Scholars Association Multidisciplinary Journal. 2026;4(1):89-110. doi:10.64726/npc4m236

BibTeX
@article{Aminu2026,
  author = {Aminu, Muhammed and Danlami, Mukhtar Danlami},
  title = {Hybrid detection model for autism spectrum disorder using convolutional neural network and autoencoder},
  journal = {Aminu Kano Academic Scholars Association Multidisciplinary Journal},
  year = {2026},
  volume = {4},
  number = {1},
  pages = {89--110},
  doi = {10.64726/npc4m236}
}
Kostenlos starten →