ATHENA: Hypergraph-Based Learning for Toddler Autism Classification from Eye-Tracking
Authors: Quang-Anh N. D., Duc Pham Minh, Thao Pham Phuong, Minh-Anh Nguyen, Thuy Linh Nguyen Thi, Hung Ha Manh
Organizations: Agora Intelligence Lab, Hanoi, 100000, Vietnam · School of Information and Communications Technology, Hanoi University of Science and Technology, Hanoi, 100000, Vietnam · International School, Vietnam National University, Hanoi, 100000, Vietnam
In this work we proposed a architecture designed for the classification of ASD in toddlers using eye-tracking data, addressing the critical need for early and precise detection. The proposed model incorporates FMMix, a feature-level augmentation technique that bolsters the model's learning capacity, and in this work we proposed the ATHENA, which captures intricate feature relationships via an adaptive hypergraph framework. Evaluated on two datasets dataset of gaze coordinates-the model achieves state-of-the-art performance, with AUC scores of 99.80% and 99.86%, respectively. It also exhibits high accuracy of 98.45% and 98.56%, sensitivity 98.01% and 98.10%, and specificity 98.74% and 99.02%, respectively. Demonstrating its superior ability to identify subtle gaze patterns associated with ASD, these results significantly surpass existing methods, highlighting the model's potential for clinical ASD screening. Future efforts will focus on validation across broader populations and integration with multi-modal data to further refine diagnostic accuracy.
Figures & tables
Figure 1: Proposal DNN Architecture with yellow denoted for the Conv2D layers, Green for Relu Activate Function, Pink represents for MaxPool2D layers, Blue is for Dropout layer, Purple for Flatten, and Orange is for the Linear layer.
Figure 2: Samples data from Carette, R et al. dataset with the red case on right representing the ASD class and black case at left showing the Non-ASD class.
Figure 3: Oculomotor trajectory mapping creation from the ASD-Eyemov dataset with black case on left is for ASD and the red case of right is for the Non-ASD class
Category
Details
Hardware
CPU
Intel Core i5-13400F
RAM
32GB
GPU
NVIDIA GeForce RTX 4070Ti SUPER
Software
OS
Windows 11 Pro
Table 1: Experimental Setup.
Model
AUC Score
Accuracy
PPV
NPV
Sensitivity
Specificity
Mahmoud E.’s Dataset
99.80%
98.45%
98.14%
96.86%
98.01%
98.74%
ASD-EyeMov Dataset
99.86%
98.56%
99.04%
98.07%
98.10%
99.02%
Table 2: Evaluation results from the proposal DNN tested on two datasets.
Model
Best AUC
Accuracy
PPV
NPV
Sensitivity
Specificity
Proposal model
99.80%
98.45%
98.14
96.86%
98.01%
98.74%
Mumenin, N. et al. [ 24 ]
-
98.12%
97.89%
-
99.03%
-
Akter, T. et al. et al. [ 30 ]
71.5%
74.2%
-
-
74.2%
68.8%
Kanhirakadavath, M. et al. [ 31 ]
97.0%
-
94.46%
90.06%
93.28%
91.38%
Ahmed, Z. et al. [ 32 ]
98.0%
98.33%
98.11%
97.25%
97.25%
98.94%
Zhao, Z. et al. [ 33 ]
86.0%
84.62%
-
-
89.47%
80.0%
Table 3: Comparison among different works using Dataset 1.
Model
AUC Score
Accuracy
PPV
NPV
Sensitivity
Specificity
Proposed model
99.86%
98.56%
99.04%
98.07%
98.10%
99.02%
Mumenin, N. et al. [ 24 ]
-
96.83%
95.43%
96.13%
99.87%
99.84%
Colonnesem F. et al. [ 25 ]
-
94.35%
93.2%
-
93.1%
-
Zhou, W. et al. [ 26 ]
-
96.78%
96.89%
-
96.29%
-
Liaqat, S. et al. [ 27 ]
64.0%
60.27%
67.81%
-
75.0%
40.0%
Tao, Y. et al. [ 28 ]
57.9%
57.9%
56.3%
-
59.2%
56.6%
Table 4: Comparison among different works using Dataset 2.
Figure 4: Performance comparison results across two datasets.
Model
AUC Score
Accuracy
PPV
NPV
Sensitivity
Specificity
PMish
96.97%
92.66%
93.98%
91.90%
86.92%
96.38%
Relu
96.75%
92.18%
90.66%
93.15%
89.35%
94.03%
Silu
97.31%
93.52%
94.57%
92.90%
88.62%
96.70%
LRelu
97.84%
93.80%
95.79%
92.68%
88.14%
97.48%
Mish
96.84%
92.09%
91.67%
92.34%
87.89%
94.81%
Swish
97.04%
92.66%
93.75%
92.03%
87.17%
96.23%
Table 5: Comparison among different activation functions using Dataset 1.
Model
AUC Score
Accuracy
PPV
NPV
Sensitivity
Specificity
PMish
99.58%
97.12%
98.54%
95.73%
95.73%
98.54%
Relu
99.83%
98.08%
97.65%
98.52%
98.58%
97.56%
Silu
99.25%
97.00%
97.83%
96.16%
96.21%
97.80%
LRelu
99.56%
98.20%
98.34%
98.05%
98.10%
98.29%
Mish
99.60%
97.00%
98.30%
95.72%
95.73%
98.29%
Swish
99.32%
97.36%
97.85%
96.86%
96.92%
97.80%
Table 6: Comparison among different activation functions using Dataset 2.
Eye-movement tracking has emerged as a promising non-invasive approach to Autism Spectrum Disorder (ASD) screening, with systematic differences in attentional allocation and revisit behaviors observed during socially interactive tasks. Existing computational methods typically characterize eye-movements using discrete gaze trajectories and fixation events, yielding representations dominated by short-range temporal dynamics and limiting models that primarily emphasize long-range dependencies. Meanwhile, gaze behavior is naturally organized across semantically meaningful Areas of Interest (AOIs), whose attention allocation and transitions provide important structural cues, yet their relationships are rarely modeled explicitly. To address these limitations, we propose a structural face AOI-guided Eye-Gaze Track Network (AOI-Net) that jointly models short-term temporal dynamics and AOI-level structural organization. A network gating mechanism adaptively integrates the complementary temporal and structural representations according to their contributions to gaze-behavior characterization. To mitigate the pronounced class imbalance commonly encountered between individuals with ASD and Typically Developing (TD) participants in clinical datasets, class-distribution-aware learning is further employed to facilitate discriminative embedding learning under skewed class distributions. Experiments on a unique and large-scale clinical eye-tracking database comprising eight stimulus subsets and more than 1,300 participants show that AOI-Net consistently outperforms state-of-the-art methods. The proposed framework also enables interpretable gaze-behavior modeling and provides a practical basis for scalable AI-driven ASD screening in real-world healthcare. The code is available at https://github.com/Zhanpei-ai/CIM-AOI-Net/tree/main/Code
Zhanpei Huang, Binbin Sun, Jialiang Chen +5
School of Computer Science and Technology, Guangdong University of Technology, Guangzhou, China · Shenzhen Maternity and Child Healthcare Hospital Affiliated to Southern Medical University, Shenzhen, China · Department of Computer Science, Hong Kong Baptist University, Hong Kong SAR, China
Accurate Autism Spectrum Disorder (ASD) screening for school-age children is crucial to identify cases that may have been missed earlier and to enable timely interventions supporting social, cognitive, and academic development. Current ASD screening relies on subjective assessments and 2D analysis methods that fail to capture spatial displacement patterns characteristic of ASD behaviors. In this study, a novel 3D temporal analysis framework is presented, built on top of DECA (Detailed Expression Capture and Animation), a 3D modeling framework, to extract comprehensive head pose parameters (including translational components Tx,Ty,Tz) and facial expressions independent of pose variations. LSTM and GRU-based temporal classifiers were trained on the extracted 3D features from video data collected from 39 participants (19 ASD, 20 TD) aged 7-12 years during Virtual Reality-Continuous Performance Test tasks. The GRU-based models demonstrated superior performance, with 3D head pose features achieving 83.9% accuracy and 3D facial features reaching 81.4% accuracy, outperforming 2D baseline approaches by 10.7% and 7.5%, respectively. Furthermore, multimodal fusion of 3D head pose and facial features with PCA-based dimensionality reduction achieved the highest accuracy of 84.6%, outperforming unimodal approaches. This work establishes a foundation for objective, automated screening tools addressing current diagnostic limitations in ASD identification for school-age populations.
Inam Qadir, Elizabeth B Varghese, Dena Al-Thani +1
College of Science and Engineering, Hamad Bin Khalifa University, Qatar Foundation, Doha, Qatar
Accurate early screening for Autism Spectrum Disorder (ASD) is a precursor to timely intervention, which is critical for improving cognitive and behavioral outcomes. We present ASD-FEAT (ASD - Feature Extraction And Tracking), a multimodal dataset derived from video recordings of infant-adult interaction sessions. The key contribution of ASD-FEAT is the combination of longitudinal coverage from infancy through 36 months, repeated interaction sessions, clinically validated developmental outcomes, expert frame-level behavioral annotations, and privacy-conscious multimodal feature representations. To the best of our knowledge, existing ASD behavioral datasets do not jointly provide these characteristics at comparable scale. To demonstrate the utility of ASD-FEAT, we use it to evaluate a computer-vision-based end-to-end pipeline relying on machine learning techniques to automatically identify ASD risk. ASD-FEAT integrates both expert-defined and deep-learned features, including face and eye landmarks, facial action units, gaze direction, head position, mel-spectrogram audio representations, and optical flow, to identify behavioral markers of social interaction. Our automated pipeline achieves an ASD classification accuracy of 76.2% and an Area Under the Receiver Operating Characteristic (AUROC) of 0.82, compared to classifiers trained on manually labeled behaviors, which yielded 81.3% accuracy and an AUROC of 0.88. We further introduce a within-visit partner contrast: a per-visit signal contrasting examiner-directed and parent-directed social behavior which, when added to the classifier, lifts the fully automated Look Face + Smile and Look Face + Vocal configurations to Matthews correlations of 0.49 and 0.45 respectively, exceeding the human-coded single-partner baseline of 0.42.
Sidrah Liaqat, Halil Helvaci, Sen-Ching Cheung +4
University of Kentucky · Department of Electrical and Computer Engineering University of Kentucky Lexington, KY 40506, USA · University of California Davis +3