Vuthea Chheang

Jicheng Li, Vuthea Chheang, Pinar Kullu et al.

Autism spectrum disorder (ASD) is a developmental disorder characterized by significant social communication impairments and difficulties perceiving and presenting communication cues. Machine learning techniques have been broadly adopted to facilitate autism studies and assessments. However, computational models are primarily concentrated on specific analysis and validated on private datasets in the autism community, which limits comparisons across models due to privacy-preserving data sharing complications. This work presents a novel privacy-preserving open-source dataset, MMASD as a MultiModal ASD benchmark dataset, collected from play therapy interventions of children with Autism. MMASD includes data from 32 children with ASD, and 1,315 data samples segmented from over 100 hours of intervention recordings. To promote public access, each data sample consists of four privacy-preserving modalities of data; some of which are derived from original videos: (1) optical flow, (2) 2D skeleton, (3) 3D skeleton, and (4) clinician ASD evaluation scores of children, e.g., ADOS scores. MMASD aims to assist researchers and therapists in understanding children's cognitive status, monitoring their progress during therapy, and customizing the treatment plan accordingly. It also has inspiration for downstream tasks such as action quality assessment and interpersonal synchrony estimation. MMASD dataset can be easily accessed at https://github.com/Li-Jicheng/MMASD-A-Multimodal-Dataset-for-Autism-Intervention-Analysis.

Alex Apffel, Huy Tran, Vuthea Chheang

In this work, we present a multimodal data acquisition workflow for the digital preservation and virtual reconstruction of at-risk historical sites in the island of Nevis. Facing threats from coastal erosion, rising sea levels, and aggressive vegetation, the archaeological heritage of Nevis requires documentation strategies that bridge the gap between high-cost professional surveying and consumer accessibility. Experimental test compared acquisition variables, specifically camera height (1m vs. 3m) and operator trajectory against high-resolution control data. Moreover, we explore the virtual reconstruction between mesh reconstruction and 3D gaussian splatting to serve as different modalities for documentation. The resulting data is fused into immersive virtual reality (VR) environments, offering a scalable, non-proprietary model for democratizing digital heritage in the Caribbean.

Martin Sanchez, Nick Tran, Vuthea Chheang

Hospital readmissions remain a challenge for healthcare systems, especially among patients with chronic conditions such as diabetes. Unplanned readmissions within 30 days are costly, strain hospital resources, and can indicate poor care coordination or discharge planning. In this work, we explore the use of machine learning to predict readmission risk for diabetic inpatients and propose a mixed reality (MR) to provide effective visualization and insights. We trained an XGBoost classifier after data cleaning, encoding, and feature engineering. The model achieved an Area Under the Receiver Operating characteristic Curve (AUROC) of 0.72 and an Area Under the Precision-Recall Curve (AUPRC) of 0.11. Key predictive factors included prior inpatient visits, discharge disposition, and glycemic control indicators such as A1C (blood sugar test) results and medication adjustments. Additionally, we developed an MR prototype that visualize patient records and predictions containing risk level, major contributing factors, and a concise summary of care. Together, the predictive model and the MR interface aim to improve clinician awareness and communication around readmission risk in real-time clinical settings.