Tariq Adnan

Tariq Adnan

PhD candidate in Computer Science (ROC-HCI Lab, University of Rochester).

Publications

You can also find my articles on my Google Scholar profile.

Journal Articles

AI-Enabled Parkinson’s Disease Screening Using Smile Videos

Published in NEJM AI, 2025

We developed a facial micro-expression analysis framework to distinguish Parkinson’s disease (PD) using short smile videos, achieving strong classification accuracy and generalizability across international cohorts. The model attained $\mathbf{\sim88\%}$ accuracy using only consumer-grade cameras, demonstrating the potential for remote and accessible PD screening.

Recommended citation: Adnan, T., Islam, M.S., Lee, S., Wasifur Rahman Chowdhury, E.M., Tithi, S.D., Noshin, K., Islam, M.R., Sarker, I., Rahman, M.S., Schneider, R.B. and Adams, J.L., 2025. AI-Enabled Parkinson’s Disease Screening Using Smile Videos. NEJM AI, 2(7), p.AIoa2400950.
Download Paper

A novel fusion architecture for detecting Parkinson’s Disease using semi-supervised speech embeddings

Published in npj Parkinson’s Disease, 2025

This work introduces a speech-based Parkinson’s disease screening framework using English pangram utterances from $\mathbf{1,306}$ participants. By fusing semi-supervised Wav2Vec2 and ImageBind speech embeddings into a multimodal classifier, the model achieved $\mathbf{\sim89\%}$ AUROC and $\mathbf{\sim86\%}$ accuracy, with strong external robustness.

Recommended citation: Adnan, T., Abdelkader, A., Liu, Z., Hossain, E., Park, S., Islam, M.S. and Hoque, E., 2025. A novel fusion architecture for detecting Parkinson’s Disease using semi-supervised speech embeddings. npj Parkinson's Disease, 11(1), p.176.
Download Paper

UACD: A Local Approach for Identifying the Most Influential Spreaders in Twitter in a Distributed Environment

Published in Social Network Analysis and Mining, 2022

We propose UACD, a novel method for identifying the most influential spreaders in large social networks by combining user-specific attributes with topological structure. Unlike traditional global centrality methods, UACD uses only local node information and offers scalability to large graphs while maintaining superior ranking performance. A distributed implementation on Amazon EC2 shows that UACD is on average $\mathbf{12.5\%}$ more accurate and $\mathbf{175\times}$ faster than state-of-the-art alternatives for influence ranking on large Twitter datasets.

Recommended citation: Adnan, T.T., Islam, M.S., Papon, T.I., Nath, S. and Adnan, M.A., 2022. Uacd: a local approach for identifying the most influential spreaders in twitter in a distributed environment. Social Network Analysis and Mining, 12(1), p.37.
Download Paper

Fast, Scalable and Geo-Distributed PCA for Big Data Analytics

Published in Information Systems, 2021

We propose TallnWide, a fast, scalable geo-distributed principal component analysis (PCA) algorithm that efficiently computes PCA on extremely high-dimensional data without intermediate memory overflow by dividing blocks of data and minimizing communication overhead. The method handles $\mathbf{10\times}$ higher dimensionality with $\mathbf{2.9\times}$ less time compared to baselines, enabling practical deployment in distributed big data settings.

Recommended citation: Adnan TMT, Tanjim MM, Adnan MA. Fast, Scalable and Geo-Distributed PCA for Big Data Analytics. Information Systems. 2021;98:101710.
Download Paper

Conference Papers

Accessible, at-home detection of Parkinson’s disease via multi-task video analysis

Published in AAAI, 2025

We introduce a large-scale multi-task video dataset ($\mathbf{3,306}$ videos from $\mathbf{845}$ participants) and a novel fusion model, UFNet, for multimodal PD screening using webcam video. UFNet significantly outperformed single-task models in diagnostic accuracy (AUROC $\mathbf{\sim93\%}$), demonstrating scalable at-home detection.

Recommended citation: Islam, M.S., Adnan, T., Freyberg, J., Lee, S., Abdelkader, A., Pawlik, M., Schwartz, C., Jaffe, K., Schneider, R.B., Dorsey, R. and Hoque, E., 2025, April. Accessible, at-home detection of Parkinson’s disease via multi-task video analysis. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 39, No. 27, pp. 28125-28133).
Download Paper

Hi5: 2D Hand Pose Estimation with Zero Human Annotation

Published in ACII, 2025

We introduce Hi5, a synthetic hand pose estimation dataset of $\mathbf{583,000}$ images generated without human annotation, leveraging high-fidelity 3D hand models with diverse genders and skin tones, dynamic environments, and consumer-grade rendering. The data synthesis pipeline precisely controls pose diversity and representation, enabling robust model training. Models trained on Hi5 outperform real-data baselines on occluded and perturbed real benchmarks, demonstrating synthetic data’s potential for scalable hand pose estimation.

Recommended citation: Hasan, M., Ozel, C., Long, N., Martin, A., Potter, S., Adnan, T., Lee, S., Zadeh, A. and Hoque, E., 2024. Hi5: 2D Hand Pose Estimation with Zero Human Annotation. arXiv preprint arXiv:2406.03599.
Download Paper

A user-centered framework to empower people with parkinson’s disease

Published in IMWUT, 2024

This study evaluated user reactions to AI-driven PD screening results and communication strategies in unsupervised home settings. The findings highlight design considerations that enhance user autonomy and trust, informing future deployments of remote PD screening tools.

Recommended citation: Rahman, W., Abdelkader, A., Lee, S., Yang, P., Islam, M.S., Adnan, T., Hasan, M., Wagner, E., Park, S., Dorsey, E.R. and Schwartz, C., 2024. A user-centered framework to empower people with parkinson's disease. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 7(4), pp.1-29.
Download Paper

Preprints

Remote AI Screening for Parkinson’s Disease: A Multimodal, Cross-Setting Validation Study

Published in Preprint (PMC / AI Screening), 2025

PARK is a web-based artificial intelligence (AI) tool for remote screening of Parkinson’s disease (PD) using video and audio recordings collected via webcam, including speech, facial expressions, and motor tasks. The system was evaluated on $\mathbf{1,865}$ participants across diverse demographic groups and settings, including supervised clinical environments and unsupervised home use. PARK achieved consistent performance with accuracy ranging from $\mathbf{80.2\%–80.6\%}$ and AUROC between $\mathbf{0.85–0.87}$ on multiple independent test sets, and demonstrated high agreement with movement disorder specialists on held-out assessments. The model generalized well across age, sex, and ethnic subgroups, included uncertainty-aware predictions to support safe use in unsupervised environments, and showed favorable usability and participant satisfaction, highlighting its potential as an accessible, scalable tool for remote PD screening.

Recommended citation: Islam, M.S., Adnan, T., Abdelkader, A., Liu, Z., Ma, E., Park, S., Azad, A., Liu, P., Pawlik, M., Hartman, E. and Shelton, E., 2025. Remote AI Screening for Parkinson’s Disease: A Multimodal, Cross-Setting Validation Study. Research Square, pp.rs-3.
Download Paper