AI For Healthcare
With the advent of advanced AI technologies, the world of healthcare is undergoing a transformative revolution. Unleashing the power of medical images and high-throughput sequencing data, these cutting-edge AI technologies are empowering healthcare professionals like never before. They offer unparalleled accuracy in diagnoses, personalized treatment plans, and real-time monitoring on an unprecedented scale. Yet, as we witness this remarkable progress, the journey towards perfecting AI healthcare encounters its fair share of obstacles. Ensuring data quality and managing its diversity, minimizing the need for expert knowledge, addressing privacy concerns, and demanding transparency and explainability are just some of the challenges that researchers and innovators in the field must overcome.
Our team is dedicated to advancing computer-aided diagnosis and integrating multi-omics data to enable personalized healthcare solutions that are precise, effective, and accessible to all. To achieve this ambitious goal, we confront two critical challenges head-on. First of all, the contradiction between high-dimensional features and scarcity of labeled data. To overcome this challenge, we delve into the intrinsic graph-structure nature of data and identify reproducible biomarkers specific to each task. By leveraging these insights, we enhance the representation of features, propelling us towards superior results. We further employ advanced machine learning algorithms such as federated learning, multi-task learning, and semi-supervised learning to achieve exceptional generalization even with limited labeled data. The second challenge lies in integrating diverse data sources into robust multimodal AI systems. We tackle this complexity by developing cutting-edge AI techniques for preprocessing, aligning, and fusing multimodal data. This allows us to harmonize the heterogeneity introduced by different data formats, paving the way for comprehensive and holistic analyses. Our team is driven by the desire to conquer these technical obstacles and unlock the boundless potential of AI in healthcare, bringing us closer to a future where better healthcare are within reach for all.
Publications
[1] Yao Hu, Z.-A. Huang, R. Liu, X. Xue, X. Sun, L, Song, K. C. Tan, “Source Free Semi-Supervised Transfer Learning for Diagnosis of Mental Disorders on fMRI Scans,” in IEEE Transactions on Pattern Analysis and Machine Intelligence, doi: 10.1109/TPAMI.2023.3298332, early access, July 2023.
[2] Z.-A. Huang, R. Liu, Z. Zhu, K.-C. Tan, “Multitask Learning for Joint Diagnosis of Multiple Mental Disorders in Resting-State fMRI,” in IEEE Transactions on Neural Networks and Learning Systems, doi: 10.1109/TNNLS.2022.3225179, early access, Dec. 2022.
[3] Z.- A. Huang, Y. Hu, R. Liu, X. Xue, Z. Zhu, L. Song, K. C. Tan, “Federated Multi-Task Learning for Joint Diagnosis of Multiple Mental Disorders on MRI Scans,” in IEEE Transactions on Biomedical Engineering, vol. 70, no. 4, pp. 1137-1149, Sep. 2022.
[4] Z.- A. Huang, J. Zhang, Z. Zhu, E. Q. Wu, K. C. Tan, “Identification of Autistic Risk Candidate Genes and Toxic Chemicals via Multilabel Learning,” in IEEE Transactions on Neural Networks and Learning Systems, vol. 32, no. 9, pp. 3971-3984, Aug. 2020.
[5] Z.- A. Huang, Z. Zhu, C. H. Yau, K. C. Tan, “Identifying autism spectrum disorder from resting-state fMRI using deep belief network,” in IEEE Transactions on Neural Networks and Learning Systems, vol. 32, no. 7, pp. 2847-2861, July 2020.