Research Experiences
- Utilize Multi-Instance learning techniques in digital pathology for prostate cancer molecular subtype classification, potentially improving accuracy and robustness of the model’s predictions.
- Introduce supervised End-to-end model for the tile-level features to improve the performance of the MIL model in the PAM50 dataset.
- Investigate the impact of pre-trained foundational models to the performance when combined with MIL for molecular subtype classification.
- Incorporate multi-scale representations of digital path imaging data to enhance the model’s ability to predict molecular subtypes in PAM50 dataset.
- Drafted a paper including an in-depth comparison of the proposed method with current state-of-the-art techniques, demonstrating its effectiveness through detailed ablation studies.
- Implement attention gates and residual connections within the U-Net architecture to better capture intricate tumor boundaries and improve segmentation accuracy, especially for small or complex regions.
- Implement weakly supervised methods for analyzing pathology images to predict two-year survival of patients diagnosed with brain tumors.
- Assisted in collecting data and determining performance within constrained movement environments.
- Developed a Matlab program to process and analyze EMG signals and filter the noise of EMG data.
- Designed and implemented a rehabilitation remote game for stroke patients with adaptable difficulty and real-time evaluation..
- Designed and constructed low-cost low fidelity adjustable mock-ups for human subject studies.
- Preliminary data presented at the CDHW Symposium in October 2022.
- Implemented Graph-based Clustering and k-means++ with a pre-generated anchor graph on Matlab.
- Collected and interpreted data, compared against the baseline algorithm in different data sets, and applied regression testing.
- Tested the effectiveness of L2 and L1 norms and verified that L21 norms outperformed other norms.
- Implemented KNN and PCA algorithms in Python and designed the lab procedure.
- Co-published reports on “Robust Landmark Graph-based Clustering for High-Dimensional Data”.
Publications
American Urological Association Annual Meeting, (submitted).
Human Factors and Ergonomics Society, 2023.
Neurocomputing Journals, 2022.
IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2019.