On December 2, 2025, the University of Science and Technology of Hanoi (USTH) successfully held the PhD dissertation defense for PhD Candidate Do Oanh Cuong from the Information and Communication Technology program. The dissertation was titled: “Deep Learning Methods for Image Analysis and Understanding: Application to Biomedicine.”
The research was supervised by Assoc. Prof. Tran Giang Son – Director of the Department of Information and Communication Technology, USTH, and Assoc. Prof. Luong Chi Mai from the Institute of Information Technology, Vietnam Academy of Science and Technology (VAST).
The defense committee consisted of seven distinguished members, chaired by Prof. Jean-Marc Lavest, Principle Rector of the USTH. The committee included Assoc. Prof. Nguyen Huu Quynh (CMC University), Prof. Jean-Christophe Burie (University of La Rochelle, France), Assoc. Prof. Nguyen Phuong Thai (University of Engineering and Technology, VNU), Assoc. Prof. Nguyen Long Giang (Institute of Information Technology, VAST), Dr. Nghiem Thi Phuong (USTH), and Dr. Tran Hoang Tung (USTH). The defense was also attended by the candidate’s academic supervisors, colleagues, family, and friends.
The Committee reviewed and approved the candidate’s academic background, research achievements, and study process. The members highly appreciated Do Oanh Cuong’s academic results, research outcomes, and dedicated efforts throughout the program.
During the session, PhD Candidate Do Oanh Cuong presented a summary of his dissertation and key findings. Nowadays, diagnostic imaging plays a pivotal role in the medical examination and treatment process. However, physicians often encounter challenges in synthesizing data from multiple sources or detecting subtle pathological signs in single images.
The dissertation, titled “Deep Learning Methods for Image Analysis and Understanding: Application to Biomedicine”, utilizes Artificial Intelligence (AI) to address these challenges with the following notable contributions:
Firstly, inspired by computational photography technology (similar to the Deep Fusion feature), we developed a deep learning model to fuse data from multi-modal images (such as MRI and PET). Not only does the system combine images but also improves visual information fidelity, enhances contrast and details, helping radiologists and doctors visualize lesions and diagnose more clearly (such as in Alzheimer’s diagnosis).
Secondly, another key highlight of the dissertation is the proposal of a method to combine two Deep Learning models – SwinTransformer and VGG19 – to improve COVID-19 screening on Chest X-rays. The SwinTransformer allows for the analysis of global features across lung regions, while VGG19 excels at detecting fine details. Leveraging these complementary strengths creates an automated diagnostic system achieving an accuracy of up to 99.32%, which is more stable and reliable than using individual models alone.
These advancements bring significant practical value to the medical field: providing high-quality images to support rapid decision-making, while simultaneously offering diagnostic support tools.
These findings of the dissertation have been published in 03 articles in international scientific journals and conference proceedings. According to the Committee’s assessment, the dissertation reflects the serious academic and research process of PhD Candidate Do Oanh Cuong. The research results possess high scientific value and practical applicability.
Following a closed discussion and secret ballot, the Committee approved the dissertation with a unanimous 7/7 vote. On behalf of the Committee, Prof. Jean-Marc Lavest congratulated Dr. Do Oanh Cuong on his successful defense and congratulated the Department of Information and Communication Technology on having another PhD graduate.
