Équipe Images et Contenus (IC) : PHAM Tri-Cong

Thématiques de recherche : Pham Tri Cong is currently a researcher at the L3i Laboratory – University of La Rochelle, with two primary research directions: document analysis and medical image analysis.
In document analysis, he develops methods for incremental learning, novelty detection, and few-shot learning to enable robust document classification under limited data conditions. He also investigates table structure recognition in financial documents, aiming to automate structured data extraction for enterprise and banking applications.
In the healthcare domain, his work focuses on deep learning for medical imaging. His notable contributions include algorithms for skin cancer diagnosis from dermoscopy images, and more recently, an AI-powered real-time cancer screening solution using ultrasound and endoscopy images. This system supports earlier and more accurate tumor detection in clinical settings, with the potential to enhance screening efficiency and assist physicians in timely diagnosis.

Points forts des activités de recherche : "The distinctive strength of Pham Tri Cong’s research lies in his ability to integrate fundamental theoretical advances with real-world applications, spanning both document analysis and medical image analysis.

In document analysis, he has developed intelligent document processing methods to tackle practical challenges such as class imbalance, the emergence of new categories, and limited labeled data. His representative works include:

• Deep Metric Learning for End-to-End Document Classification
• Few-Shot Document Classification in Real Applications: Boosting Precision with Novelty Detection
• Exemplar Sampling Algorithm for Instance Incremental Learning on Imbalanced Document Datasets
• Incremental Learning and Ambiguity Rejection for Document Classification

Currently, his research focuses on table structure recognition in financial documents, with the goal of automating the extraction of structured information from corporate and banking reports. This line of work has strong implications for financial technology, auditing, and decision-making systems.

In parallel, his research in medical image analysis has resulted in impactful contributions. His study on melanoma detection from dermoscopy images demonstrated that his deep learning models could outperform dermatologists in cancer diagnosis, underscoring the transformative potential of AI in healthcare. More recently, he has advanced the development of an AI-assisted real-time cancer screening system based on ultrasound and endoscopy imaging. This innovation provides direct clinical value by enabling technicians to identify abnormal signs and suspicious tumor regions for closer inspection. The results can then be consolidated and escalated to specialists for final diagnosis. This product has strong potential for integration into cancer screening programs, supporting early detection and improving patient outcomes.

In recognition of these achievements, in January 2025, Pham Tri Cong was awarded the prestigious SIU Prize, an award that honors outstanding doctoral dissertations by Vietnamese and Vietnamese-descendant researchers worldwide in the field of Computer Science. The award not only acknowledges his research excellence but also affirms the potential of his work to be translated into innovative solutions with global impact.