[M2 internship] in artificial intelligence applied to digital health

Analysis and integration of class imbalance in deep learning architectures for melanoma detection

Keywords: Class Imbalance, Long-Tailed Learning, Deep Learning, Melanoma, Medical Imaging.

Internship Topic

Melanoma is an aggressive and potentially fatal skin cancer, representing a major public health issue with an increasing incidence in France. Computer-Aided Diagnosis (CAD) systems, particularly those based on deep neural networks applied to dermoscopic images, have shown promising performance for early melanoma detection.

However, datasets used in this context are often highly imbalanced, as some lesion categories are much rarer than others. This imbalance introduces significant bias in model training and degrades performance on minority classes. Numerous approaches have been proposed in the literature to address this issue, including resampling strategies, loss re-weighting, and decoupled learning [1, 2]. In this internship, the objective is to further investigate loss-function-based approaches, particularly margin-based loss functions [3, 4]. For instance, modifications of the cross-entropy loss will be explored to enforce larger margins between rare and dominant classes, inspired by recent advances in long-tailed visual recognition [5].

Internship Objective

To study, develop, and integrate class-imbalance-aware loss functions into deep learning architectures for dermoscopic image classification.

Methodology

• State-of-the-art review on class imbalance and long-tailed learning.

• Implementation of advanced loss functions (margin-based loss, re-weighting).

• Training and evaluation of deep neural networks on imbalanced datasets.

• Comparative performance analysis on minority and majority classes.

Expected Outcomes

• Improved robustness of models to class imbalance.

• Enhanced classification performance on minority classes.

• Potential scientific publication.

Lab presentation

CESI LINEACT (UR 7527), Laboratory for Digital Innovation for Businesses and Learning to Support the Competitiveness of Territories, anticipates and accompanies the technological mutations of sectors and services related to industry and construction. The historical proximity of CESI with companies is a determining element for our research activities. It has led us to focus our efforts on applied research close to companies and in partnership with them. A human-centered approach coupled with the use of technologies, as well as territorial networking and links with training, have enabled the construction of cross-cutting research; it puts humans, their needs and their uses, at the center of its issues and addresses the technological angle through these contributions.

Its research is organized according to two interdisciplinary scientific teams and several application areas.

• Team 1 "Learning and Innovating" mainly concerns Cognitive Sciences, Social Sciences and Management Sciences, Training Techniques and those of Innovation. The main scientific objectives are the understanding of the effects of the environment, and more particularly of situations instrumented by technical objects (platforms, prototyping workshops, immersive systems...) on learning, creativity and innovation processes.

• Team 2 "Engineering and Digital Tools" mainly concerns Digital Sciences and Engineering. The main scientific objectives focus on modeling, simulation, optimization and data analysis of cyber physical systems. Research work also focuses on decision support tools and on the study of human-system interactions in particular through digital twins coupled with virtual or augmented environments.

Research intersects across the application domains of the Factory of the Future and the City of the Future.

Bibliography

[1] Lu YANG et al. « A Survey on Long-Tailed Visual Recognition ». en. In : Int J Comput Vis 130.7 (juill. 2022), p. 1837-1872. ISSN : 1573-1405. DOI : 10.1007/ s11263-022-01622-8.

[2] Yifan ZHANG et al. Deep Long-Tailed Learning : A Survey. arXiv :2110.04596 [cs]. Oct. 2021. DOI : 10.48550/arXiv.2110.04596.

[3] Foahom Gouabou, A. C., Iguernaissi, R., Damoiseaux, J. L., Moudafi, A., & Merad, D. (2022). End-to-End Decoupled Training: A Robust Deep Learning Method for Long-Tailed Classification of Dermoscopic Images for Skin Lesion Classification. Electronics, 11(20), 3275

[4] Foahom Gouabou, A. C., Iguernaissi, R., Damoiseaux, J. L., Moudafi, A., & Merad, D. (2022). End-to-End Decoupled Training: A Robust Deep Learning Method for Long-Tailed Classification of Dermoscopic Images for Skin Lesion Classification. Electronics, 11(20), 3275

[5] Youngkyu HONG et al. « Disentangling label distribution for long-tailed visual recognition ». In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2021, p. 6626-6636. DOI : 10.48550/arXiv.2012. 00321.