The field of medicine is witnessing a significant integration of Artificial Intelligence (AI) technologies. Novel Computer-Aided Design (CAD) methods are emerging as valuable tools to support pathologists, easing their workload and reducing operational costs.Yet, the creation of specialized solutions tailored to specific tasks, such as aiding in the diagnosis of Follicular Lymphoma (FL), faces hindrances including data collection and annotation challenges. In response to the intricate demands of pathologists’ tasks, we propose an AI-driven approach for detecting centrocytes (CBs) and centroblasts (CCs) in FL cases. Notably, identifying centroblasts poses a noteworthy challenge, even for experienced professionals, due to potential confusion with centrocytes. Our method is devised to identify cells and assign them into three distinct categories: centroblasts, centrocytes or other. While existing studies predominantly concentrate on centroblast detection, our investigation extends to evaluating its effectiveness in categorizing centroblasts against non-centroblast cells—encompassing centrocytes and other classes—aligning with configurations outlined in current literature. Under this specific setup, we achieved an F1-score, precision, and recall of 63%. Our exploration into centroblast detection using the proposed approach highlights the practicality of detecting centroblasts within whole slide images (WSIs) scanned at x20 resolution. Despite leveraging lower-resolution images, we have showcased the ability to achieve centroblast classification results. Looking ahead, the proposed algorithm holds potential for extension to support follicular lymphoma grading.