Detection of Tumor-Infiltrating Lymphocytes in the Images of Immunohistochemistry using Deep Learning Techniques

Deep learning is playing a pivotal role in revolutionizing the field of medicine, particularly in its ability to enhance the interpretation of histopathological images and unravel the intricate relationship between the immune system and cancer. This technology has the potential to offer a comprehensive understanding of immune cell behavior within cancer, ultimately leading to more effective diagnostic and therapeutic strategies and, in turn, improving patient outcomes. This survey is dedicated to exploring the intricate dynamics of lymphocytes in balancing pro- and anti-inflammatory activities, underscoring the critical role of tumor-infiltrating lymphocytes (TILs) as essential cancer biomarkers. With the limitations of traditional methods such as immunohistochemistry and manual cell counting in mind, the medical community is increasingly turning to machine learning. The ultimate objective is achieving precise lymphocyte detection and spatial analysis through whole-slide imaging, promising substantial advancements across various medical domains, with a primary focus on enhancing patient quality of life and healthcare precision. Notably, the use of automated methods, particularly deep learning, is being explored to address inter-individual variability in the assessment of crucial prognostic factors like Ki-67 and TILs. The introduction of the LYON19 dataset and PathoNet addresses the demand for publicly available benchmarks, with PathoNet exhibiting superior performance to date, particularly in terms of the harmonic mean measure, thus contributing significantly to medical progress.