Programa del congreso
Resúmenes y datos de las sesiones para este congreso. Seleccione una fecha o ubicación para mostrar solo las sesiones en ese día o ubicación. Seleccione una sola sesión para obtener una vista detallada (con resúmenes y descargas, si están disponibles).
|
Resumen diario |
| Sesión | |
|
Ju-S4.1-IAIm: Inteligencia artifical en imagen médica (II) Lugar: Salón de Actos Presidente de la sesión: Monica Abella García Presidente de la sesión: José Antonio Pérez Carrasco | |
    | |
9:30 - 9:45
Coloring Cancer: GAN Powered Biomarker Synthesis from H&E Stains 1: University of Castilla-La Mancha, Ciudad Real, Spain; 2: University Hospital of Jerez de la Frontera, Spain; 3: General University Hospital of Ciudad Real, Spain This study investigates digital staining techniques for breast cancer biopsy samples, with the aim of virtually generating immunohistochemical stains for key biomarkers such as HER2, Ki67, PR, and ER from routine hematoxylin and eosin (H&E) whole-slide images. Generative adversarial networks (GANs) are employed to transform H\&E-stained WSIs into synthetic IHC-stained images. Four models were evaluated and compared in terms of their ability to preserve morphological features and replicate the appearance of real IHC stains. The models evaluated in this work are CycleGAN, CUT, StainGAN, and HistAuGAN. The goal is to reduce reliance on costly biomarker staining procedures by providing an efficient and accessible alternative through digital pathology. The results show that CUT and HistAuGAN achieve superior preservation of cellular morphology and improved visual consistency relative to real IHC images. Although preliminary, these findings highlight the potential of digital staining techniques to be integrated into clinical workflows, reducing both costs and diagnostic time.
| |