Próxima Charla ReLeLa: IA Estamos! LoAC: Length-Optimized Active Contour Loss for Boundary Segmentation in Biomedical Images

Estimada comunidad,

Tenemos el agrado de invitarles a participar en una nueva sesión de la Charla ReLeLa: IA Estamos!

En esta oportunidad nos acompañará Maximiliano Westerhout, Asistente de Investigación y Magíster en Data Science, quien nos presentará una innovadora propuesta metodológica para mejorar la segmentación de imágenes biomédicas, preservando detalles finos y estructuras geométricas.

 Detalles del Evento

Nombre de la Charla: LoAC: Length-Optimized Active Contour Loss for Boundary Segmentation in Biomedical Images

Expositor: Maximiliano Westerhout, Research Assistant & Master in Data Science.

Fecha: Miércoles 7 de octubre 2026

Hora: 12:30 horas.

Lugar: Sala P303 - Philippe Flajolet, Departamento de Ciencias de la Computación (DCC), 3er piso, Edificio Poniente (Av. Beauchef 851, Santiago).

Abstract

How can we improve biomedical image segmentation while preserving fine details and geometric structure?

This talk introduces LoAC (Length Optimized Active Contour Loss), a novel loss function designed to improve boundary delineation and contour consistency in biomedical image segmentation. LoAC enhances the optimization of active-contour-based losses by incorporating multi-orientation gradient information, magnitude and orientation constraints, and normalization strategies.

Experiments on the PH2 skin lesion segmentation dataset show that LoAC achieves strong performance, with the highest mean Dice Score, IoU, and Recall among the evaluated methods, including Cross Entropy, Dice, NAC, and ABeDice. The results highlight the potential of geometry-aware loss functions for producing more accurate and consistent biomedical image segmentation.

Speaker Bio: Maximiliano Westerhout holds a Master’s degree in Data Science (MDS) and currently works as a Research Assistant.

¡Los y las esperamos para seguir conversando y aprendiendo sobre las últimas aplicaciones de la Inteligencia Artificial en biomedicina!

Saludos cordiales,