arXiv:2510.18637v2 Announce Type: replace-cross 
Abstract: Semantic segmentation of electron microscopy (EM) images of biological samples remains a challenge in the life sciences. EM data captures details of biological structures, sometimes with such complexity that even human observers can find it overwhelming. We introduce {\epsilon}-Seg, a method based on hierarchical variational autoencoders (HVAEs), employing center-region masking, sparse label contrastive learning (CL), a Gaussian mixture model (GMM) prior, and clustering-free label prediction. Center-region masking and the inpainting loss encourage the model to learn robust and representative embeddings to distinguish the desired classes, even if training labels are sparse (0.05% of the total image data or less). For optimal performance, we employ CL and a GMM prior to shape the latent space of the HVAE such that encoded input patches tend to cluster wrt. the semantic classes we wish to distinguish. Finally, instead of clustering latent embeddings for semantic segmentation, we propose a MLP semantic segmentation head to directly predict class labels from latent embeddings. We show empirical results of {\epsilon}-Seg and baseline methods on 2 dense EM datasets of biological tissues and demonstrate the applicability of our method also on fluorescence microscopy data. Our results show that {\epsilon}-Seg is capable of achieving competitive sparsely-supervised segmentation results on complex biological image data, even if only limited amounts of training labels are available.

طور الباحثون طريقة جديدة تُدعى ε-Seg لمواجهة التحديات المتعلقة بالتجزئة الدلالية في صور المجهر الإلكتروني للعينات البيولوجية. تعتمد هذه الطريقة المبتكرة على استخدام مشفرات متغيرة هرمية وتقنيات متقدمة مثل إخفاء منطقة المركز والتعلم التبايني مع تسميات نادرة. تكمن أهمية ε-Seg في قدرتها على تحسين تحليل الهياكل البيولوجية المعقدة، مما يسهل على العلماء تفسير البيانات المعقدة التي كانت في السابق مرهقة.

Investigadores han desarrollado un nuevo método llamado ε-Seg para abordar los desafíos de la segmentación semántica en imágenes de microscopía electrónica de muestras biológicas. Este enfoque innovador utiliza autoencoders variacionales jerárquicos y técnicas avanzadas como el enmascaramiento de región central y el aprendizaje contrastivo con etiquetas escasas. La importancia de ε-Seg radica en su potencial para mejorar el análisis de estructuras biológicas complejas, facilitando a los científicos la interpretación de datos intrincados que antes eran abrumadores.

Des chercheurs ont développé une nouvelle méthode appelée ε-Seg pour relever les défis de la segmentation sémantique dans les images de microscopie électronique de spécimens biologiques. Cette approche innovante utilise des autoencodeurs variationnels hiérarchiques et des techniques avancées telles que le masquage de région centrale et l'apprentissage contrastif avec des étiquettes rares. L'importance d'ε-Seg réside dans son potentiel à améliorer l'analyse de structures biologiques complexes, facilitant ainsi l'interprétation de données complexes qui étaient auparavant écrasantes.

Researchers have developed a new method called ε-Seg to tackle the challenges of semantic segmentation in electron microscopy images of biological samples. This innovative approach utilizes hierarchical variational autoencoders and advanced techniques like center-region masking and sparse label contrastive learning. The significance of ε-Seg lies in its potential to enhance the analysis of complex biological structures, making it easier for scientists to interpret intricate data that was previously overwhelming.

{\epsilon}-Seg: Sparsely Supervised Semantic Segmentation of Microscopy Data

Was this article worth reading? Share it

LucidQuery AI

The Visualizer

Attentive AI

Supametas.AI

Zemith-3bda3b

SVGenius

Ready to build your own newsroom?