cs.CVNov 16, 2026

Interpretable deep concepts of organoid morphology

Authors: Lion Jasper Gleiter, Tingying Peng

Organizations: Helmholtz Munich, Helmholtz AI, Neuherberg, Germany · Technical University of Munich, School of CIT, Munich, Germany

Abstract

Image-based classification of organoid morphologies can indicate drug response, viability, or successful differentiation. In recent work, learned representations outperform handcrafted features as the basis for downstream classifiers. However, their black-box nature hides the morphological cues a classifier relies on. Based on the foundation models DINOv3 (general-purpose) and ViTAMIn-O (organoid-specific), we train sparse autoencoders (SAEs) to learn interpretable concepts related to distinct organoid morphologies. We annotate concepts with spatial labels (organoid, organoid wall, background) and compute their dataset specificity and task relevance across five tasks. The spatial labels of DINOv3 concepts yield coarse organoid segmentations on out-of-domain datasets. Filtering concepts by spatial label in the DINOv3-SAE improves downstream AUROC for three of the five tasks. Qualitative analysis of the most relevant concepts further reveals a confounding factor introduced by our preprocessing in one task. Through such insights, interpretable concepts could help improve the reliability of classifiers and guide the future development of organoid representation models. The trained SAEs will be made available upon publication.

Figures & tables

Explore similar work

Sep 8, 2026

MorphoOrgaAgent: A Foundation-Model-Based Multi-Agent System for Autonomous Organoid Analysis

Organoids are three-dimensional tissue models whose morphology provides important insights into tumor development, disease progression, and drug testing. Extracting these morphological features relies heavily on manual segmentation, which is time-consuming and labor-intensive. Furthermore, performing quantitative statistical analysis typically requires custom coding skills and a mathematical background, presenting a major barrier for experimental biologists. To address these challenges, we introduce MorphoOrgaAgent, a multi-agent framework that achieves zero-shot organoid segmentation, automated data analysis, and report generation based on natural language input. The framework consists mainly of three core components: a TaskUnderstandingAgent that identifies requested measurements and visualization types; a hybrid segmentation module that combines Cellpose-derived geometric prompts with text prompts to guide SAM3 for zero-shot organoid instance segmentation; and a ReportAgent that computes quantitative metrics and compiles them alongside generated visualizations into a structured report. We further introduce MorphoOrgaVQA, a benchmark designed for quantitative evaluation of agent systems in organoid morphology analysis. Experimental results demonstrate that MorphoOrgaAgent handles both explicit and descriptive user requests, produces measurements closely matching ground truth, and generates complete analysis reports without requiring manual programming. The complete source code and MorphoOrgaVQA benchmark are publicly available at https://github.com/peng-lab/MorphoOrgaAgent.
Mar 24, 2026

Sparse Autoencoders for Interpretable Medical Image Representation Learning

Vision foundation models (FMs) achieve state-of-the-art performance in medical imaging. However, they encode information in abstract latent representations that clinicians cannot interrogate or verify. The goal of this study is to investigate Sparse Autoencoders (SAEs) for replacing opaque FM image representations with human-interpretable, sparse features. We train SAEs on embeddings from BiomedParse (biomedical) and DINOv3 (general-purpose) using 909,873 CT and MRI 2D image slices from the TotalSegmentator dataset. We find that learned sparse features: (a) reconstruct original embeddings with high fidelity (R2 up to 0.941) and recover up to 87.8% of downstream performance using only 10 features (99.4% dimensionality reduction), (b) preserve semantic fidelity in image retrieval tasks, (c) correspond to specific concepts that can be expressed in language using large language model (LLM)-based auto-interpretation. (d) bridge clinical language and abstract latent representations in zero-shot language-driven image retrieval. Our work indicates SAEs are a promising pathway towards interpretable, concept-driven medical vision systems. Code repository: https://github.com/pwesp/sail.
Apr 2, 2026

Automatic Image-Level Morphological Trait Annotation for Organismal Images

Morphological traits are physical characteristics of biological organisms that provide vital clues on how organisms interact with their environment. Yet extracting these traits remains a slow, expert-driven process, limiting their use in large-scale ecological studies. A major bottleneck is the absence of high-quality datasets linking biological images to trait-level annotations. In this work, we demonstrate that sparse autoencoders trained on foundation-model features yield monosemantic, spatially grounded neurons that consistently activate on meaningful morphological parts. Leveraging this property, we introduce a trait annotation pipeline that localizes salient regions and uses vision-language prompting to generate interpretable trait descriptions. Using this approach, we construct Bioscan-Traits, a dataset of 80K trait annotations spanning 19K insect images from BIOSCAN-5M. Human evaluation confirms the biological plausibility of the generated morphological descriptions. We assess design sensitivity through a comprehensive ablation study, systematically varying key design choices and measuring their impact on the quality of the resulting trait descriptions. By annotating traits with a modular pipeline rather than prohibitively expensive manual efforts, we offer a scalable way to inject biologically meaningful supervision into foundation models, enable large-scale morphological analyses, and bridge the gap between ecological relevance and machine-learning practicality.