Sparse Autoencoders for Expert-Assisted Interpretation of Histopathology Foundation Model Embeddings

Computer Science > Computer Vision and Pattern Recognition arXiv:2610.40905 (cs) [Submitted on 29 Oct 2026] Title: Sparse Autoencoders for Expert-Assisted Interpretation of Histopathology Foundation Model Embeddings Authors: Šimon Šmída, Wanda Benesova.

计算机科学 > 计算机视觉与模式识别 arXiv:2610.40905 (cs) [提交于 2026 年 10 月 29 日] 标题：用于专家辅助解读组织病理学基础模型嵌入的稀疏自编码器 作者：Šimon Šmída, Wanda Benesova。

Abstract: Histopathology foundation models learn rich embedding representations, but their internal structure remains difficult to inspect. Concept-level interpretability is particularly relevant in computational pathology because pathologists reason through recurring morphological and tissue-level patterns, including tissue architecture, cellular morphology, staining variation, and artifacts.

摘要：组织病理学基础模型学习到了丰富的嵌入表示，但其内部结构仍然难以检查。概念层面的可解释性在计算病理学中尤为重要，因为病理学家通过反复出现的形态学和组织层面模式进行推理，这些模式包括组织结构、细胞形态、染色差异和伪影。

In this work, we investigate whether sparse autoencoders (SAEs) can decompose embeddings from histopathology foundation models into sparse latent features suitable for expert-assisted interpretation. We train vanilla and Batch-TopK SAE variants on NCT-CRC embeddings, using CONCH as the primary foundation model and UNI2-h as an additional robustness evaluation, and characterize the resulting latent features through reconstruction fidelity, sparsity, coherence, label enrichment, and downstream class-signal preservation.

在这项工作中，我们研究了稀疏自编码器（SAE）是否能将组织病理学基础模型的嵌入分解为适合专家辅助解读的稀疏潜在特征。我们在 NCT-CRC 嵌入上训练了普通版和 Batch-TopK 版的 SAE 变体，使用 CONCH 作为主要基础模型，并以 UNI2-h 作为额外的稳健性评估，通过重构保真度、稀疏性、连贯性、标签富集度和下游类别信号保留情况来表征所得的潜在特征。

The learned sparse representations preserve most downstream tissue-class information, and their latent features retrieve more coherent and label-concentrated patch groups than raw embedding dimensions or PCA components. Discussions with pathology experts further motivated a WSI-level latent visualization workflow that places feature activations into larger tissue context.

学习到的稀疏表示保留了大部分下游组织类别信息，且其潜在特征比原始嵌入维度或 PCA 分量能检索到更连贯、标签更集中的切片组。与病理学专家的讨论进一步推动了一种全切片图像（WSI）层面的潜在可视化工作流程，该流程将特征激活置于更大的组织背景中。

Overall, the additional UNI2-h evaluation suggests that the observed behavior is not specific to CONCH, supporting the use of sparse autoencoders for expert-assisted analysis of histopathology foundation-model embeddings, while emphasizing that latent features should be treated as candidate concepts requiring pathological assessment rather than automatically validated histopathological concepts.

总体而言，额外的 UNI2-h 评估表明所观察到的行为并非 CONCH 所特有，这支持了使用稀疏自编码器进行组织病理学基础模型嵌入的专家辅助分析，同时也强调潜在特征应被视为需要病理评估的候选概念，而非自动验证的组织病理学概念。