Linhao Qu | ViLab https://vilab.team/author/linhao-qu/ Linhao Qu Hugo Blox Builder (https://hugoblox.com)en-usMon, 27 Oct 2025 00:00:00 +0000 https://vilab.team/media/icon_hu2896232876136423579.png Linhao Qu https://vilab.team/author/linhao-qu/ Enhancing zero-shot brain tumor subtype classification via fine-grained patch-text alignment https://vilab.team/publication/enhancing-zero-shot-brain-tumor-subtype-classification-via-f/ Mon, 27 Oct 2025 00:00:00 +0000 https://vilab.team/publication/enhancing-zero-shot-brain-tumor-subtype-classification-via-f/ <p>本文提出细粒度补丁对齐网络（FG-PAN），用于脑肿瘤亚型的零样本分类。该方法包含局部特征细化模块，通过建模代表性补丁间的空间关系增强视觉特征；以及细粒度文本描述生成模块，利用大语言模型生成病理感知的类别语义原型。通过对齐细粒度视觉与语义特征，并引入坐标感知聚合机制，FG-PAN在整张病理切片级别实现了更准确的亚型判别，缓解了标注数据稀缺和形态差异细微带来的挑战。</p> Semamil: Semantic reordering with retrieval-guided state space modeling for whole slide image classification https://vilab.team/publication/semamil-semantic-reordering-with-retrieval-guided-state-spac/ Fri, 01 Aug 2025 00:00:00 +0000 https://vilab.team/publication/semamil-semantic-reordering-with-retrieval-guided-state-spac/ <p>本文针对全切片图像分类中多实例学习忽略上下文、Transformer计算复杂、状态空间模型打乱语义顺序的问题，提出SemaMIL方法。该方法包含语义重排模块，通过可逆置换将语义相似的图像块聚类排列；以及语义引导检索状态空间模块，选择代表性查询子集调整状态空间参数，实现高效全局建模。在四个WSI数据集上验证了方法的有效性。</p>