<p>Microscopic hyperspectral imaging (MHSI) of unstained tissue provides quantitative, label-free cues for pathology, but practical diagnosis is hindered by weak morphological contrast and high-dimensional spectra. Patch-wise classification is therefore unstable: discriminative spectral signatures are subtle, spatially sparse, and easily confounded by noise and tissue heterogeneity. To address this, we construct a new unstained breast MHSI dataset and formulate slice-level diagnosis as a multiple instance learning (MIL) problem. We propose a Multi-Scale Hierarchical Attention Network (MS-HAN) tailored to hyperspectral MIL. Each instance (patch) is encoded by an Inception-like multi-branch extractor that operates at a fixed spatial resolution using parallel convolution kernels to capture spectral–spatial patterns at different receptive fields. To reduce high intra-class spectral variability, we introduce a prototype-based clustering regularization that softly assigns instance embeddings to learnable centers and refines the representation. We then apply dual attention directly on the spatial feature map: channel (spectral) attention generates band-wise weights from global spatial descriptors, explicitly modeling inter-band dependencies, followed by spatial attention producing a 2D attention map to localize informative cellular regions. These modules are trained end-to-end with only slice-level labels. Finally, a hierarchical aggregator models inter-patch dependencies via self-attention and performs attention pooling to obtain the slice representation for classification. On a strictly patient-split cohort of 60 patients, MS-HAN achieved 86.7% accuracy and 0.92 AUC, outperforming strong MIL baselines (e.g., TransMIL and DS-MIL). McNemar’s test demonstrated statistically significant improvements over ABMIL (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(p=0.0251\)</EquationSource> </InlineEquation>) and DS-MIL (<InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(p=0.0198\)</EquationSource> </InlineEquation>), with marginal significance against CLAM and TransMIL (<InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(p&lt;0.1\)</EquationSource> </InlineEquation>). Ablations verified the necessity of the prototype regularization and hyperspectral-specific attention. Attention visualizations highlighted regions consistent with tumor-related morphology and emphasized informative spectral ranges without pixel-level annotations, pending expert validation. MS-HAN suggests that hyperspectral-specific feature refinement and hierarchical MIL aggregation may improve robust, stain-free breast cancer detection from microscopic MHSI. Further multi-center validation and expert review of attention explanations are needed to establish clinical utility.</p>

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

MultiScale hierarchical attention network for stain free breast cancer detection in microscopic hyperspectral imaging

  • Zhuowei Chen,
  • Qingyu Yang,
  • Geng Qin,
  • Xiaoying Ma,
  • Zhuo Lu,
  • Haiyan Li,
  • Binghua Su

摘要

Microscopic hyperspectral imaging (MHSI) of unstained tissue provides quantitative, label-free cues for pathology, but practical diagnosis is hindered by weak morphological contrast and high-dimensional spectra. Patch-wise classification is therefore unstable: discriminative spectral signatures are subtle, spatially sparse, and easily confounded by noise and tissue heterogeneity. To address this, we construct a new unstained breast MHSI dataset and formulate slice-level diagnosis as a multiple instance learning (MIL) problem. We propose a Multi-Scale Hierarchical Attention Network (MS-HAN) tailored to hyperspectral MIL. Each instance (patch) is encoded by an Inception-like multi-branch extractor that operates at a fixed spatial resolution using parallel convolution kernels to capture spectral–spatial patterns at different receptive fields. To reduce high intra-class spectral variability, we introduce a prototype-based clustering regularization that softly assigns instance embeddings to learnable centers and refines the representation. We then apply dual attention directly on the spatial feature map: channel (spectral) attention generates band-wise weights from global spatial descriptors, explicitly modeling inter-band dependencies, followed by spatial attention producing a 2D attention map to localize informative cellular regions. These modules are trained end-to-end with only slice-level labels. Finally, a hierarchical aggregator models inter-patch dependencies via self-attention and performs attention pooling to obtain the slice representation for classification. On a strictly patient-split cohort of 60 patients, MS-HAN achieved 86.7% accuracy and 0.92 AUC, outperforming strong MIL baselines (e.g., TransMIL and DS-MIL). McNemar’s test demonstrated statistically significant improvements over ABMIL ( \(p=0.0251\) ) and DS-MIL ( \(p=0.0198\) ), with marginal significance against CLAM and TransMIL ( \(p<0.1\) ). Ablations verified the necessity of the prototype regularization and hyperspectral-specific attention. Attention visualizations highlighted regions consistent with tumor-related morphology and emphasized informative spectral ranges without pixel-level annotations, pending expert validation. MS-HAN suggests that hyperspectral-specific feature refinement and hierarchical MIL aggregation may improve robust, stain-free breast cancer detection from microscopic MHSI. Further multi-center validation and expert review of attention explanations are needed to establish clinical utility.