RESUMO
Pathological images are considered the gold standard for clinical diagnosis and cancer grading. Automatic segmentation of pathological images is a fundamental and crucial step in constructing powerful computer-aided diagnostic systems. Medical microscopic hyperspectral pathological images can provide additional spectral information, further distinguishing different chemical components of biological tissues, offering new insights for accurate segmentation of pathological images. However, hyperspectral pathological images have higher resolution and larger area, and their annotation requires more time and clinical experience. The lack of precise annotations limits the progress of research in pathological image segmentation. In this paper, we propose a novel semi-supervised segmentation method for microscopic hyperspectral pathological images based on multi-consistency learning (MCL-Net), which combines consistency regularization methods with pseudo-labeling techniques. The MCL-Net architecture employs a shared encoder and multiple independent decoders. We introduce a Soft-Hard pseudo-label generation strategy in MCL-Net to generate pseudo-labels that are closer to real labels for pathological images. Furthermore, we propose a multi-consistency learning strategy, treating pseudo-labels generated by the Soft-Hard process as real labels, by promoting consistency between predictions of different decoders, enabling the model to learn more sample features. Extensive experiments in this paper demonstrate the effectiveness of the proposed method, providing new insights for the segmentation of microscopic hyperspectral tissue pathology images.
RESUMO
Spatiotemporal heterogeneity of tumors provides an escape mechanism for breast cancer cells, which can obstruct the investigation of tumor progression. While molecular profiling obtained from mass spectrometry imaging (MSI) is rich in biochemical information, it lacks the capacity for in vivo analysis. Ultrasound diagnosis has a high diagnostic accuracy but low chemical specificity. Here, we describe a noninvasive ultrasound elastography (UE)-guided MSI strategy (UEg-MSI) that integrates physical and biochemical characteristics of tumors acquired from both in vivo and in vitro imaging. Using UEg-MSI, both elasticity histopathology metabolism "fingerprints" and reciprocal crosstalk are revealed, indicating the intact, multifocal spatiotemporal heterogeneity of spontaneous tumorigenesis of the breast from early, middle, and late stages. Our results demonstrate a gradual increase in malignant degree of primary focus in cervical and thoracic mammary glands. This progression is characterized by increased stiffness according to elasticity scores, histopathological changes from hyperplasia to increased nests of neoplastic cells and necrotic areas, and regional metabolic heterogeneity and reprogramming at the spatiotemporal level. De novo fatty acid (FA) synthesis focused on independent (such as ω-9 FAs) and dependent (such as ω-6 FAs) dietary FA intake in the core cancerous nest areas in the middle and late stages of tumor or in the peripheral microareas in the early stage of the tumor. SM-Cer signaling pathway and GPs biosynthesis and degradation, as well as glycerophosphoinositol intensity, changed in multiple characteristic microareas. The UEg-MSI strategy holds the potential to expand MSI applications and enhance ultrasound-mediated cancer diagnosis. It offers new insight into early cancer discovery and the occurrence of metastasis.