Abstract:Thyroid cancer is the most common endocrine malignancy, and accurately distinguishing between benign and malignant thyroid tumors is crucial for developing effective treatment plans in clinical practice. Pathologically, thyroid tumors pose diagnostic challenges due to improper specimen sampling. In this study, we have designed a three-stage model using representation learning to integrate pixel-level and slice-level annotations for distinguishing thyroid tumors. This structure includes a pathology structure recognition method to predict structures related to thyroid tumors, an encoder-decoder network to extract pixel-level annotation information by learning the feature representations of image blocks, and an attention-based learning mechanism for the final classification task. This mechanism learns the importance of different image blocks in a pathological region, globally considering the information from each block. In the third stage, all information from the image blocks in a region is aggregated using attention mechanisms, followed by classification to determine the category of the region. Experimental results demonstrate that our proposed method can predict microscopic structures more accurately. After color-coding, the method achieves results on unstained pathology slides that approximate the quality of Hematoxylin and eosin staining, reducing the need for stained pathology slides. Furthermore, by leveraging the concept of indirect measurement and extracting polarized features from structures correlated with lesions, the proposed method can also classify samples where membrane structures cannot be obtained through sampling, providing a potential objective and highly accurate indirect diagnostic technique for thyroid tumors.

What problem does this paper attempt to address?

This paper attempts to solve the problem of benign and malignant classification of thyroid tumors, especially the diagnostic challenges in pathological diagnosis due to improper sample collection. Specifically, the article proposes a three - stage model to more accurately distinguish between benign and malignant thyroid tumors by integrating pixel - level and slice - level annotation information through representation learning. ### Problem Background Thyroid cancer is the most common endocrine malignant tumor. Accurately distinguishing between benign and malignant thyroid tumors is crucial for formulating effective treatment plans. However, in pathology, the diagnosis of thyroid tumors faces challenges such as improper sample collection. Therefore, developing a method that can improve the diagnostic accuracy is of great significance. ### The Method Proposed in the Paper To address the above problems, this paper designs a three - stage model: 1. **Pathological Structure Recognition**: Use the pathological structure recognition method by predicting structures related to thyroid tumors (such as cell, fiber, and colloid structures). 2. **Encoder - Decoder Network**: Extract pixel - level annotation information by learning the feature representations of image patches. This stage aims to connect annotation information at different levels and generate probability maps for structure classification. 3. **Attention - mechanism - based Learning**: In the third stage, use the attention mechanism to aggregate all the image patch information within the region for the final classification task. This mechanism learns the importance of different image patches in the entire pathological region and globally considers the information of each patch. ### Innovation Points - **Application of Polarization Modes**: Utilize polarization imaging technology to extract polarization features from lesion - related structures, providing an indirect measurement method. - **Virtual Staining**: Through color coding, this method can achieve an effect similar to hematoxylin and eosin (H&E) staining on unstained pathological slices, reducing the need for stained pathological slices. - **Multi - scale Information Fusion**: Combine pixel - level and slice - level annotations to improve the accuracy of classification. ### Experimental Results The experimental results show that the proposed model is more accurate in predicting microstructures. In addition, by introducing noise and verifying the robustness of the model, the effectiveness and stability of this method are proved. ### Summary This paper proposes a new thyroid tumor classification method by introducing polarization modes and representation learning, which not only improves the diagnostic accuracy but also provides a more objective and efficient diagnostic tool for the clinic.

From Pixel to Slide image: Polarization Modality-based Pathological Diagnosis Using Representation Learning

Space and Level Cooperation Framework for Pathological Cancer Grading

A Polarization-Imaging-Based Machine Learning Framework for Quantitative Pathological Diagnosis of Cervical Precancerous Lesions

Thyroid Cancer Malignancy Prediction From Whole Slide Cytopathology Images

Interactive thyroid whole slide image diagnostic system using deep representation

Deep Learning-Based Recognition of Different Thyroid Cancer Categories Using Whole Frozen-Slide Images

Dual Polarization Modality Fusion Network for Assisting Pathological Diagnosis

Mapping microstructural features of pathological tissues by pixel clustering of Mueller matrix images

Unsupervised Learning of Pixel Clustering in Mueller Matrix Images for Mapping Microstructural Features in Pathological Tissues

Diagnostic Text-guided Representation Learning in Hierarchical Classification for Pathological Whole Slide Image

Annotation-Free Deep Learning-Based Prediction of Thyroid Molecular Cancer Biomarker BRAF (V600E) from Cytological Slides

TshFNA-Examiner: A Nuclei Segmentation and Cancer Assessment Framework for Thyroid Cytology Image

Pyramid multi-loss vision transformer for thyroid cancer classification using cytological smear

Thyroid carcinoma detection on whole histologic slides using hyperspectral imaging and deep learning

An Ensemble Learning Method for Detection of Head and Neck Squamous Cell Carcinoma Using Polarized Hyperspectral Microscopic Imaging

Cytopathology image analysis method based on high-resolution medical representation learning in medical decision-making system

A multi-resolution model for histopathology image classification and localization with multiple instance learning

PathUp: Patch-wise Timestep Tracking for Multi-class Large Pathology Image Synthesising Diffusion Model

Computer-aided Detection of Squamous Carcinoma of the Cervix in Whole Slide Images

Deep learning-based cell segmentation for rapid optical cytopathology of thyroid cancer

DeepTree: Pathological Image Classification Through Imitating Tree-Like Strategies of Pathologists