Tissue slice classification method based on microscopic hyperspectral imaging technology

Abstract

The invention provides a tissue slice classification method based on a microscopic hyperspectral imaging technology. The method comprises the following steps of firstly, preprocessing the microscopichyperspectral data so as to eliminate the noise influence and the data redundancy; establishing and training three types of convolutional neural network (CNN) models; wherein the one-dimensional CNN model realizes the spectral dimension feature extraction and classification, and the two-dimensional CNN model and the three-dimensional CNN model realize the spectrum-spectral dimension joint featureextraction and classification respectively; and for an actual to-be-detected microscopic hyperspectral image, obtaining a final classification result through the quantitative qualitative analysis andthe model output result voting. According to the present invention, the deep learning convolutional neural network models are used for extracting and classifying the deep features, so that the overallclassification precision and the speed are improved, and meanwhile the automatic data collection and classification process of the pathological sections is perfected.

Description

technical field [0001] The invention belongs to the technical field of medical image signal processing, and in particular relates to a method for classifying tissue slices based on microscopic hyperspectral imaging technology. Background technique [0002] Medical hyperspectral imaging is a comprehensive cross-cutting technology based on clinical medicine, imaging, medical sensors, pathological tissue analysis and other technologies. It belongs to the new field application of hyperspectral technology in recent years. [0003] The microscopic hyperspectral technology integrated with the microscopic imaging system can obtain microscopic hyperspectral images with higher spatial resolution, making it possible to study microscopic objects (such as tissue slices, cells, microorganisms, etc.). Microscopic hyperspectral imaging technology combines two traditional optical diagnostic technologies of spectral analysis and optical imaging, and can simultaneously provide information on t...

AI Technical Summary

Problems solved by technology

The results do not meet the requirements of precision medicine for accurate positioning of diseases, especially in the face of more complex disease subtypes and branches, the image features are not obvious enough, and the difference in spectral features is small, which is difficult to extract by traditional algorithms and simple discriminant models Deep features for efficient discriminative classification

Images