A hyperspectral microscopic imaging analysis system for label-free tumor tissue

Abstract

The application discloses a hyperspectral microscopic imaging analysis system for label-free tumor tissue, comprising a super surface slide, a hyperspectral imaging system and a medical optical analysis algorithm system connected in sequence; the super surface slide comprises a glass substrate, a super surface microstructure and a biocompatible coating arranged in sequence from bottom to top, and a tissue slice is placed on the super surface microstructure. The application aims at the problems of high cost, low sensitivity, difficulty in early diagnosis and excessive dependence on doctors' experience of the current traditional cancer diagnosis technology, realizes the differentiation and identification of tumor tissue in a label-free manner in the visible light range, develops a spectral imaging system with fast, dynamic and high spectral resolution together with an image sensor, extracts spectral information in the form of image position to assist doctors in pathological diagnosis, and reduces the possibility of missed diagnosis or misdiagnosis.

Description

Technical Field

[0001] This invention relates to a hyperspectral microscopic imaging analysis system for label-free tumor tissue, belonging to the field of hyperspectral imaging technology. Background Technology

[0002] In the process of pathological diagnosis, pathological analysis of biopsies is the "gold standard" for cancer diagnosis. Conventional pathological analysis methods, such as histopathological staining and morphological assessment of suspicious tissues, are challenging in identifying abnormal cells that resemble healthy cells in morphology, and the diagnostic results depend to some extent on the subjectivity of the pathologist. Using molecular markers to differentiate disease states is complex and time-consuming, and the lack of markers for certain pathologies severely limits the detection range. A series of label-free tissue imaging techniques have been developed, including photoacoustic and ultraviolet imaging, quantitative mass spectrometry and biochemical analysis imaging, stimulated Raman scattering spectroscopy, and quantitative phase imaging, but these techniques each have their advantages and limitations in terms of analysis time, cost, ease of use, and compatibility. Currently, hyperspectral imaging technology is also gradually being applied in the biomedical field. Hyperspectral imaging technology combines spectra and images to form a three-dimensional data cube containing two-dimensional spatial image data and one-dimensional spectral data. It can not only acquire spatial information of objects but also obtain spectral information of each pixel in the image to analyze the internal properties of the object. Hyperspectral imaging technology has the characteristics of accuracy, speed, high sensitivity, non-contact, non-destructive, and high safety, meeting the development requirements of modern medical testing technology. However, most research remains at the experimental level because rapidly and accurately acquiring and processing spatially resolved spectra from millions of image pixels is a major challenge. Furthermore, in studies using hyperspectral imaging as an auxiliary analytical technique, label-free detection has not yet been truly achieved because most work still relies on staining and labeling for detection.

[0003] Traditional cancer technologies currently suffer from problems such as high cost, low sensitivity, difficulty in early diagnosis, and over-reliance on physician experience. Summary of the Invention

[0004] This invention provides a hyperspectral microscopic imaging analysis system for label-free tumor tissue, which performs image-based analysis.

[0005] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows:

[0006] A hyperspectral microscopic imaging analysis system for label-free tumor tissue includes a metasurface glass slide, a hyperspectral imaging system, and a medical optical analysis algorithm system connected in sequence.

[0007] The metasurface slide comprises a glass substrate, a metasurface microstructure, and a biocompatible coating arranged sequentially from bottom to top, with tissue sections placed on the metasurface microstructure.

[0008] The hyperspectral imaging system includes a microscope objective, a beam splitter, a telescopic lens and a CCD camera arranged from bottom to top, wherein the beam splitter is connected to the AOTF through an illumination module;

[0009] An objective lens, a polarizer, and a half-wave plate are sequentially arranged between the AOTF and the reflector. The femtosecond laser is reflected by the reflector after passing through the optical isolator.

[0010] The medical optics analysis algorithm system includes multispectral image acquisition, band selection, image preprocessing, self-learning function, and image analysis.

[0011] Preferably, the femtosecond laser beam is generated by the high nonlinear fiber in the optical isolator module through nonlinear effects including self-phase modulation, Raman scattering, optical soliton fission, and four-wave mixing, which causes the laser to broaden and produce the output beam.

[0012] Preferably, the illumination module consists of an achromatic lens, an aperture stop, and a field stop, and provides uniform and bright illumination light within the microscope's field of view.

[0013] Preferably, the metasurface glass slide is based on Fano resonance, guided mode resonance, surface lattice resonance, and bound state spectral modulation mechanism in the continuous domain.

[0014] Preferably, the sleeve lens, in conjunction with a CCD camera, acquires microscopic images of the sample and transmits them to a medical optical analysis algorithm system.

[0015] Preferably, the constituent material of the metasurface glass slide is a low-loss dielectric material.

[0016] Compared with existing technologies, this invention addresses the problems of high cost, low sensitivity, difficulty in early diagnosis, and over-reliance on physician experience in current traditional cancer diagnosis technologies. It aims to achieve label-free differentiation and identification of tumor tissue in the visible light range by developing a fast, dynamic, high-spectral-resolution spectral imaging system that, together with an image sensor, extracts spectral information from image locations for analysis and judgment. Specifically, based on various spectral modulation mechanisms, a high-spectral-resolution metasurface slide is created using low-loss media materials and structures, enabling label-free sensing of minute changes in tissue composition. An integrated, collaboratively controlled system for field-of-view stitching, automatic slide changing, and data acquisition is designed and developed using programming software to achieve rapid identification of cancer tissue slice images. A deep convolutional neural network model is independently built and trained, and a new deep learning network framework is established. A streamlined software interface is developed to realize functions such as hyperspectral image acquisition, band selection, spectral image preprocessing, self-learning, and image analysis. Ultimately, this system enables lesion location analysis and supports high-throughput batch processing analysis of thousands of images. Attached Figure Description

[0017] Figure 1 This invention relates to a fully automated, large-field-of-view hyperspectral microscopy optical path system for label-free tumor tissue hyperspectral microscopy imaging analysis.

[0018] Figure 2 This is a three-dimensional schematic diagram of a spectrally modulated metasurface glass slide for the hyperspectral microscopic imaging analysis system for label-free tumor tissues according to the present invention.

[0019] Figure 3 This invention provides in-situ analysis software for a hyperspectral microscopic imaging analysis system for label-free tumor tissue, in conjunction with a microscopic imaging analysis system.

[0020] Figure 4 This is a schematic diagram illustrating the function of the present invention in using near-field resonance effect to distinguish subtle refractive index differences between tumor tissue and normal tissue in a label-free manner. Detailed Implementation

[0021] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0022] like Figures 1 to 3 As shown, a hyperspectral microscopic imaging analysis system for label-free tumor tissue includes a metasurface glass slide, a hyperspectral imaging system, and a medical optical analysis algorithm system connected in sequence.

[0023] The metasurface slide consists of a glass substrate, a metasurface microstructure, and a biocompatible coating arranged sequentially from bottom to top, with tissue sections placed on the metasurface microstructure.

[0024] The hyperspectral imaging system includes, from bottom to top, a microscope objective, a beam splitter, a telescope lens, and a CCD camera, wherein the beam splitter is connected to the AOTF via an illumination module.

[0025] An objective lens, a polarizer, and a half-wave plate are sequentially arranged between the AOTF and the reflector. The femtosecond laser is reflected by the reflector after passing through the optical isolator.

[0026] The medical optics analysis algorithm system includes multispectral image acquisition, band selection, image preprocessing, self-learning function, and image analysis.

[0027] The femtosecond laser, generated by a highly nonlinear fiber in an optical isolator module, undergoes laser broadening through nonlinear effects including self-phase modulation, Raman scattering, optical soliton fission, and four-wave mixing. The resulting output beam possesses not only high power and high spatial coherence but also a broad spectral range from visible to near-infrared. In a fully automated large-field-of-view hyperspectral microscopy system, a supercontinuum laser, in conjunction with an AOTF (Optical Optical Transformer Filter), performs multi-wavelength channel filtering on the supercontinuum white light. The incident light then passes through an illumination module and a microscope objective before reaching a spectrally modulated metasurface slide. The illumination module, composed of an achromatic lens, an aperture stop, and a field stop, provides uniform and bright illumination within the microscope's field of view.

[0028] like Figure 3 As shown, the metasurface slide utilizes near-field resonance to distinguish subtle refractive index differences between tumor and normal tissues in a label-free manner, which is highly advantageous for early cancer detection. A sleeve lens, in conjunction with a CCD camera, acquires microscopic images of the sample and transmits them to the medical optical analysis algorithm system. Furthermore, the spectrally modulated metasurface slide is mounted on an XY motorized displacement platform; the bidirectional movement of the platform allows for field-of-view stitching. After a sample test is completed, an automatic slide-changing mechanism automatically replaces the sample, improving detection efficiency. The medical optical analysis algorithm system includes image processing and precise prediction algorithm software that utilizes machine learning algorithms to learn and classify samples, enabling accurate prediction of cancer cells.

[0029] High-spectral-resolution metasurface glass slides utilize spectral modulation mechanisms such as Fano resonance, guided-mode resonance, surface lattice resonance, and bound states in the continuous domain. These slides achieve high spectral resolution (down to 1 nm), high reflectivity (over 90%), and an effective sensing area (approximately 4 cm²) by employing novel dielectric materials, structures, and superatoms. 2 Innovative design of metasurface glass slides.

[0030] The hyperspectral imaging system technology for metasurfaces is based on a metasurface glass slide. AOTF is introduced as the core component of the hyperspectral imaging system, which can obtain extremely high resolution (less than 1 nm) single-wavelength light and highly collimated incident illumination light (less than 10°) in microseconds or even shorter times. This allows for the development of a fast, dynamic, and high-resolution spectral imaging system that can be combined with an image sensor.

[0031] The integrated control technology enables rapid identification of cancer tissue slice images. A system integrating and coordinating the control of field stitching, automatic slide changing, and data acquisition processes was designed and developed using programming software. This system achieves high field stitching accuracy (approximately 0.3μm), high identification accuracy (greater than 98%), and a large-capacity slide supply (greater than 100 slides).

[0032] To achieve functions such as hyperspectral image acquisition, band selection, spectral image preprocessing, self-learning, and image analysis, the proposed hyperspectral image data processing and recognition algorithm for tumor tissue requires the independent construction and training of a deep convolutional neural network model and the establishment of a new deep learning network framework. A streamlined software interface is also developed to ultimately achieve lesion location analysis and support high-throughput batch processing analysis of thousands of images. The proposed image processing and algorithm exhibits high accuracy (approximately 9%), high sensitivity (approximately 96%), and high specificity (approximately 98%).

[0033] The constituent materials of metasurface glass slides are not limited to a single low-loss dielectric material;

[0034] The femtosecond laser in a hyperspectral imaging system can be replaced by a picosecond laser.

[0035] The metal plasma nanoplatelet slide is replaced with a dielectric metasurface slide to achieve label-free spectral detection.

[0036] Hyperspectral imaging analysis in the visible light band can quickly and conveniently identify and distinguish tissues in different health states by using colorimetry to compare the differences in output colors.

[0037] Using AOTF to replace traditional spectroscopic devices such as mechanical disc filters, prisms, interferometers, and filters to achieve a fast, dynamically adjustable, high spectral resolution, and compact hyperspectral imaging system;

[0038] The development of a large field-of-view microscopic splicing and automatic slide feeding module to improve instrument detection efficiency and automation, capable of accommodating at least 100 slides;

[0039] The design utilizes a Kohler illumination module to provide a uniform illumination environment and collimation of the incident illumination light for the microscope field of view; the integrated collaborative control system is designed to realize the system of processes such as field stitching, automatic slide changing and automatic slide supply, and data acquisition, meeting the requirements for rapid identification and diagnosis of cancer tissue sections; the independently developed image processing algorithm and the in-situ analysis software designed in conjunction with the microscopic imaging system are designed to realize multiple functions such as multispectral image acquisition, band selection, image preprocessing, self-learning function, and image analysis.

[0040] This invention addresses the problems of high cost, low sensitivity, difficulty in early diagnosis, and over-reliance on physician experience in current traditional cancer diagnostic technologies. It aims to achieve label-free differentiation and identification of tumor tissue in the visible light range. A rapid, dynamic, and high-spectral-resolution spectral imaging system, integrated with an image sensor, is developed to assist physicians in pathological diagnosis by extracting spectral information from image locations, reducing the possibility of missed or misdiagnosed cases. Specifically, based on various spectral modulation mechanisms, a high-spectral-resolution metasurface slide is created using low-loss media materials and structures. This allows for the perception of minute changes in tissue composition in a label-free manner, which is crucial for early cancer diagnosis. An integrated, collaboratively controlled system for field-of-view stitching, automatic slide changing, and data acquisition is designed and developed using programming software to achieve rapid identification of cancer tissue slice images. A deep convolutional neural network model is independently built and trained, and a new deep learning network framework is established. A streamlined software interface is developed to realize functions such as hyperspectral image acquisition, band selection, spectral image preprocessing, self-learning, and image analysis. Ultimately, this system enables lesion location analysis and supports high-throughput batch processing analysis of thousands of images.

[0041] Finally, it should be noted that the above embodiments are only illustrative of the technical solutions of the present invention, and not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A hyperspectral microscopic imaging analysis system for label-free tumor tissue, characterized by, The application relates to a hyperspectral imaging system for pathological tissue sample analysis, which comprises a super-structured slide, a hyperspectral imaging system and a medical optical analysis algorithm system connected in sequence. The super-structured slide comprises a glass substrate, a super-structured microstructure and a biocompatible coating arranged in sequence from bottom to top, and a tissue slice is placed on the super-structured microstructure. The hyperspectral imaging system comprises a microscope objective, a beam splitter, a sleeve lens and a CCD camera arranged in sequence from bottom to top, wherein the beam splitter is connected with an AOTF through an illumination module. An objective, a polarizer and a half-wave plate are arranged in sequence between the AOTF and a mirror, and a femtosecond laser is reflected by the mirror after passing through an optical isolator. The medical optical analysis algorithm system comprises multispectral image acquisition, waveband screening, image preprocessing, self-learning function and image analysis; and the super-structured slide is based on Fano resonance, guided-mode resonance, surface-lattice resonance and bound-state spectrum regulation mechanism in continuous domain.

2. The system for label-free tumor tissue analysis by hyperspectral microscopic imaging according to claim 1, characterized in that, The femtosecond laser passes through a high-nonlinear optical fiber in a light isolator generation module, and the laser is widened through nonlinear effects such as self-phase modulation, Raman scattering, optical soliton fission and four-wave mixing to generate an outgoing light beam.

3. The system for label-free tumor tissue analysis by hyperspectral microscopic imaging according to claim 1, characterized in that, The illumination module is composed of an achromatic lens, an aperture stop and a field stop and provides uniform and bright illumination light in a microscope field of view.

4. The system for label-free tumor tissue analysis by hyperspectral microscopic imaging according to claim 1, characterized in that, The sleeve lens cooperates with the CCD camera to collect a microscopic image of the sample and transmit the image to the medical optical analysis algorithm system.

5. The system for label-free tumor tissue analysis by hyperspectral microscopic imaging according to claim 1, wherein, The composition material of the super-structured slide is a low-loss dielectric material.

Images