CN111398250A - Tumor diagnosis method based on molecular fragment spectrum generated by interaction of light and substance - Google Patents
Tumor diagnosis method based on molecular fragment spectrum generated by interaction of light and substance Download PDFInfo
- Publication number
- CN111398250A CN111398250A CN202010133904.7A CN202010133904A CN111398250A CN 111398250 A CN111398250 A CN 111398250A CN 202010133904 A CN202010133904 A CN 202010133904A CN 111398250 A CN111398250 A CN 111398250A
- Authority
- CN
- China
- Prior art keywords
- molecular fragment
- neural network
- diagnosis
- network model
- interaction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/62—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light
- G01N21/71—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light thermally excited
- G01N21/718—Laser microanalysis, i.e. with formation of sample plasma
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Evolutionary Biology (AREA)
- Software Systems (AREA)
- Optics & Photonics (AREA)
- Plasma & Fusion (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention relates to a tumor pathological diagnosis method based on a molecular fragment spectrum generated by interaction of light and a substance, belongs to the technical field of spectrum detection, and solves the problems that the condition of preoperative imaging diagnosis cannot be highly consistent, pathological biopsy is required to be performed in an operation, the frozen pathological biopsy depends on the experience of a pathologist, and the diagnosis accuracy is not high. And acquiring laser-induced breakdown spectroscopy data of the tumor by using a laser-induced breakdown spectroscopy measurement system. And (3) identifying the molecular fragment spectral bands in the data as diagnostic identification features, calculating the intensity ratio of the molecular fragment spectral bands, and expanding the quantity of the diagnostic identification features. And constructing a brain-like calculation pulse neural network model according to the ratio characteristics, and applying the obtained pulse neural network model to diagnose and identify unknown molecular fragment spectral data.
Description
Technical Field
The invention relates to a tumor pathological diagnosis method based on a molecular fragment spectrum generated by interaction of light and a substance, and belongs to the technical field of spectrum detection.
Background
The incidence of the malignant tumor is increased year by year at present, according to the national cancer report of China family, China cancer big data authority report, troubled by cancer published by the China national cancer center for 2019, the ratio of the malignant tumor accounts for 26% of the main causes of death of Chinese urban residents. The average relative survival rate of 5 years is only 40.5 percent and is far lower than the level of developed countries. The primary treatment for malignant tumors is still surgical resection. The postoperative survival time and quality of life of the patient are affected by the surgical resection. Because the preoperative imaging diagnosis can not be highly consistent with the intraoperative condition, pathological biopsy needs to be performed in the operation, a frozen section biopsy mode is usually adopted, although the speed of the mode is faster than that of the traditional pathological detection, the time is still 10 to 30 minutes, and the freezing causes the liquid in cells and among cells to form pricks, damages partial tissue morphological structures, influences judgment of doctors and has low pathological biopsy diagnosis accuracy. An auxiliary technical means capable of carrying out pathological diagnosis in real time and rapidly is urgently needed in clinical application occasions, and auxiliary opinions are provided for pathologists.
A L ase-induced breakdown spectroscopy (L IBS) is used as a novel substance element detection technology, and is widely researched and applied in the fields of explosive detection, clinical medicine sample detection, cultural heritage identification, alloy processing, space exploration, agriculture, food analysis and the like at present.
However, during the induced breakdown process, the molecular fragment spectra produced are generally less and therefore, it is more difficult to distinguish directly from the intensity of the spectra. Machine learning methods have been widely used in the field of spectral processing, but many conventional machine learning classification recognition models have difficulty in achieving a good effect in terms of recognition of molecular fragment spectra. Effective discrimination based on molecular fragments also urgently requires an effective spectral recognition algorithm, so that the tumor targeting technology based on the molecular fragment spectrum can be practically applied.
Disclosure of Invention
The invention aims to solve the problem that pathological biopsy in operation is difficult to rapidly and accurately carry out in clinical tumor pathological diagnosis application, and provides a molecular fragment spectrum diagnosis method based on laser-induced breakdown spectroscopy. The method generates molecular fragment spectrum by ablation and breakdown according to the interaction of laser and tissues, and uses advanced brain-like computational pulse neural network to classify and identify, thereby completing pathological diagnosis in tumor operation and solving the practical application problem of clinical pathological diagnosis of tumor.
The technical scheme of the invention is as follows:
a method for diagnosing tumors based on molecular fragment spectra generated by interaction between light and a substance, the method comprising the steps of:
1. and collecting spectral data.
And a laser-induced breakdown spectroscopy experimental system is adopted to acquire corresponding spectral data aiming at various tumors.
2. Molecular fragment bands are identified.
According to the corresponding positions of the wavelengths of the common molecular fragment bands, the molecular fragment bands and the intensities thereof in the spectral data are identified, and as diagnostic identification characteristics, for example, the common biological tissue molecular fragment bands are shown in table 1.
Table 1 examples of common biological tissue molecular fragment band wavelengths
3. And expanding the diagnosis identification characteristic quantity and calculating the molecular fragment band intensity ratio.
And (4) calculating the intensity ratio of the obtained molecular fragment bands pairwise to serve as a new diagnosis and identification characteristic.
4. And constructing a brain-like computational impulse neural network model.
1) Determining a coding mode, and coding the sample data into a pulse sequence;
2) constructing a traditional neural network model as an initial pulse neural network model, inputting a pulse sequence into a pulse neural network, and calculating to obtain an output pulse sequence;
3) comparing the expected pulse sequence with the actual output pulse sequence to obtain an error, and adjusting the weight W according to the error;
4) and when the difference between the actual output pulse sequence of the neural network and the expected output pulse sequence is minimum, fixing the weight W to obtain a pulse neural network model.
5. And (3) applying the obtained pulse neural network model to diagnose and identify unknown molecular fragment spectral data.
Advantageous effects
The pathological diagnosis in the operation of the tumor has important practical significance, the diagnosis result directly influences the subsequent surgical excision judgment of a doctor and influences the prognosis of a patient, and the current common frozen biopsy causes the liquid in cells and between cells to form pricks, damages partial tissue morphological structures and influences the judgment of the doctor. The invention provides a method for tumor pathological diagnosis based on a molecular fragment spectrum generated by interaction of light and a substance. The molecular characteristics of different tumors and different stage areas of the same tumor are different, the element content level is different, and the molecular fragments of the tumor under the action of laser are different. The laser-induced breakdown spectroscopy technology is one of analysis methods with potential in the current material identification and detection, can provide molecular fragment spectra generated by laser ablation breakdown, and can achieve better diagnosis and identification effects on visual spectral data without obvious differences by combining an adaptive preprocessing method and a pulse neural network identification method. The method obviously improves the pathological diagnosis speed of the clinical tumor, has better stability, can realize higher diagnosis accuracy, does not need to completely depend on the diagnosis of a clinical pathological doctor, saves manpower and material resources, solves the problems of high dependence on the experience of the doctor, slower speed and lower accuracy of the pathological diagnosis of the clinical tumor, and provides a foundation for the popularization of the practical application of the molecular fragment spectrum clinical biopsy method.
Drawings
FIG. 1 is a schematic flow diagram of the process of the present invention;
FIG. 2 is a schematic diagram of a laser-induced breakdown spectroscopy measurement system;
FIG. 3 is a diagram showing the diagnosis result of infiltrative glioma in brain tumor in the example.
Detailed Description
To better illustrate the objects and advantages of the present invention, the following description is provided in conjunction with the accompanying drawings and examples.
Examples
Take brain tumors of interest for neurosurgery as an example. Some brain tumors, such as glioma, are higher in brain tumor and are usually found in various parts of the brain. Glioma is invasive, can spread to other surrounding tissues, needs to be subjected to extensive excision to prevent relapse, is distinguished from other tumor diagnoses, and has important significance for judging an excision area in operation. The application of the tumor pathological diagnosis method based on molecular fragment spectrum generated by interaction of light and substances in diagnosis of glioma, brain membrane tumor, craniopharyngioma and epidermoma is illustrated by taking the diagnosis of glioma, meningioma, craniopharyngioma and epidermoma as an example.
1. And collecting spectral data.
The laser induced breakdown spectroscopy experimental device shown in fig. 2 is set up, wherein 1 is an indicating calibration laser, 2 is an exciting laser, 3 is a reflector, 4 is a focusing lens, 5 is plasma generated by induced breakdown, 6 is an optical receiver, 7 is an optical detector, 8 is a three-dimensional translation table, 9 is a time delay device, 10 is a spectrometer, and 11 is a computer.
Different tumor pathological sample blocks to be detected from different patients are respectively placed on a sample three-dimensional translation table 8, and a laser induced breakdown spectrum is collected, wherein a laser adopts Nd: YAG laser fundamental frequency 1064nm output wavelength, frequency 1Hz, single pulse energy 40 mJ. And adjusting the three-dimensional sample table to enable the acquisition point of each luminescence spectrum to be located at different positions on the sample so as to avoid ablation of the tumor sample to be almost complete and excitation of the impurity spectrum. The excitation spectrum of every 10 laser pulses is averaged as an average spectrum. All tumor diagnostic samples were from 12 patients, 4 of whom glioma, 3 of whom meningioma, 3 of whom involuntary tumor, and 2 of whom craniopharyngioma were detected, wherein the spectral detection data of tumors from 2 of whom glioma, 2 of whom meningioma, 2 of whom involuntary tumor, and 1 of whom craniopharyngioma were detected were used to train the constructed model, for a total of 355 spectral data, and the spectral detection data of tumors from 4 of whom patient were used to test the disclosed inventive method, for a total of 123 spectral data.
2. Molecular fragment bands are identified.
According to the common biological tissue molecular fragment band wavelength shown in the table 1, 6 molecular fragment band clusters are identified and obtained from the spectral data, 20 molecular fragment bands are contained, and the intensities of the peaks in each molecular fragment band cluster are summed to be used as the intensity of the molecular fragment band cluster to be used as the diagnosis identification characteristic.
3. And expanding the diagnosis identification characteristic quantity and calculating the molecular fragment band intensity ratio.
And (4) calculating the intensity ratios of the obtained molecular fragment bands pairwise to obtain 15 intensity ratios as new diagnosis and identification characteristics.
4. And constructing a brain-like computational impulse neural network model.
1) Encoding sample data into a pulse sequence by adopting an Izhikevich model neuron encoding mode;
2) constructing a traditional neural network model as an initial pulse neural network model, setting the number of hidden layers to be 1, inputting a pulse sequence into the pulse neural network, and calculating to obtain an output pulse sequence;
3) comparing the expected pulse sequence with the actual output pulse sequence to obtain an error, and optimizing and adjusting the weight W according to the error by adopting a Cuckoo Search algorithm;
4) and when the difference between the actual output pulse sequence of the neural network and the expected output pulse sequence is minimum, fixing the weight W to obtain a pulse neural network model.
5. And (3) diagnosing and identifying unknown 123 molecular fragment spectral data by using the obtained pulse neural network model, and diagnosing which belong to the infiltrative glioma and which belong to other tumors. The diagnosis result is shown in fig. 3, the diagnosis identification accuracy reaches 88.6%, the diagnosis time is within 10s, and the accuracy reaches a higher level in the existing diagnosis method.
The analysis result proves that the proposed tumor pathological diagnosis method based on the molecular fragment spectrum generated by the interaction of light and a substance can carry out pathological diagnosis on clinical tumors under the condition of quick real-time, the diagnosis result is similar to the current optimal level, a quick, real-time and accurate technical means is provided for the pathological detection of the tumors in clinical operation, and the problem of difficult objective diagnosis is solved.
Claims (4)
1. A method for diagnosing tumors based on molecular fragment spectra generated by interaction between light and a substance, comprising: the method comprises the following specific steps:
1) collecting spectral data;
2) identifying molecular fragment bands as diagnostic identification features;
3) expanding the diagnosis and identification characteristic quantity, and calculating the intensity ratio of the molecular fragment spectral bands;
4) constructing a brain-like calculation impulse neural network model;
5) and (3) applying the obtained pulse neural network model to diagnose and identify unknown molecular fragment spectral data.
2. The method of claim 1, wherein the molecular fragment spectrum generated by the interaction between light and a substance is used as the basis for tumor diagnosis, and the method comprises the following steps: the identification of the molecular fragment band described in the step 2) is taken as a diagnosis identification characteristic, and comprises the identification of a single molecular fragment band peak and a molecular fragment band cluster consisting of a plurality of molecular fragment band peaks.
3. The method of claim 1, wherein the molecular fragment spectrum generated by the interaction between light and a substance is used as the basis for tumor diagnosis, and the method comprises the following steps: the specific step of expanding the diagnosis and identification feature quantity in the step 3) is to calculate the intensity ratio of the obtained molecular fragment spectral bands pairwise to be used as a new diagnosis and identification feature.
4. The method of claim 1, wherein the molecular fragment spectrum generated by the interaction between light and a substance is used as the basis for tumor diagnosis, and the method comprises the following steps: the specific steps of constructing the brain-like computational impulse neural network model in the step 4) are as follows:
1) determining a coding mode, and coding the sample data into a pulse sequence;
2) constructing a traditional neural network model as an initial pulse neural network model, inputting a pulse sequence into a pulse neural network, and calculating to obtain an output pulse sequence;
3) sequencing the spectral data of all unknown classes from small to large according to the calculation result of the distance value for the spectral data of each known class;
4) and when the difference between the actual output pulse sequence of the neural network and the expected output pulse sequence is minimum, fixing the weight W to obtain a pulse neural network model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010133904.7A CN111398250A (en) | 2020-03-02 | 2020-03-02 | Tumor diagnosis method based on molecular fragment spectrum generated by interaction of light and substance |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010133904.7A CN111398250A (en) | 2020-03-02 | 2020-03-02 | Tumor diagnosis method based on molecular fragment spectrum generated by interaction of light and substance |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111398250A true CN111398250A (en) | 2020-07-10 |
Family
ID=71432106
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010133904.7A Pending CN111398250A (en) | 2020-03-02 | 2020-03-02 | Tumor diagnosis method based on molecular fragment spectrum generated by interaction of light and substance |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111398250A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111832477A (en) * | 2020-07-13 | 2020-10-27 | 华中科技大学 | Novel coronavirus detection method and system |
CN114366037A (en) * | 2022-02-16 | 2022-04-19 | 北京理工大学 | Spectral diagnostic system for guidance in brain cancer operation and working method thereof |
-
2020
- 2020-03-02 CN CN202010133904.7A patent/CN111398250A/en active Pending
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111832477A (en) * | 2020-07-13 | 2020-10-27 | 华中科技大学 | Novel coronavirus detection method and system |
CN114366037A (en) * | 2022-02-16 | 2022-04-19 | 北京理工大学 | Spectral diagnostic system for guidance in brain cancer operation and working method thereof |
CN114366037B (en) * | 2022-02-16 | 2023-11-24 | 北京理工大学 | Spectral diagnosis system for guiding in brain cancer operation and working method thereof |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6333326B2 (en) | Analysis method of biological samples by spectral imaging | |
US10012591B2 (en) | Measurement of tissue structures | |
US20060281068A1 (en) | Cytological methods for detecting a disease condition such as malignancy by Raman spectroscopic imaging | |
CN109297949B (en) | Tumor cell detection method and device by combining microscopic image with transmission Raman spectrum | |
US20060253261A1 (en) | Digitizing biology | |
CN114923893B (en) | Nondestructive marking-free rapid breast cancer Raman spectrum pathological grading and staging method | |
WO2007056560A2 (en) | System and method for cytological analysis by raman spectroscopic imaging | |
CN108732121B (en) | Method for identifying benign and malignant tumors based on infrared spectrum | |
CN111398250A (en) | Tumor diagnosis method based on molecular fragment spectrum generated by interaction of light and substance | |
US20050250091A1 (en) | Raman molecular imaging for detection of bladder cancer | |
AU2010247132B2 (en) | Tissue sample analysis | |
US20180024064A1 (en) | Method of detecting fibrous tissue in a biological specimen using co-localized images generated by second harmonic generation and two photon emission | |
Lin et al. | Discrimination of lung tumor and boundary tissues based on laser-induced breakdown spectroscopy and machine learning | |
Teng et al. | Pathological identification of brain tumors based on the characteristics of molecular fragments generated by laser ablation combined with a spiking neural network | |
CN117038060A (en) | Raman spectrum molecular detection and imaging device based on machine learning cascade | |
CN112716447A (en) | Oral cancer classification system based on deep learning of Raman detection spectral data | |
Pezzei et al. | Fourier transform infrared imaging analysis in discrimination studies of bladder cancer | |
CN1804593A (en) | Method for distinguishing epithelial cell cancer property by single cell Raman spectrum | |
EP2887050A1 (en) | Method for marker-free demarcation of tissues | |
WO2007011571A2 (en) | Digitizing biology | |
Liu et al. | For cervical cancer diagnosis: Tissue Raman spectroscopy and multi-level feature fusion with SENet attention mechanism | |
CN113960130A (en) | Machine learning method for diagnosing thyroid cancer by adopting open ion source | |
CN111624191A (en) | Off-body universal brain tumor biopsy and boundary determining device | |
US20240122493A1 (en) | System for detection of breast cancer margin and method thereof | |
CN114384057B (en) | Tumor early diagnosis system based on Raman spectrum |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20200710 |