CRISPR-based high-throughput lung cancer gene screening system and method

Abstract

The invention discloses a high-throughput lung cancer gene screening system and method based on CRISPR (clustered regularly interspaced short palindromic repeats), and relates to the technical field of gene screening. According to the method, full-lung three-dimensional visualization and automatic classification and identification of the lung tissue tumor can be realized without tissue slicing, and the mass center of the tumor, the structural characteristics and the interaction relationship with the host tissue are extracted in combination with a space coordinate system. On the basis, the system constructs a multi-parameter risk evaluation index system, proposes a tumor comprehensive evaluation score S based on the shortest distance d between the tumor and the lung lobe surface in combination with the tumor volume V and the ratio R of tumor cells to host cells, and introduces a dual threshold judgment mechanism to perform risk grading on different tumor types. Compared with traditional two-dimensional slice morphology or sequencing enrichment scoring, the quantitative evaluation method can accurately identify structure invasion period and in-situ stable tumors.

Description

Technical Field

[0001] The present invention relates to the field of gene screening technology, and specifically to a CRISPR-based high-throughput lung cancer gene screening system and method. Background Art

[0002] With the rapid development of gene editing, bioimaging, and artificial intelligence, biomedical research is moving from macroscopic pathological analysis to molecular mechanism analysis and precise classification. In line with this trend, functional gene screening technology based on CRISPR has become an important tool for tumor mechanism research, showing great potential in high-throughput mutation screening in lung adenocarcinoma mouse models. However, gene editing technology alone is still difficult to systematically evaluate the spatiotemporal impact of mutant genes on tumor progression. Therefore, there is an urgent need for a new platform that integrates gene editing, tissue transparency, high-resolution 3D imaging, and AI image analysis for multidimensional data collection and intelligent analysis of lung cancer development and progression, thereby improving screening throughput and evaluation accuracy.

[0003] While current CRISPR-based mouse lung cancer research methods, such as the Tuba-seq system and two-dimensional pathology sectioning combined with sequencing, can achieve parallel screening of tumor-driving genes to a certain extent, they still face a series of key technical bottlenecks. First, Tuba-seq relies on barcode sequencing, which cannot provide information on the three-dimensional spatial distribution and true morphological structure of the tumor, thus affecting the understanding of the tumor microenvironment and invasion pathways. Second, traditional 2D tissue sectioning methods can only capture the local planar morphology of the tumor, and the limited number of sections leads to significant deviations in tumor burden assessment. Third, tumor heterogeneity, spatial invasion behavior, and microstructural features are difficult to quantify and integrate in these methods, severely limiting the precise correlation between gene function and lesion morphology. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a CRISPR-based high-throughput lung cancer gene screening system and method, which solves the problems mentioned in the background technology.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: comprising the following steps:

[0006] S1. Using the triple-gene engineered mouse model KCZ, tumor suppressor genes were knocked out through CRISPR-Cas9 technology and lentivirus-mediated gene editing. At the same time, the mice were dissected at set time points to remove lung tissue samples.

[0007] S2. Transparenting the lung tissue sample using TESOS clearing technology, imaging the transparentized lung tissue sample using a self-assembled light sheet microscope to obtain three-dimensional tumor image data, and then extracting and classifying features from the three-dimensional tumor image data using a deep learning algorithm to obtain tumor classification results and tumor data;

[0008] S3. Based on the tumor data, calculate and output the tumor volume V, and based on the tumor volume V, calculate and output the ratio R of tumor cells to host cells, and set a ratio threshold Rth and the ratio R of tumor cells to host cells for preliminary comparative evaluation;

[0009] S4. Based on the preliminary comparative evaluation results, triggering a centroid analysis, wherein the centroid analysis calculates the tumor centroid and compares it with the centroid of the non-critical area to calculate the shortest distance d between the tumor and the lung lobe surface;

[0010] S5. Based on the shortest distance d between the tumor and the surface of the lung lobe, combined with the tumor volume V and the ratio of tumor cells to host cells R, a comprehensive calculation is performed to output the tumor comprehensive assessment score S. At the same time, a comprehensive assessment threshold Sth is set to perform a secondary comparative assessment with the tumor comprehensive assessment score S to determine the tumor risk situation.

[0011] Preferably, said S1 includes S11 and S12;

[0012] S11, selecting three gene elements to construct the three-gene engineered mouse model KCZ, and pre-inserting the LSL fragment into the three-gene engineered mouse model KCZ to control the three gene elements;

[0013] The three genetic elements include the oncogene KrasG12D, the RNA-guided nuclease Cas9, and the green fluorescent protein ZsGreen;

[0014] The oncogene KrasG12D is used to drive lung adenoma mutation;

[0015] The RNA-guided nuclease Cas9 is used to knock out the target gene;

[0016] The green fluorescent protein ZsGreen is used to mark the cell area where editing occurs;

[0017] The front section of the three-gene element is provided with a LoxP-STOP-loxP element that stops the expression of the gene in the back section. When the Cre recombinase is exogenously introduced, it recognizes and removes the LSL fragment, activates the three-gene element, drives lung cancer, performs gene editing and visual marking;

[0018] S12. Instilling the lentivirus containing the gRNA into the triple-gene engineered mouse model KCZ via endotracheal intubation to induce gene editing, and then using CRISPR-Cas9 gene knockout technology to knock out the tumor suppressor gene in the triple-gene engineered mouse model KCZ. Dissecting the mice at two time points, 10 weeks and 16 weeks, to remove lung tissue samples.

[0019] The lung tissue samples include early stage samples and late stage samples;

[0020] The early samples were collected by autopsy at 10 weeks after gene knockout to capture the early morphology of lung adenoma;

[0021] The late samples were obtained by complete removal of lung tissue and fixation at 16 weeks after gene knockout for assessment of lung adenoma burden, heterogeneity, and progression.

[0022] Preferably, said S2 includes S21 and S22;

[0023] S21. After extracting the lung tissue sample, transparentize the lung tissue sample using the TESOS clearing technology, then image the transparentized lung tissue sample using a light sheet microscope to obtain three-dimensional tumor image data, construct a lung cancer gene screening server, transmit the three-dimensional tumor image data to the lung cancer gene screening server, segment the three-dimensional tumor image data using an image segmentation algorithm to obtain a tumor segmentation image set, and calibrate the tumor region in the tumor segmentation image set to obtain a calibration function Q of the tumor region, where the calibration function Q=1 represents a tumor region, and the calibration function Q=0 represents a non-tumor region;

[0024] S22. Using a convolutional neural network (CNN) to construct a deep learning model, and inputting a large number of labeled lung tissue tumor sample data sets into the deep learning model to train the deep learning model. Then, the segmented tumor segmentation image set is input into the trained deep learning model. The deep learning model extracts and classifies features based on the morphological characteristics of the tumor, the morphological characteristics of the cells, the cell density, and the coordinate position, and outputs tumor classification results and tumor data to identify the tumor condition of the current lung tissue sample.

[0025] The tumor classification results include adenocarcinoma AC, adenoma AD, atypical glandular hyperplasia AAH and hyperplasia H;

[0026] The tumor data includes three-dimensional spatial coordinates (x, y, z), the area of the tumor cell region A and the area of the host cell region B;

[0027] Among them, the three-dimensional space coordinates (x, y, z) are x, y, and z, respectively, representing the horizontal axis coordinate, and vertical axis coordinate, respectively.

[0028] Preferably, said S3 includes S31, S32 and S33;

[0029] S31 . The calibration function Q of the extracted tumor area outputs a three-dimensional spatial coordinate (x, y, z) of 1, and integrates the three-dimensional volume of the tumor to obtain the tumor volume V.

[0030] Preferably, S32, extract the host cell area and analyze the spatial relationship with the tumor center to obtain the distance term D, and use the numerical normalization method MinMax to eliminate the dimensional influence of the distance term D, tumor volume V, tumor cell area A and host cell area B, and combine the distance term D and tumor volume V that eliminate the two influences with the area A of the tumor cell area and the area B of the host cell area in the tumor data for comprehensive calculation to obtain the ratio R of tumor cells to host cells, and measure the biological proportion of tumor cells in the tissue.

[0031] Preferably, S33, based on the tumor classification result, labeling the risk level, wherein the risk level is labeled as follows;

[0032] Adenocarcinoma AC is marked as level one, and the output result is 1;

[0033] Adenoma AD is marked as level 2, and the output result is 0;

[0034] Atypical gonadal hyperplasia (AAH) is marked as level one, and the output result is 1;

[0035] Overproliferation H is marked as level 2, and the output result is 0;

[0036] The ratio R of tumor cells to host cells of each tumor classification result is extracted, and the ratio R of tumor cells to host cells is used as a continuous predictor variable, the risk level is used as the classification label, and the middle value between the first and second levels is taken as the ratio threshold Rth;

[0037] The obtained ratio R of tumor cells to host cells is then compared with the ratio threshold Rth for preliminary evaluation to analyze the risk of the ratio of tumor cell areas relative to host cell areas. Based on the preliminary comparative evaluation results, a trigger centroid analysis is performed. The specific evaluation contents are as follows;

[0038] When the ratio of tumor cells to host cells R ≥ the ratio threshold Rth, it indicates that the ratio of tumor cells is abnormal, and the centroid analysis is triggered.

[0039] When the ratio of tumor cells to host cells R is less than the threshold Rth, it means that the host cell area is dominant and has not yet expanded over a large area. At this time, monitoring should continue.

[0040] Preferably, said S4 includes S41 and S42;

[0041] S41, by using the centroid coordinates (x tumor ,y tumor , z tumor ), and at the same time, based on the remaining lung lobe image set after segmentation of the tumor segmentation image set, the centroid coordinates (x lung ,y lung , z lung );

[0042] S42. Based on the obtained centroid coordinates of the tumor area and the centroid coordinates of the lung lobe surface, calculate the shortest distance d between the tumor and the lung lobe surface, analyze whether the tumor is located in the surface area of the lung tissue, and determine the possible invasiveness.

[0043] Preferably, said S5 includes S51 and S52;

[0044] S51. Extract the shortest distance d between the tumor and the lung lobe surface, combine it with the tumor volume V and the ratio of tumor cells to host cells R, perform dimensionless processing, and then perform comprehensive weighted calculation to obtain the comprehensive tumor assessment score S.

[0045] Preferably, S52, based on the lung tissue sample data already in the database, extract each tumor comprehensive assessment score S, and average them to set as the comprehensive assessment threshold Sth, then perform a secondary comparative evaluation of the comprehensive assessment threshold Sth with the obtained tumor comprehensive assessment score S to determine the tumor risk. The specific evaluation content is as follows;

[0046] When the tumor comprehensive assessment score S> the comprehensive assessment threshold Sth, it means that the current tumor is in the structural invasion stage;

[0047] When the tumor comprehensive assessment score S ≤ the comprehensive assessment threshold Sth, it indicates that the current tumor is in the in situ stage.

[0048] A CRISPR-based high-throughput lung cancer gene screening system and a CRISPR-based high-throughput lung cancer gene screening method, characterized by comprising a gene knockout module, a feature extraction and classification module, a gene screening and evaluation module, a centroid analysis module, and a comprehensive analysis module;

[0049] The gene knockout module uses the triple-gene engineered mouse model KCZ and uses CRISPR-Cas9 technology and lentivirus-mediated gene editing to knock out tumor suppressor genes. At the same time, time points are set to prompt the mice to be dissected and lung tissue samples to be taken;

[0050] The feature extraction and classification module uses TESOS clearing technology to clear lung tissue samples, uses a self-assembled light sheet microscope to image the cleared lung tissue samples to obtain three-dimensional tumor image data, and then uses a deep learning algorithm to extract features and classify the three-dimensional tumor image data to obtain tumor classification results and tumor data;

[0051] The gene screening and evaluation module calculates and outputs the tumor volume V based on the tumor data, and calculates and outputs the ratio R of tumor cells to host cells based on the tumor volume V, and sets a ratio threshold Rth and the ratio R of tumor cells to host cells for preliminary comparative evaluation;

[0052] The centroid analysis module triggers centroid analysis based on the preliminary comparison and evaluation results. The centroid analysis calculates the tumor centroid and compares it with the centroid of the non-critical area to calculate the shortest distance d between the tumor and the lung lobe surface.

[0053] The comprehensive analysis module performs a comprehensive calculation based on the shortest distance d between the tumor and the lung lobe surface, combined with the tumor volume V and the ratio R of tumor cells to host cells, and outputs a comprehensive tumor assessment score S. At the same time, a comprehensive assessment threshold Sth is set to perform a secondary comparative assessment with the comprehensive tumor assessment score S to determine the tumor risk situation.

[0054] The present invention provides a CRISPR-based high-throughput lung cancer gene screening system and method. It has the following beneficial effects:

[0055] (1) This method combines CRISPR-Cas9 gene knockout technology in a triple-gene engineered mouse model to achieve parallel high-throughput functional verification of multiple tumor suppressor genes, significantly improving the efficiency of screening for lung adenoma driver genes. Compared with the traditional transgenic mouse method that relies on single-gene construction, this scheme not only greatly shortens the modeling cycle, but also enables structural analysis of multi-gene mutations, effectively reducing the scale of experiments and resource consumption. By setting two time points for sample collection, early and late, it is possible to dynamically capture the occurrence, development and heterogeneous evolution of lung adenocarcinoma, and promote in-depth research on the early mutation spectrum, spatial behavior mechanism and tissue adaptive evolution of lung cancer.

[0056] (2) This method integrates TESOS tissue transparency, light sheet microscopy three-dimensional imaging technology and deep learning algorithms. The present invention can achieve full lung three-dimensional visualization and automatic classification and identification of lung tissue tumors without the need for tissue sections, and extract the tumor center of mass, structural characteristics and interactive relationship with host tissue in combination with the spatial coordinate system. On this basis, a multi-parameter risk assessment index system is systematically constructed, and a comprehensive tumor assessment score S is proposed based on the shortest distance d between the tumor and the lung lobe surface combined with the tumor volume V and the ratio of tumor cells to host cells R. A dual threshold judgment mechanism is introduced to perform risk classification for different tumor types. Compared with traditional two-dimensional section morphology or sequencing enrichment scores, this quantitative assessment method can accurately identify structural invasive stages and in situ stable tumors, thereby improving the scientificity and robustness of risk prediction.

[0057] (3) This method can not only be used for functional screening of lung cancer mutation drivers and phenotypes in basic research, but also provide repeatable, quantifiable, and scalable data support for preclinical drug target validation and efficacy evaluation. Through automated classification and scoring, the spatial-structural-functional triple modeling of tumors can provide clear decision-making path guidance for targeted intervention. When the tumor comprehensive assessment score S is higher than the preset comprehensive assessment threshold Sth, it is automatically prompted to enter the intervention channel; those with lower scores are recommended to continue observation and dynamic monitoring, thus constructing an intelligent treatment recommendation framework guided by structural behavior, which has broad translational medical value and clinical promotion prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 Schematic diagram of the steps of a CRISPR-based high-throughput lung cancer gene screening method of the present invention;

[0059] Figure 2 This is a schematic diagram of the process of a CRISPR-based high-throughput lung cancer gene screening system of the present invention;

[0060] Figure 3 This is a schematic diagram of the three-dimensional spatial coordinate positioning of lung tissue samples;

[0061] Figure 4 This is the three-dimensional tumor image data of lung tissue, where green represents the tumor, red represents the lung tissue, and larger tumors are marked with yellow arrows. DETAILED DESCRIPTION

[0062] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0063] Example 1

[0064] See also Figure 1 、 Figure 3 and Figure 4 The present invention provides a high-throughput lung cancer gene screening method based on CRISPR. To achieve the above purpose, the present invention is implemented by the following technical solution: comprising the following steps:

[0065] S1. Using the triple-gene engineered mouse model KCZ, tumor suppressor genes were knocked out through CRISPR-Cas9 technology and lentivirus-mediated gene editing. At the same time, the mice were dissected at set time points to remove lung tissue samples.

[0066] S2. Transparent lung tissue samples using TESOS clearing technology. Images of the transparent lung tissue samples are taken using a self-assembled light-sheet microscope to obtain three-dimensional tumor image data. Deep learning algorithms are then used to extract and classify features from the three-dimensional tumor image data to obtain tumor classification results and tumor data.

[0067] S3. Based on the tumor data, calculate and output the tumor volume V, and based on the tumor volume V, calculate and output the ratio R of tumor cells to host cells, and set a ratio threshold Rth and the ratio R of tumor cells to host cells for preliminary comparative evaluation;

[0068] S4. Based on the preliminary comparative assessment results, trigger centroid analysis. The centroid analysis calculates the tumor centroid and compares it with the centroid of the non-critical area to calculate the shortest distance d between the tumor and the lung lobe surface.

[0069] S5. Based on the shortest distance d between the tumor and the lung lobe surface, combined with the tumor volume V and the ratio of tumor cells to host cells R, a comprehensive calculation is performed to output the tumor comprehensive assessment score S. At the same time, a comprehensive assessment threshold Sth is set to perform a secondary comparative assessment with the tumor comprehensive assessment score S to determine the tumor risk situation.

[0070] In this embodiment, the method first constructs a three-gene engineered mouse model KCZ and introduces oncogenes KrasG12D, Cas9, and ZsGreen expression elements into it. Under lentivirus-mediated gRNA introduction, the site-directed knockout of multiple tumor suppressor genes in the lung tissue is induced, effectively simulating the mutation background of lung adenocarcinoma and the natural occurrence process of the tumor. Subsequently, TESOS tissue clearing treatment and light sheet microscopy technology are used to achieve full-organ three-dimensional visualization scanning of lung tissue, and image features are extracted, automatic segmentation and classification are completed through a deep learning model to accurately identify the tumor type and extract its structural data. Based on the spatial data output by image segmentation and calibration, the system further calculates the tumor volume V, and combines the host area area with the spatial interaction relationship to calculate the ratio R of tumor cells to host cells, and realizes a preliminary risk assessment by presetting the threshold RthR_{th}Rth. If R exceeds the threshold, the centroid analysis module is further triggered to calculate the shortest distance d between the tumor centroid and the lung lobe surface to assess whether it has the possibility of extravasation and invasion. Ultimately, the system outputs a quantitative risk score S based on the tumor volume V, cell ratio R, and spatial distance d. This score is then combined with the database-generated comprehensive assessment threshold Sth for a secondary comparative assessment, enabling quantitative identification and hierarchical management of tumor bioburden, structural invasiveness, and spatial evolutionary behavior. Compared to existing technologies, this approach significantly improves the throughput and efficiency of gene screening, shortening the traditional transgenic model construction cycle. Secondly, it enables precise quantification of tumor spatial behavior and morphology based on 3D imaging and AI recognition, resolving the issue of traditional 2D slicing and sequencing methods being unable to obtain spatial distribution and structural information. Thirdly, the constructed multi-parameter fusion model enhances the ability to comprehensively judge tumor development stages and risk levels, providing a structured, visual, and data-based decision-making support platform for tumor mechanism research and preclinical target screening, with significant scientific research and promotion value and clinical translation potential.

[0071] Example 2

[0072] See also Figure 3 and Figure 4 Specifically: S1 includes S11 and S12;

[0073] S11, selecting three gene elements to construct a three-gene engineered mouse model KCZ, and pre-inserting an LSL fragment into the three-gene engineered mouse model KCZ to control the three gene elements;

[0074] The three genetic elements include the oncogene KrasG12D, the RNA-guided nuclease Cas9, and the green fluorescent protein ZsGreen;

[0075] Oncogene KrasG12D was used to drive lung adenoma mutations to simulate the real lung adenocarcinoma background;

[0076] RNA-guided nuclease Cas9 is used to perform knockout of target genes;

[0077] Green fluorescent protein ZsGreen is used to mark the cell area where editing occurs, allowing real-time visualization of the infected area;

[0078] The front section of the three-gene element is equipped with a LoxP-STOP-loxP element to stop the expression of the gene in the back section. When the Cre recombinase is exogenously introduced, it recognizes and removes the LSL fragment, activating the three-gene element to drive lung cancer, perform gene editing and visual labeling;

[0079] S12. Lentivirus containing GRNA was instilled into the triple-gene engineered mouse model KCZ through endotracheal intubation to induce gene editing. Then, combined with CRISPR-Cas9 gene knockout technology, tumor suppressor genes in the triple-gene engineered mouse model KCZ were knocked out. At two time points, 10 and 16 weeks, the mice were dissected and lung tissue samples were taken.

[0080] Lung tissue samples included early and late stage samples;

[0081] Early samples were collected by autopsy at 10 weeks after gene knockout to capture the initial morphology of lung adenoma;

[0082] Late-stage samples were obtained by dissecting intact lung tissue and fixing it at 16 weeks after gene knockout to assess lung adenoma burden, heterogeneity, and progression.

[0083] In the present embodiment, the method selects the three-gene engineered mouse model KCZ with genetic control elements as a research vector. The model pre-introduces the three major elements of oncogene KrasG12D, Cas9 enzyme required for the CRISPR-Cas9 system, and ZsGreen green fluorescent protein, and sets the three gene elements to a silent state by inserting the LSL fragment (loxP-STOP-loxP). The recognition and excision of the LSL fragment are achieved by exogenous injection of Cre recombinase, thereby synchronously activating the Kras oncogenic mutation to drive lung adenoma formation and start the Cas9-mediated gene editing mechanism. At the same time, the cell area where the editing occurs is marked by ZsGreen, realizing the trinity startup mechanism of lung cancer occurrence, functional knockout and visual tracking. The lentivirus carrying the gRNA sequence is accurately infused into the lung tissue of KCZ mice by tracheal intubation technology to achieve CRISPR-mediated knockout of specific tumor suppressor genes in vivo, thereby simulating the natural formation process of lung adenoma under different mutation conditions. Two anatomical time points of 10 weeks and 16 weeks are set to collect early and late lung tissue samples respectively. The 10-week samples are used to capture the morphology and spatial behavior of the tumor in its initial stage, and the 16-week samples are used to observe changes in tumor burden, heterogeneous evolution, and invasive tendencies, ensuring that the study has both time span and developmental integrity. Through the above-mentioned implementation mode, the present invention realizes the functional integration of simultaneous induction of carcinogenic mutations, target gene knockout, and visual tracking in the same animal model, significantly improving the controllability and efficiency of lung cancer pathogenesis research. Compared with the traditional single transgenic construction method, the experimental cycle is greatly shortened, the amount of mice used is reduced, and the screening throughput is improved. In addition, the introduction of the ZsGreen signal makes the subsequent tissue separation, three-dimensional imaging, and AI recognition processes more efficient and accurate, providing a stable, clear, and high-throughput data foundation for subsequent tumor feature analysis and comprehensive risk modeling, fundamentally improving the ability to analyze the multi-gene mutation mechanism of lung cancer and screen intervention targets.

[0084] Example 3

[0085] See also Figure 3 and Figure 4 Specifically: S2 includes S21 and S22;

[0086] S21. After extracting the lung tissue sample, use the TESOS clearing technology to clear the lung tissue sample, then image the cleared lung tissue sample using a light sheet microscope to obtain three-dimensional tumor image data, build a lung cancer gene screening server, transmit the three-dimensional tumor image data to the lung cancer gene screening server, segment the three-dimensional tumor image data using an image segmentation algorithm to obtain a tumor segmentation image set, and calibrate the tumor region in the tumor segmentation image set to obtain a calibration function Q of the tumor region, where the calibration function Q=1 represents a tumor region, and the calibration function Q=0 represents a non-tumor region;

[0087] S22. Using a convolutional neural network (CNN) to construct a deep learning model, and inputting a large number of labeled lung tissue tumor sample data sets into the deep learning model to train the deep learning model. Then, the segmented tumor segmentation image set is input into the trained deep learning model. The deep learning model extracts and classifies features based on the morphological characteristics of the tumor, the morphological characteristics of the cells, the cell density, and the coordinate position, and outputs the tumor classification results and tumor data to identify the tumor condition of the current lung tissue sample.

[0088] Tumor classification results include adenocarcinoma AC, adenoma AD, atypical glandular hyperplasia AAH, and hyperproliferation H;

[0089] Tumor data include three-dimensional spatial coordinates (x, y, z), tumor cell area A, and host cell area B;

[0090] Among them, the three-dimensional space coordinates (x, y, z) are x, y, and z, respectively, representing the horizontal axis coordinate, and vertical axis coordinate, respectively.

[0091] In this embodiment, the method first performs TESOS transparency treatment on lung tissue samples obtained from gene knockout mice, so that the lung tissue maintains its intact morphology while having high light transmittance. Subsequently, a self-assembled light sheet microscope device is used to perform a full lung scan on the transparent lung tissue to obtain three-dimensional tumor image data with tissue in situ retention characteristics. The obtained data is transmitted to the lung cancer gene screening server in real time, and the tumor area is automatically identified using an image segmentation algorithm, and a tumor segmentation image set is constructed. At the same time, by generating a tumor area calibration function Q(x, y, z), the tumor area is marked as 1 and the non-tumor area is marked as 0, providing digital support for subsequent spatial integration and volume calculation. A deep learning model with a convolutional neural network CNN structure is further constructed, and a large number of labeled three-dimensional lung tissue image samples are input for training, so that the model has the ability to efficiently recognize tumor morphology, cell morphological characteristics, cell density and spatial coordinate position. After the model training is completed, the tumor segmentation image set is input into the model, and the tumor classification results including adenocarcinoma AC, adenoma AD, atypical glandular hyperplasia AAH and hyperproliferation H are output, and key tumor data are simultaneously extracted, such as three-dimensional spatial coordinates (x, y, z), tumor cell area A, host cell area B, etc., to provide a full parameter basis for subsequent volume calculation, proportion assessment and risk analysis. Through this embodiment, the present invention realizes the full process automation processing from sample preparation, three-dimensional scanning, image structure segmentation to deep learning recognition, which not only avoids the problems of tissue damage and information loss caused by traditional 2D slicing, but also greatly improves the accuracy and consistency of tumor spatial recognition. In particular, with the support of the CNN model, the system can accurately classify early tumor types with subtle morphological differences that are difficult to identify with the naked eye, and realize individualized spatial modeling analysis based on coordinates and morphological features.

[0092] Example 4

[0093] See also Figure 1 Specifically: S3 includes S31, S32 and S33;

[0094] S31, using the calibration function Q of the extracted tumor area to output the three-dimensional spatial coordinates (x, y, z) of 1, and integrating the three-dimensional volume of the tumor to obtain the tumor volume V;

[0095] The tumor volume V is calculated and output by the following algorithm formula;

[0096] V=∫∫∫ Q＝1 Q(x, y, z) dx dy dz;

[0097] Wherein, Q=1 means that only the calibration function Q of the tumor area outputs the three-dimensional spatial coordinate (x, y, z) of 1, and integration is performed, d represents the small integral variable, dx represents the small integral variable on the horizontal axis, dy represents the small integral variable on the vertical axis, dz represents the small integral variable on the vertical axis, t is integrated, and Q(x, y, z) represents the three-dimensional spatial coordinate (x, y, z) of the calibration function Q of the tumor area.

[0098] S32. Extract the host cell region and analyze its spatial relationship with the tumor center to obtain a distance term D. Use the numerical normalization method MinMax to eliminate the dimensional effects of the distance term D, tumor volume V, tumor cell region area A, and host cell region area B. Combine the distance term D and tumor volume V, which have eliminated these two effects, with the tumor cell region area A and host cell region area B in the tumor data, to perform a comprehensive calculation to obtain a ratio R of tumor cells to host cells, which is used to measure the biological proportion of tumor cells in the tissue.

[0099] The ratio R of tumor cells to host cells is calculated and output by the following algorithm formula;

[0100]

[0101] Where A i represents the area of the ith tumor cell region, B j represents the area of the jth host cell region, Dj represents the distance term from the jth host cell region to the tumor center, n represents the total number of tumor regions, m represents the total number of host regions, and E represents the distance adjustment factor to prevent division by 0, and to a very small constant, such as 10 -5 ;

[0102] The area A of the i-th tumor cell region i Through image segmentation extraction, the tumor volume is normalized. i / V, which eliminates the scale bias caused by differences in tumor size, making it more suitable for comparison across samples;

[0103] Considering the spatial relationship between the host cell region and the tumor, the distance term D from the jth host cell region to the tumor center is introduced. j , through B j 2 / (D j +E) emphasizes that the "interaction intensity" of host cells close to the tumor area is higher, and host cells with a large area and close distance have a greater impact on the proportion.

[0104] S33. Based on the tumor classification results, the risk level is marked, wherein the risk level is marked as follows;

[0105] Adenocarcinoma AC is marked as level one, and the output result is 1;

[0106] Adenoma AD is marked as level 2, and the output result is 0;

[0107] Atypical gonadal hyperplasia (AAH) is marked as level one, and the output result is 1;

[0108] Overproliferation H is marked as level 2, and the output result is 0;

[0109] The ratio of tumor cells to host cells R was extracted for each tumor classification result, and the ratio of tumor cells to host cells R was used as a continuous predictor variable, the risk level was used as the classification label, and the middle value between level 1 and level 2 was taken as the ratio threshold Rth (as shown in Table 1);

[0110] The obtained ratio of tumor cells to host cells R is then compared with the ratio threshold Rth for preliminary evaluation to analyze the risk of the tumor cell area relative to the host cell area. Based on the preliminary comparative evaluation results, a trigger centroid analysis is performed. The specific evaluation contents are as follows;

[0111] When the ratio of tumor cells to host cells R ≥ the proportion threshold Rth, it means that the proportion of tumor cells in the area is abnormal and the host tissue participation is low, which may be an invasive lesion. At this time, the centroid analysis is triggered;

[0112] When the ratio of tumor cells to host cells R is less than the threshold Rth, it indicates that the host cell area is dominant and has not yet expanded to a large area. At this time, monitoring should continue;

[0113] Table 1, Correspondence table of tumor sample risk levels;

[0114] Sample No. Tumor type Risk Level The ratio of tumor cells to host cells R M01 AC Level 1 (1) 0.72 M02 AD Level 2(0) 0.33 M03 AAH Level 1 (1) 0.65 M04 H Level 2(0) 0.14

[0115] For example, in Table 1, high risk is greater than 0.6 and low risk is less than 0.4. In this case, the proportion threshold Rth takes the middle value of 0.5.

[0116] In this embodiment, the image processing and structural quantitative algorithms are organically integrated to construct a quantifiable, classifiable, and triggerable lung cancer risk assessment mechanism. First, the calibrated tumor area is integrated in three dimensions, and the tumor volume V is calculated using the microelement accumulation method of voxel data, thereby achieving accurate quantification of the real spatial load of the tumor. This calculation is performed on the basis of image segmentation and calibration, and has the advantages of high resolution and high restoration, laying the foundation for subsequent multi-parameter normalization processing. The area data of the tumor cell area and the host cell area are extracted, and the dimensional offset between different samples is eliminated by normalization. The distance term Dj from each host cell area to the tumor center is further combined to construct a comprehensive biological ratio index R between tumor cells and host cells. This ratio not only takes into account the area difference, but also adjusts the distance term D jEmphasize the "biological significance" of spatial interaction, so as to more realistically reflect the dominance and invasion trend of tumors in tissue structures. The normalization method also effectively avoids the evaluation bias caused by the difference in absolute volume of tumors, making the data between different samples comparable. Based on the tumor type classification results output by the aforementioned image recognition model, the system automatically assigns it a risk level label and extracts the proportion value R of the corresponding sample. By establishing a mapping between risk level and proportion, the system automatically generates a demarcation threshold Rth, such as 0.5, for preliminary comparative evaluation of new samples. If R≥Rth of a new sample, it is judged to be dominated by tumor cells and may have stronger invasiveness, and the subsequent centroid analysis process is automatically triggered; otherwise, it enters a continuous monitoring state. Through this embodiment, the present invention not only realizes a complete closed-loop process from "image calibration → numerical extraction → spatial relationship modeling → risk level assignment", but also constructs a risk quantification framework with adaptive threshold update, spatial structure trade-off and multi-dimensional data fusion. Compared with traditional methods that only consider "volume size" or "type classification", this invention further introduces the relationship between host tissue participation and spatial distribution, enabling early signs of tumor invasion to be identified and timely warnings to be issued, greatly improving the accuracy, systematicness and early response capabilities of tumor risk assessment, and providing strong criteria support for subsequent intervention recommendations.

[0117] Example 5

[0118] See also Figure 1 Specifically: S4 includes S41 and S42;

[0119] S41, by using the centroid coordinates (x tumor ,y tumor , z tumor ), and at the same time, based on the remaining lung lobe image set after segmentation of the tumor segmentation image set, the centroid coordinates (x lung ,y lung , z lung );

[0120] S42. Calculate the shortest distance d between the tumor and the lung lobe surface based on the obtained coordinates of the tumor region's centroid and the lung lobe surface's centroid, analyze whether the tumor is located on the lung tissue surface, and determine its potential invasiveness.

[0121] The shortest distance d between the tumor and the lung lobe surface is calculated and output by the following algorithm formula;

[0122] d=min[(x tumor ,y tumor , z tumor )-(x lung ,y lung , zlung )];

[0123] In the formula, min means taking the minimum value, x tumor Represents the horizontal axis coordinate of the center of mass of the tumor area, y tumor Represents the vertical coordinate of the center of mass of the tumor area, z tumor Represents the vertical axis coordinate of the center of mass of the tumor area, x lung Represents the horizontal axis coordinate of the center of mass of the lung lobe area, y lung Represents the vertical coordinate of the center of mass of the lung lobe area, z lung Represents the vertical axis coordinate of the centroid of the lung lobe area.

[0124] In this embodiment, the method establishes a geometric relationship between the spatial location of the tumor and the anatomical structure of the lung, further identifying whether the tumor has the potential risk of invasion by adhering to the lung surface or extravasating. Specifically, the method includes two sub-steps: S41 and S42.

[0125] First, based on the segmented 3D tumor image data, the spatial centroid coordinates (x tumor ,y tumor , z tumor ), and extract the lung lobe image data from the remaining tissue after tumor segmentation, and identify the centroid coordinates (x lung ,y lung , z lung ). During this processing, the system semantically partitions the image based on the tumor calibration function and tissue label information, and extracts the surface tissue as an independent region, effectively distinguishing the internal structure of the lung from the edge area. Based on the above-mentioned centroid point set, the Euclidean distance formula is used to calculate the shortest spatial distance d between the centroid of each tumor and all lung lobe surface points. This shortest distance value d can be further used for structural risk assessment: if the value is significantly smaller, it indicates that the tumor may grow close to the surface of the lung and has the potential risk of invading the pleura, blood vessels or bronchial areas; otherwise, it is relatively deep or isolated growth. Through the implementation of this step, the present invention not only achieves accurate quantification of the spatial position of the tumor, but also establishes an algorithm system for judging tumor behavior patterns based on centroid geometric distance, which makes up for the defect that traditional 2D slices or linear measurements cannot obtain real spatial relationships.

[0126] Example 6

[0127] See also Figure 1 Specifically: S5 includes S51 and S52;

[0128] S51. Extract the shortest distance d between the tumor and the lung lobe surface, combine it with the tumor volume V and the ratio of tumor cells to host cells R, perform dimensionless processing, and then perform comprehensive weighted calculation to obtain a comprehensive tumor assessment score S;

[0129] The comprehensive tumor assessment score S is calculated and output by the following algorithm formula;

[0130]

[0131] Where a1, a2, and a3 represent the preset weight values of the tumor volume V, the ratio of tumor cells to host cells R, and the shortest distance d between the tumor and the lung lobe surface, respectively, and a1+a2+a3=1. The specific values are set by the user.

[0132] S52. Based on the lung tissue sample data already in the database, extract the comprehensive assessment score S for each tumor, take the average value, and set it as the comprehensive assessment threshold Sth. Then, perform a secondary comparative evaluation of the comprehensive assessment threshold Sth with the obtained comprehensive tumor assessment score S to determine the tumor risk. The specific evaluation content is as follows;

[0133] When the tumor comprehensive assessment score S> the comprehensive assessment threshold Sth, it means that the current tumor is in the structural invasion stage;

[0134] When the tumor comprehensive assessment score S ≤ the comprehensive assessment threshold Sth, it indicates that the current tumor is in the in situ stage.

[0135] In this embodiment, in the integrated lung cancer screening and evaluation process of the method, as the final decision module, the key parameters tumor volume V, the ratio of tumor cells to host cells R, and the closest distance d of the tumor to the lung lobe surface obtained in the previous steps are integrated to quantitatively evaluate and grade the overall risk level of the tumor. First, the three core indicators are dimensionlessly processed to eliminate the unit influence of each physical quantity so that they have a unified numerical range for comparability. Subsequently, a comprehensive scoring model is constructed, and the following formula is used to calculate the comprehensive evaluation score S of the tumor: where a1, a2, and a3 are the weight coefficients of each indicator respectively. This multi-factor fusion model effectively integrates the biological load, cell proportion characteristics, and spatial invasion trend of the tumor, providing quantitative indicators for downstream judgment. Based on the accumulated lung cancer mouse model database, the comprehensive scores S of a large number of historical samples are statistically analyzed, and their mean is calculated as the demarcation point for risk judgment, that is, the comprehensive evaluation threshold Sth is set. The scoring results of new samples are then compared with this threshold for judgment: when S>Sth, the tumor is relatively large, has strong cell dominance, and is close to the lung surface, indicating that it may be in the structural invasion stage and an intervention strategy should be initiated; when S≤Sth, it is judged to be still in the in situ stage and is suitable for continued observation or low-intensity management. Through this implementation, the present invention constructs a quantifiable and standardized tumor risk assessment system, effectively integrating image anatomical parameters with biometric data, and for the first time realizing a three-dimensional risk assessment mechanism combining volume, morphology, and space.

[0136] Example 7

[0137] See also Figure 1 and Figure 2 , a CRISPR-based high-throughput lung cancer gene screening system, including a gene knockout module, a feature extraction and classification module, a gene screening and evaluation module, a centroid analysis module, and a comprehensive analysis module;

[0138] The gene knockout module uses the triple-gene engineered mouse model KCZ and uses CRISPR-Cas9 technology and lentivirus-mediated gene editing to knock out tumor suppressor genes. At the same time, time points are set to prompt the mice to be dissected and lung tissue samples to be taken;

[0139] The feature extraction and classification module uses TESOS clearing technology to clear lung tissue samples, uses a self-assembled light sheet microscope to image the cleared lung tissue samples to obtain three-dimensional tumor image data, and then uses a deep learning algorithm to extract features and classify the three-dimensional tumor image data to obtain tumor classification results and tumor data;

[0140] The gene screening and evaluation module calculates and outputs the tumor volume V based on the tumor data, and calculates and outputs the ratio R of tumor cells to host cells based on the tumor volume V, and sets a ratio threshold Rth and the ratio R of tumor cells to host cells for preliminary comparative evaluation;

[0141] The centroid analysis module triggers centroid analysis based on the preliminary comparison and evaluation results. The centroid analysis calculates the tumor centroid and compares it with the centroid of the non-critical area to calculate the shortest distance d between the tumor and the lung lobe surface.

[0142] The comprehensive analysis module performs a comprehensive calculation based on the shortest distance d between the tumor and the lung lobe surface, combined with the tumor volume V and the ratio R of tumor cells to host cells, and outputs a comprehensive tumor assessment score S. At the same time, a comprehensive assessment threshold Sth is set to perform a secondary comparative assessment with the comprehensive tumor assessment score S to determine the tumor risk situation.

[0143] While the embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that various changes, modifications, substitutions, and alterations can be made to the embodiments without departing from the principles and spirit of the invention.

Claims

1. A CRISPR-based high-throughput lung cancer gene screening method, characterized by: The following steps are involved: S1. Using the triple-gene engineered mouse model KCZ, tumor suppressor genes were knocked out through CRISPR-Cas9 technology and lentivirus-mediated gene editing. At the same time, the mice were dissected at set time points to remove lung tissue samples. S2. Transparenting the lung tissue sample using TESOS clearing technology, imaging the transparentized lung tissue sample using a self-assembled light sheet microscope to obtain three-dimensional tumor image data, and then extracting and classifying features from the three-dimensional tumor image data using a deep learning algorithm to obtain tumor classification results and tumor data; S3. Based on the tumor data, calculate and output the tumor volume V, and based on the tumor volume V, calculate and output the ratio R of tumor cells to host cells, and set a ratio threshold Rth and the ratio R of tumor cells to host cells for preliminary comparative evaluation; S4. Based on the preliminary comparative evaluation results, triggering a centroid analysis, wherein the centroid analysis calculates the tumor centroid and compares it with the centroid of the non-critical area to calculate the shortest distance d between the tumor and the lung lobe surface; S5. Based on the shortest distance d between the tumor and the surface of the lung lobe, combined with the tumor volume V and the ratio of tumor cells to host cells R, a comprehensive calculation is performed to output the tumor comprehensive assessment score S. At the same time, a comprehensive assessment threshold Sth is set to perform a secondary comparative assessment with the tumor comprehensive assessment score S to determine the tumor risk situation.

2. The CRISPR-based high-throughput lung cancer gene screening method according to claim 1, characterized in that: Said S1 includes S11 and S12; S11, selecting three gene elements to construct the three-gene engineered mouse model KCZ, and pre-inserting the LSL fragment into the three-gene engineered mouse model KCZ to control the three gene elements; The three genetic elements include the oncogene KrasG12D, the RNA-guided nuclease Cas9, and the green fluorescent protein ZsGreen; The oncogene KrasG12D is used to drive lung adenoma mutation; The RNA-guided nuclease Cas9 is used to knock out the target gene; The green fluorescent protein ZsGreen is used to mark the cell area where editing occurs; The front section of the three-gene element is provided with a LoxP-STOP-loxP element that stops the expression of the gene in the back section. When the Cre recombinase is exogenously introduced, it recognizes and removes the LSL fragment, activates the three-gene element, drives lung cancer, performs gene editing and visual marking; S12. Instilling the lentivirus containing the gRNA into the triple-gene engineered mouse model KCZ via endotracheal intubation to induce gene editing, and then using CRISPR-Cas9 gene knockout technology to knock out the tumor suppressor gene in the triple-gene engineered mouse model KCZ. Dissecting the mice at two time points, 10 weeks and 16 weeks, to remove lung tissue samples. The lung tissue samples include early stage samples and late stage samples; The early samples were collected by autopsy at 10 weeks after gene knockout to capture the early morphology of lung adenoma; The late-stage samples were obtained by complete removal of lung tissue and fixation at 16 weeks after gene knockout for assessment of lung adenoma burden, heterogeneity, and progression.

3. The CRISPR-based high-throughput lung cancer gene screening method according to claim 2, characterized in that: Said S2 includes S21 and S22; S21. After extracting the lung tissue sample, transparentize the lung tissue sample using the TESOS clearing technology, then image the transparentized lung tissue sample using a light sheet microscope to obtain three-dimensional tumor image data, construct a lung cancer gene screening server, transmit the three-dimensional tumor image data to the lung cancer gene screening server, segment the three-dimensional tumor image data using an image segmentation algorithm to obtain a tumor segmentation image set, and calibrate the tumor region in the tumor segmentation image set to obtain a calibration function Q of the tumor region, where the calibration function Q=1 represents a tumor region, and the calibration function Q=0 represents a non-tumor region; S22. Using a convolutional neural network (CNN) to construct a deep learning model, and inputting a large number of labeled lung tissue tumor sample data sets into the deep learning model to train the deep learning model. Then, the segmented tumor segmentation image set is input into the trained deep learning model. The deep learning model extracts and classifies features based on the morphological characteristics of the tumor, the morphological characteristics of the cells, the cell density, and the coordinate position, and outputs tumor classification results and tumor data to identify the tumor condition of the current lung tissue sample. The tumor classification results include adenocarcinoma AC, adenoma AD, atypical glandular hyperplasia AAH and hyperplasia H; The tumor data includes three-dimensional spatial coordinates (x, y, z), the area of the tumor cell region A and the area of the host cell region B; Among them, the three-dimensional space coordinates (x, y, z) are x, y, and z, respectively, representing the horizontal axis coordinate, and vertical axis coordinate, respectively.

4. The CRISPR-based high-throughput lung cancer gene screening method according to claim 3, characterized in that: Said S3 includes S31, S32 and S33; S31 . The calibration function Q of the extracted tumor area outputs a three-dimensional spatial coordinate (x, y, z) of 1, and integrates the three-dimensional volume of the tumor to obtain the tumor volume V.

5. The CRISPR-based high-throughput lung cancer gene screening method according to claim 4, characterized in that: S32. Extract the host cell region and analyze its spatial relationship with the tumor center to obtain the distance term D. Use the numerical normalization method MinMax to eliminate the dimensional influence of the distance term D, tumor volume V, tumor cell region area A, and host cell region area B. The distance term D and tumor volume V that eliminate the two influences are combined with the area A of the tumor cell region and the area B of the host cell region in the tumor data for comprehensive calculation to obtain the ratio R of tumor cells to host cells, and measure the biological proportion of tumor cells in the tissue.

6. The CRISPR-based high-throughput lung cancer gene screening method according to claim 5, characterized in that: S33. Based on the tumor classification results, the risk level is marked, wherein the risk level is marked as follows; Adenocarcinoma AC is marked as level one, and the output result is 1; Adenoma AD is marked as level 2, and the output result is 0; Atypical gonadal hyperplasia (AAH) is marked as level one, and the output result is 1; Overproliferation H is marked as level 2, and the output result is 0; The ratio R of tumor cells to host cells of each tumor classification result is extracted, and the ratio R of tumor cells to host cells is used as a continuous predictor variable, the risk level is used as the classification label, and the middle value between the first and second levels is taken as the ratio threshold Rth; The obtained ratio R of tumor cells to host cells is then compared with the ratio threshold Rth for preliminary evaluation to analyze the risk of the ratio of tumor cell areas relative to host cell areas. Based on the preliminary comparative evaluation results, a trigger centroid analysis is performed. The specific evaluation contents are as follows; When the ratio of tumor cells to host cells R ≥ the ratio threshold Rth, it indicates that the ratio of tumor cells is abnormal, and the centroid analysis is triggered. When the ratio of tumor cells to host cells R is less than the threshold Rth, it means that the host cell area is dominant and has not yet expanded over a large area. At this time, monitoring should continue.

7. The CRISPR-based high-throughput lung cancer gene screening method according to claim 6, characterized in that: Said S4 includes S41 and S42; S41, by using the centroid coordinates (x tumor ,y tumor , z tumor ), and at the same time, based on the remaining lung lobe image set after segmentation of the tumor segmentation image set, the lobe surface centroid coordinates (x lung ,y lung , z lung ); S42. Based on the obtained centroid coordinates of the tumor area and the centroid coordinates of the lung lobe surface, calculate the shortest distance d between the tumor and the lung lobe surface, analyze whether the tumor is located in the surface area of the lung tissue, and determine the possible invasiveness.

8. The CRISPR-based high-throughput lung cancer gene screening method according to claim 1, characterized in that: Said S5 includes S51 and S52; S51. Extract the shortest distance d between the tumor and the lung lobe surface, combine it with the tumor volume V and the ratio of tumor cells to host cells R, perform dimensionless processing, and then perform comprehensive weighted calculation to obtain the comprehensive tumor assessment score S.

9. The CRISPR-based high-throughput lung cancer gene screening method according to claim 8, characterized in that: S52. Based on the lung tissue sample data already in the database, extract the comprehensive assessment score S for each tumor, take the average value, and set it as the comprehensive assessment threshold Sth. Then, perform a secondary comparative evaluation of the comprehensive assessment threshold Sth with the obtained comprehensive tumor assessment score S to determine the tumor risk. The specific evaluation content is as follows; When the tumor comprehensive assessment score S> the comprehensive assessment threshold Sth, it means that the current tumor is in the structural invasion stage; When the tumor comprehensive assessment score S ≤ the comprehensive assessment threshold Sth, it indicates that the current tumor is in the in situ stage.

10. A CRISPR-based high-throughput lung cancer gene screening system, applied to the CRISPR-based high-throughput lung cancer gene screening method according to any one of claims 1 to 9, characterized in that: It includes gene knockout module, feature extraction and classification module, gene screening and evaluation module, centroid analysis module and comprehensive analysis module; The gene knockout module uses the triple-gene engineered mouse model KCZ and uses CRISPR-Cas9 technology and lentivirus-mediated gene editing to knock out tumor suppressor genes. At the same time, time points are set to prompt the mice to be dissected and lung tissue samples to be taken; The feature extraction and classification module uses TESOS clearing technology to clear lung tissue samples, uses a self-assembled light sheet microscope to image the cleared lung tissue samples to obtain three-dimensional tumor image data, and then uses a deep learning algorithm to extract features and classify the three-dimensional tumor image data to obtain tumor classification results and tumor data; The gene screening and evaluation module calculates and outputs the tumor volume V based on the tumor data, and calculates and outputs the ratio R of tumor cells to host cells based on the tumor volume V, and sets a ratio threshold Rth and the ratio R of tumor cells to host cells for preliminary comparative evaluation; The centroid analysis module triggers centroid analysis based on the preliminary comparison and evaluation results. The centroid analysis calculates the tumor centroid and compares it with the centroid of the non-critical area to calculate the shortest distance d between the tumor and the lung lobe surface. The comprehensive analysis module performs a comprehensive calculation based on the shortest distance d between the tumor and the lung lobe surface, combined with the tumor volume V and the ratio R of tumor cells to host cells, and outputs a comprehensive tumor assessment score S. At the same time, a comprehensive assessment threshold Sth is set to perform a secondary comparative assessment with the comprehensive tumor assessment score S to determine the tumor risk situation.

Images