Circulating tumor cell bimodal co-localization model and application thereof

Abstract

The invention discloses a circulating tumor cell bimodal co-localization model and application thereof. The bimodal co-localization model comprises a morphological analysis unit and an immunofluorescence signal analysis unit for circulating tumor cells, wherein the morphological analysis unit comprises Diff staining, and the immunofluorescence signal analysis unit comprises a CD45 mouse antibody, a 488-mouse secondary antibody, a CK rabbit antibody and a 594-rabbit secondary antibody. The circulating tumor cell bimodal co-localization model has the advantages that in the application process of the circulating tumor cell bimodal co-localization model, a non-antibody-dependent nano magnetic bead enrichment technology is adopted, so that the completeness and cell viability of CTC are improved, and the success rate of subsequent multi-step dyeing is ensured; the use amount of the sample and the collection frequency are reduced; the use amount of the peripheral blood sample is reduced through double analysis (morphology and molecular markers), so that the clinical universality is improved; in combination with morphology and immunofluorescence detection, the risk of false positive and false negative is reduced, and the diagnosis accuracy of CTC is improved.

Description

Technical Field

[0001] This invention belongs to the field of biomedical detection technology and relates to a method for detecting circulating tumor cells in peripheral blood, specifically a bimodal colocalization model of circulating tumor cells and its application. Background Technology

[0002] Circulating tumor cells (CTCs) are a leading indicator of tumor metastasis, originating from the spread of primary or metastatic tumors. CTCs provide direct evidence of tumor development and progression. Currently, clinical detection of CTCs faces the following difficulties and limitations: High difficulty in detection: CTCs are present in very low concentrations in blood and are similar in morphology and properties to normal blood cells. Efficiently identifying CTCs and ensuring accuracy during enrichment and separation is a significant challenge. False negatives and false positives in CTC detection: Antibody-dependent enrichment methods, such as using surface markers like EpCAM, may miss CTCs expressing low levels or lacking this marker, leading to false negatives. Physical methods for separating CTCs (such as density gradients and microfluidics) may result in false positives due to the similarity between CTCs and other blood cells. Furthermore, the morphological characteristics of separated CTCs are highly heterogeneous. While Diff staining can provide morphological information, accurately identifying CTCs morphologically remains a challenge. Automation and Intelligentization of CTC Detection and Analysis: Traditional CTC enrichment, separation, and detection rely on manual operation, which is cumbersome and time-consuming. Utilizing automation and AI algorithms to improve detection efficiency and accuracy is the future direction. Technical Costs and Clinical Applications of CTC Detection: Existing CTC detection technologies are costly, and most CTC tests only provide information on the number of CTCs, failing to simultaneously display cell morphology and molecular information, significantly limiting their clinical application.

[0003] Current technologies often separate the observation of circulating tumor cell morphology from the detection of immunofluorescence antigens, easily leading to blind spots in information acquisition. With the advancement of precision oncology, CTCs, as "real-time liquid biopsy samples," are rapidly gaining application in early tumor screening, efficacy evaluation, and drug resistance monitoring. However, existing products (CellSearch, ISET, microfluidic chips, etc.) remain at the stage of "single staining + manual microscopy" or "pure immunofluorescence + flow cytometry," failing to meet the dual requirements of sensitivity and throughput in clinical practice. For example: ① The immunomagnetic bead-IF route relies on a single EpCAM / CK antigen; epithelial-mesenchymal transition causes the capture rate to drop to <50%, and the non-specific binding of magnetic beads to leukocytes leads to false positives; ② The recovery rate of microfluidic-IF chips decreases simultaneously with increasing purity; the signal attenuates by 30-50% after multiple rounds of fluorescence elution, resulting in significant batch-to-batch variations; ③ Manual morphological microscopy has low sensitivity, with a 38% false negative rate, a fragmented and time-consuming process, and cannot meet the demands for high throughput and real-time results. Furthermore, it is highly subjective, leading to significant interpretation differences.

[0004] In summary, traditional CTC detection has long faced three major technical challenges: (1) insufficient morphological sensitivity, resulting in a high rate of false negatives; (2) immunofluorescence is susceptible to antigen drift, non-specific binding, or human error, leading to false positives; and (3) fully manual or semi-automatic microscopic counting is inefficient and has poor repeatability. Summary of the Invention

[0005] Technical problems addressed: To overcome the shortcomings of existing technologies, this invention integrates cell morphology and immunofluorescence dual staining to compensate for the low sensitivity and false positive or false negative results of single staining, and avoids misinterpretations due to subjective factors when manually interpreting staining results, significantly improving accuracy and laying the foundation for building a platform for tumor staging, prognostic assessment, and precision medicine. Furthermore, by combining AI algorithm data models and optimizing the data reading process, it can provide rapid and reliable data references for clinical practice. Therefore, this invention provides a dual-modal co-localization model of circulating tumor cells and its applications.

[0006] Technical solution: A bimodal colocalization model for circulating tumor cells, comprising: a morphological analysis unit and an immunofluorescence signal analysis unit for circulating tumor cells; wherein, the morphological analysis unit includes Diff staining, and the immunofluorescence signal analysis unit includes: CD45 mouse antibody and 488-mouse secondary antibody, CK rabbit antibody and 594-rabbit secondary antibody.

[0007] Preferably, the model integrates an AI algorithm judgment unit, which receives morphological data from the morphological analysis unit and / or fluorescence data from the immunofluorescence signal analysis unit.

[0008] The above-described bimodal colocalization model is applied in the construction of a peripheral blood circulating tumor cell detection platform.

[0009] Preferably, the application includes the following steps:

[0010] S1. Enriching circulating tumor cells using non-antibody-dependent magnetic nanobeads.

[0011] Based on the differences in glucose metabolism between circulating tumor cells and normal cells, charge-modified magnetic nanobeads were designed to give them selective affinity for circulating tumor cells. Then, the specific binding of the magnetic nanobeads to the surface of circulating tumor cells was used to capture and enrich them. Among them, the charge-modified magnetic nanobeads are the materials disclosed in Chinese Patent ZL202210784391.5, "A Polymer-Modified Magnetic Nanomaterial, Its Preparation Method and Application".

[0012] S2 staining and scanning were used to obtain morphological data of circulating tumor cells.

[0013] Take the cell suspension obtained in S1 and drop it onto a wet cell slide. Let it air dry at room temperature. Add Diff-A solution to the slide to complete protein cross-linking and fixation. After Diff-A solution dries, add Diff-B solution to cover the cell area. Gently shake the slide and discard the Diff-B solution. Immediately add 3-5 drops of Diff-C solution to cover the cell area. Gently shake the slide and immediately wash away excess dye. Wipe away excess staining solution and water around the slide with absorbent paper. After air drying, mount the slide. Add neutral resin to the cell area and let it stand for 10-15 minutes until the resin and slide are completely adhered. Use a scanner to perform bright-field scanning to obtain cell morphology data.

[0014] The purpose of mounting the slide with neutral resin for 10-15 minutes in this step is that cell morphology scanning requires long-term focusing with an oil immersion lens or dry lens. Without mounting, water evaporation causes cells to collapse and lose focus. The resin mounting provides mechanical plane and refractive index matching, ensuring that the scanner's autofocus success rate is ≥99%.

[0015] S3, Pretreatment before immunofluorescence staining

[0016] Immerse the S2-treated slides in xylene for 3-5 minutes, remove the coverslip, replace with xylene, and continue immersion for 3-5 minutes until the neutral resin is washed away. Remove the slides and air dry. Then place the slides in destaining solution and repeat the immersion and washing twice in a shaker for 5 minutes each time. Next, place the slides in ddH2O and repeat the immersion and washing three times in a shaker for 5 minutes each time. Remove the slides and remove excess water to obtain a colorless cell layer. Place the cell layer in antigen retrieval buffer and boil it under high temperature and pressure for 2 minutes or in a boiling water bath for 10 minutes to expose the antigen epitopes. After cooling to room temperature, remove the slides and wash them three times with PBS for 5 minutes each time. After removing excess water, add 5% BSA blocking solution to the cell area and incubate at room temperature for 30-60 minutes.

[0017] The purpose of using xylene to remove the resin before decolorization in this step is that if chemical decolorization is carried out directly, the neutral resin residue will hinder the penetration of the decolorizing solution and result in uneven decolorization; the xylene step ensures 100% removal of the resin and subsequent antigen retrieval with stable pH.

[0018] The purpose of using a pressure cooker to boil at high temperature and pressure for 2 minutes or a boiling water bath for 10 minutes is that, after fixation with Diff-A and acid etching with decolorizing solution, the antigen epitopes are masked by aldehyde cross-linking; high temperature and pressure reversible cross-linking restores the binding sites of CK and CD45 antibodies, and the signal-to-noise ratio is improved by 3-5 times.

[0019] The purpose of permeabilization is to ensure that some targets (such as Ki-67 in the nucleus) can penetrate the nuclear membrane; without permeabilization, the detection rate of nucleoproteins is less than 30%. This step is controlled by an optional script to ensure high fidelity for extramembrane targets.

[0020] The purpose of blocking with 5% BSA for 30-60 minutes is that the residual charge sites after decolorization are prone to adsorb secondary antibody, which can easily increase the background fluorescence by 2-3 times.

[0021] S4. Immunofluorescence detection: Scanning to obtain protein expression results.

[0022] After the blocking incubation is complete, remove the blocking solution and add 50-100 μL of primary antibody working solution (the primary antibody working solution is made by diluting the anti-human CD45 mouse monoclonal antibody stock solution and the anti-human pan-CK rabbit monoclonal antibody with antibody dilution buffer, where the CD45 mouse antibody is diluted at a ratio of 1 / 200-1 / 500 and the pan-CK rabbit antibody is diluted at a ratio of 1 / 500-1 / 1000). Place the solution in an antibody incubation chamber, keep it moist, and incubate overnight at 4°C for 16-18 hours. After incubation, remove the incubation chamber and allow it to return to room temperature for 15-30 minutes. Then wash three times with phosphate-buffered saline (PBST) to remove excess water, and add 50-100 μL of secondary antibody working solution (the secondary antibody working solution is made by diluting the anti-human CD45 mouse monoclonal antibody stock solution and the anti-human pan-CK rabbit monoclonal antibody with antibody dilution buffer, where the CD45 mouse antibody is diluted at a ratio of 1 / 200-1 / 500 and the pan-CK rabbit antibody is diluted at a ratio of 1 / 500-1 / 1000). 488 fluorescent dye for goat anti-mouse IgG and conjugated with Alexa Goat anti-rabbit IgG with 594 fluorescent dye was diluted with antibody diluent (488 anti-mouse secondary antibody was diluted to 1 / 500-1 / 1000, and 594 anti-rabbit secondary antibody was diluted to 1 / 500-1 / 1000). The diluted antibody was then returned to the incubation chamber and incubated at room temperature in the dark for 60 minutes. The slide was then removed, washed three times with PBST to remove excess water, and 20 μL of anti-fluorescence quenching mounting medium containing DAPI was added. The slide was incubated at room temperature in the dark for 2-5 minutes. A coverslip was then placed on top, excess liquid was wiped off, and the coverslip was sealed with clear nail polish. Immunofluorescence data were then read using a fluorescence scanner.

[0023] The purpose of using transparent nail polish for sealing in this step is that: DAPI anti-quenching agent contains glycerin, which is easily volatilized within 48 hours; the nail polish forms an airtight ring to ensure that the fluorescence signal decays by less than 5% within 7 days, meeting the hospital's follow-up examination requirements.

[0024] S5. Combining the results of S2 and S4, we can provide data references for the clinical detection of circulating tumor cells in peripheral blood.

[0025] Preferably, the application method integrates an AI algorithm judgment unit, specifically: the cell morphology data obtained in S2 and the immunofluorescence data obtained in S4 are input into the AI ​​algorithm judgment unit through network or local storage, a three-dimensional coordinate system is established using DAPI nuclear positioning signals, the morphological contour and fluorescence signal are mapped to the same pixel matrix, generating "morphology-fluorescence" data, which is then input into the AI ​​training database.

[0026] Preferably, the decolorizing solution in S3 includes methanol, ethanol, denatured alcohol, or hydrochloric acid alcohol.

[0027] The principle of the bimodal colocalization model of circulating tumor cells described in this invention is as follows: A temporal scanning strategy is employed to spectrally separate the visible light absorption spectrum of DAPI morphological dyes with the narrow-band emission spectrum of Alexa Fluor series fluorescent dyes; this enhances the lateral resolution of subcellular structures such as the nucleolus, nuclear membrane, and cytoplasmic granules, meeting the discrimination requirements between CTCs and leukocytes; a post-fixation step is added after morphological staining to retain antigenic epitopes; a tumor target + CD45 dual-color combination is selected, using a "positive-negative-exclusion" logic to lock epithelial-derived tumor cells while eliminating leukocyte interference; a three-dimensional coordinate system is established using DAPI nuclear localization signals, mapping morphological contours and fluorescence signals to the same pixel matrix to generate "morphological-fluorescence" data, which is then input into the AI ​​training database.

[0028] This invention employs a dual-channel strategy of "first co-localization, then AI fusion" to encode complementary information from two modalities into computable high-dimensional features in a single step, resulting in more comprehensive identification, more accurate results, and significantly improved efficiency. Its core technology is the use of deep learning to establish a joint "morphology-molecule" probabilistic model, transforming human experience into transferable and scalable mathematical expressions. This systematically and comprehensively addresses the three major pain points of sensitivity, specificity, and efficiency, providing a standardized, reproducible, and high-throughput new paradigm for CTC clinical testing.

[0029] Beneficial effects: (1) In the application of the bimodal colocalization model of circulating tumor cells described in this invention, the non-antibody-dependent nanomagnetic bead enrichment technology is used to improve the integrity and cell activity of CTCs and ensure the success rate of subsequent multi-step staining; reduce sample volume and collection times - the amount of peripheral blood samples is reduced by dual analysis (morphology and molecular markers), which improves clinical universality; improve detection accuracy - the combination of morphological and immunofluorescence detection reduces the risk of false positives and false negatives and improves the diagnostic accuracy of CTCs. (2) After the model is integrated with the AI ​​algorithm model, it can further improve the capture efficiency of circulating tumor cells and realize the three-layer closed loop of information complementarity, noise suppression and algorithm iteration. Attached Figure Description

[0030] Figure 1 This is a graph showing the application results of the method described in Example 1 in the screening of circulating tumor cells;

[0031] Figure 2-4 The image shows the test results of three peripheral blood samples using the method described in Example 2.

[0032] Figure 5 This is a flowchart of the AI ​​algorithm in Example 2, which has been fully trained and verified, in a practical application. Detailed Implementation

[0033] The following embodiments further illustrate the content of the present invention, but should not be construed as limiting the present invention. Modifications and substitutions made to the methods, steps, or conditions of the present invention without departing from the spirit and essence of the invention are all within the scope of the present invention. Unless otherwise specified, the technical means used in the embodiments are conventional means well known to those skilled in the art.

[0034] In this embodiment, Diff staining was performed using circulating tumor cell staining solution purchased from Zhihui Medical Devices (Zhejiang) Co., Ltd., product model RS01. Diff-A solution corresponds to D1 in the staining solution kit, Diff-B solution corresponds to D2 in the staining solution kit, and Diff-C solution corresponds to D3 in the staining solution kit.

[0035] Antigen retrieval buffer was purchased from Beijing Zhongshan Jinqiao Biotechnology Co., Ltd., product number ZLI-9065;

[0036] The antibody diluent was purchased from Suzhou Xinsaimei Biotechnology Co., Ltd., product number WB100D;

[0037] CD45 antibody was purchased from abcam, product code ab8216;

[0038] Pan-CK antibody was purchased from abcam, product code ab234297;

[0039] Goat anti-mouse IgG H&L (Alexa) 488) Purchased from abcam, product number ab150113;

[0040] Goat anti-rabbit IgG H&L (Alexa) 594) Purchased from abcam, product number ab150080.

[0041] Example 1: Dual-modal colocalization model for circulating tumor cell detection

[0042] The application includes the following steps:

[0043] S1. Enriching circulating tumor cells using non-antibody-dependent magnetic nanobeads.

[0044] Based on the differences in glucose metabolism between circulating tumor cells and normal cells, charge-modified magnetic nanobeads were designed to give them selective affinity for circulating tumor cells. Then, the specific binding of the magnetic nanobeads to the surface of circulating tumor cells was used to capture and enrich them.

[0045] S2 staining and scanning were used to obtain morphological data of circulating tumor cells.

[0046] Take the cell suspension obtained in S1 and drop it onto a wet cell slide. Let it air dry at room temperature. Add Diff-A solution to the slide to complete protein cross-linking and fixation. After Diff-A solution dries, add Diff-B solution to cover the cell area. Gently shake the slide for 20 seconds and then pour off Diff-B solution. Immediately add 3-5 drops of Diff-C solution to cover the cell area. Gently shake the slide for 10 seconds and then wash off excess dye. Wipe away excess staining solution and water around the slide with absorbent paper. After air drying, mount the slide. Add neutral resin to the cell area and let it stand for 10-15 minutes until the resin adheres completely to the slide. Then, use a bright-field scanner to obtain cell morphology data.

[0047] S3, Pretreatment before immunofluorescence staining

[0048] After soaking the S2-treated slides in xylene for 3-5 minutes, remove the coverslip and replace with xylene, continuing the soaking for another 3-5 minutes until the neutral resin is washed away. Remove the slides and air dry. Then, place the slides in destaining solution and repeat the soaking and washing process twice in a shaker for 5 minutes each time. Next, place the slides in ddH2O and repeat the soaking and washing process three times in a shaker for 5 minutes each time. After removing the slides, remove excess water to obtain a colorless cell layer. Place the cell layer in antigen retrieval buffer and boil it under high temperature and pressure for 2 minutes or in a boiling water bath for 10 minutes to expose the antigen epitopes. After cooling to room temperature, remove the slides, wash with PBS buffer, and selectively permeabilize. Permeabilize the cells with 0.1% Triton X-100 for 10 minutes, and wash with PBS three times for 5 minutes each time. After removing excess water, add 5% BSA blocking solution to the cell area and incubate at room temperature for 30-60 minutes.

[0049] S4. Immunofluorescence detection: Scanning to obtain protein expression results.

[0050] After the blocking incubation is complete, remove the blocking solution and add 50-100 μL of primary antibody working solution. Place the slide in an antibody incubation chamber, keep it moist, and incubate overnight at 4°C for 16-18 hours. After incubation, remove the incubation chamber and allow it to return to room temperature for 15-30 minutes. Then wash the slide three times with phosphate-buffered saline (PBST) to remove excess water. Add 50-100 μL of secondary antibody working solution and return the slide to the incubation chamber. Incubate at room temperature in the dark for 60 minutes. Remove the slide and wash it three times with PBST to remove excess water. Add 20 μL of anti-fluorescence quenching mounting medium containing DAPI and let it stand at room temperature in the dark for 2-5 minutes. Add a coverslip, wipe off excess liquid, and seal the coverslip with clear nail polish. Use fluorescence scanning to read the immunofluorescence data.

[0051] S5. Combining the results of S2 and S4, we can provide data references for the clinical detection of circulating tumor cells in peripheral blood.

[0052] Results analysis:

[0053] This embodiment employs a non-antibody-dependent nanomagnetic bead enrichment method, enriching CTCs through physical properties and avoiding missed detections due to antigen drift. Experiments show that the recovery rate of this method can reach over 85%, significantly higher than that of traditional immunomagnetic bead enrichment methods (recovery rate <50%). Furthermore, the AI ​​algorithm can identify the morphological characteristics of the enriched CTCs, further screening for potential CTCs and providing more accurate targets for subsequent staining steps.

[0054] This embodiment employs a dual analysis of morphology and molecular markers, requiring only one sample for detection. Diff staining rapidly reveals cell morphology and structure, initially screening for potential CTCs; subsequent immunofluorescence detection, by labeling specific tumor markers (such as CK and CD45), confirms the presence of CTCs at the molecular level. Building upon this, an AI algorithm can perform pixel-level fusion of morphological and immunofluorescence data to generate a high-dimensional feature tensor, further optimizing the detection results and ensuring high sensitivity and specificity even with low sample sizes. Simultaneously, the dual analysis reduces false positive and false negative rates.

[0055] Figure 1 After CTC enrichment, blood samples were subjected to Diff staining for pathological morphological analysis. The results showed... Figure 1 The second arrow in the second row of the second small image points to a circulating tumor cell (CTC). Subsequently, the cell was destained and stained with immunofluorescence. The nuclei were observed under a fluorescence microscope using DAPI staining (blue channel), and compared with the nuclei under Diff staining in bright field. This confirmed that the cell's fluorescence results showed CD45 negativity (green channel) and pan-CK positivity (red channel), further confirming it as a CTC. Furthermore, this CTC was closely connected to two CD45-positive leukocytes, suggesting it was a cluster structure formed by circulating tumor cells and leukocytes.

[0056] Example 2

[0057] This embodiment, based on Embodiment 1, incorporates an AI algorithm. Specifically, it employs a multi-channel fluorescence scanner (bright field + three fluorescence channels) to accurately locate CTCs by comparing DAPI cell nuclei (blue) with bright field cell nuclei signals. In other words, the cell morphology data obtained in S2 and the immunofluorescence data obtained in S4 of Embodiment 1 are input into the AI ​​algorithm's judgment unit via network or local storage. A three-dimensional coordinate system is established using the DAPI nuclear localization signal, mapping the morphological contours and fluorescence signals to the same pixel matrix to generate "morphology-fluorescence" data, which is then input into the AI ​​training database.

[0058] Results analysis:

[0059] This embodiment, by integrating AI algorithms based on Embodiment 1, ensures the accurate fusion of morphological and immunofluorescence data, providing high-quality input data for the AI ​​algorithm and is key to achieving high-precision AI interpretation. Furthermore, the AI ​​fusion interpretation technology significantly improves the automation level and result stability of the detection, reduces the subjective error of manual interpretation, and is crucial for achieving high efficiency and high repeatability.

[0060] like Figure 2-4 As shown: Figure 2-4These are panoramic images of bright-field and corresponding immunofluorescence staining obtained sequentially by a scanner from three peripheral blood samples after CTC enrichment, Diff staining, and immunofluorescence staining. CTCs were precisely located by comparing DAPI (blue) with the bright-field nuclear signals (indicated by arrows). In the immunofluorescence images, the green channel represents CD45, the red channel represents pan-CK, and the blue channel represents DAPI. After the above registration, cells meeting the following criteria were retained: 1. Morphological criteria (at least 4 out of 6 criteria must be met): ① Cell diameter (long end) ≥ 15 μm, ② Nuclear-cytoplasmic ratio > 0.8, ③ Deeply stained and irregular nucleus, ④ Thickened nuclear membrane with indentations or wrinkles, resulting in an irregular nuclear membrane shape, ⑤ Giant nucleoli or abnormal mitotic figures, ⑥ Abnormal nuclear type, such as lobed or mulberry-shaped nuclei. If ⑤ is present, and two other diagnostic criteria are met, CTCs can also be diagnosed; 2. Immunophenotype: CD45 negative, pan-CK positive. Cell morphology characteristics and immunophenotype (CD45- / pan-CK) + The complete feature vectors of CTC and its subtypes (CTC, CTM, CHC) after double verification are imported into the AI ​​training database.

[0061] Figure 5 This demonstrates the practical application of a fully trained and validated AI algorithm on real-world samples. The algorithm first receives a Diff stained brightfield panoramic image acquired by a scanner. Figure 5 (Left) Without human intervention, the system automatically performs the identification steps and ultimately identifies and labels circulating tumor cells (CTC), circulating tumor microemboli (CTM), and circulating atypical cells (CHC) simultaneously in the same field of view.

[0062] Example 3

[0063] Based on Example 1, those skilled in the art can replace the CTC enrichment method, cell morphology staining method, and fluorescence detection method according to different detection needs:

[0064] For example, replace S1 in Example 1 with microfluidic chip enrichment or density gradient centrifugation enrichment. Microfluidic chip enrichment relies on the physical size and surface markers of cells, resulting in high CTC recovery rates, but may have a high rate of missed detection for some small CTCs. Nanoparticle enrichment, on the other hand, enriches based on physical properties, has no size limitations for CTCs, and is more widely applicable. Advantages: Microfluidic chip enrichment may be faster, suitable for high-throughput screening. Density gradient centrifugation enriches CTCs based on differences in cell density; it is simple to operate, but has a lower CTC recovery rate and may contain more leukocytes, increasing the complexity of subsequent processing. Advantages: Low cost, suitable for laboratories with limited resources.

[0065] Replace S2 in Example 1 with H&E staining or Giemsa staining. H&E staining provides richer contrast between the nucleus and cytoplasm, but the antigen activity retention rate after destaining is lower, which may affect the sensitivity of subsequent immunofluorescence detection. Advantages: H&E staining is widely used in pathological diagnosis, and pathologists are more familiar with interpreting its results. Giemsa staining provides higher contrast for the nucleolus, but the antigen activity retention rate after destaining is lower, and the destaining process is more complex. Advantages: Giemsa staining is widely used in blood cell morphology analysis and is suitable for detecting CTCs in blood.

[0066] The S4 immunofluorescence assay in Example 1 is replaced with in situ hybridization or flow cytometry. In situ hybridization detects specific mRNA or DNA sequences in CTCs using fluorescently labeled probes. While in situ hybridization provides molecular-level results, it is complex and time-consuming. Its advantage lies in its high specificity for detecting gene mutations and expression in tumor cells. Flow cytometry detects CTCs using fluorescently labeled antibodies, enabling rapid analysis of large numbers of cells and making it suitable for high-throughput detection. However, it retains less cell morphology information and relies heavily on the intensity and specificity of the fluorescence signal. Its advantage is its fast detection speed, making it suitable for large-scale sample screening.

[0067] If the above alternatives are used and AI algorithms need to be integrated, the AI ​​model needs to be retrained, quantized, and burned into the system.

[0068] It should be noted that most operations in the complete data processing chain of this invention are established once and maintained as needed, such as: formula verification and stability testing of staining / decolorizing reagents; factory calibration of multi-channel optical paths; offline training, quantization and burning of AI model weight files; daily quality control slide scanning; monitoring of scanner LED / laser power attenuation; quality control sample testing when reagent batches change, etc.

Claims

1. A bimodal colocalization model of circulating tumor cells, characterized in that, The dual-modal colocalization model includes a morphological analysis unit for circulating tumor cells and an immunofluorescence signal analysis unit; wherein, the morphological analysis unit includes Diff staining, and the immunofluorescence signal analysis unit includes CD45 mouse antibody and 488-mouse secondary antibody, CK rabbit antibody and 594-rabbit secondary antibody.

2. The bimodal colocalization model of circulating tumor cells according to claim 1, characterized in that, The model integrates an AI algorithm judgment unit, which receives morphological data from the morphological analysis unit and / or fluorescence data from the immunofluorescence signal analysis unit.

3. The application of the dual-modal colocalization model as described in claim 1 in constructing a peripheral blood circulating tumor cell detection platform.

4. The application according to claim 3, characterized in that, Includes the following steps: S1. Enriching circulating tumor cells using non-antibody-dependent magnetic nanobeads. Based on the differences in glucose metabolism between circulating tumor cells and normal cells, charge-modified magnetic nanobeads were designed to give them selective affinity for circulating tumor cells. Then, the specific binding of the magnetic nanobeads to the surface of circulating tumor cells was used to capture and enrich them. S2 staining and scanning were used to obtain morphological data of circulating tumor cells. Take the cell suspension obtained in S1 and drop it onto a wet cell slide. Let it air dry at room temperature. Add Diff-A solution to the slide to complete protein cross-linking and fixation. After Diff-A solution dries, add Diff-B solution to cover the cell area. Gently shake the slide and discard the Diff-B solution. Immediately add 3-5 drops of Diff-C solution to cover the cell area. Gently shake the slide and immediately wash away excess dye. Wipe away excess staining solution and water around the slide with absorbent paper. After air drying, mount the slide. Add neutral resin to the cell area and let it stand for 10-15 minutes until the resin and slide are completely adhered. Use a scanner to perform bright-field scanning to obtain cell morphology data. S3, Pretreatment before immunofluorescence staining After soaking the S2-treated slides in xylene for 3-5 minutes, remove the coverslip and replace with xylene, continuing the soaking for another 3-5 minutes until the neutral resin is washed away. Remove the slides and air dry. Then, place the slides in destaining solution and repeat the soaking and washing process twice in a shaker for 5 minutes each time. Next, place the slides in ddH2O and repeat the soaking and washing process three times in a shaker for 5 minutes each time. After removing the slides, remove excess water to obtain a colorless cell layer. Place the cell layer in antigen retrieval buffer and boil it under high temperature and pressure for 2 minutes or in a boiling water bath for 10 minutes to expose the antigen epitopes. After cooling to room temperature, remove the slides, wash with PBS buffer, and selectively permeabilize. Permeabilize the cells with 0.1% Triton X-100 for 10 minutes, and wash with PBS three times for 5 minutes each time. After removing excess water, add 5% BSA blocking solution to the cell area and incubate at room temperature for 30-60 minutes. S4. Immunofluorescence detection: Scanning to obtain protein expression results. After the blocking incubation is complete, remove the blocking solution and add 50-100 μL of primary antibody working solution. Place the slide in an antibody incubation chamber, keep it moist, and incubate overnight at 4°C for 16-18 hours. After incubation, remove the incubation chamber and allow it to return to room temperature for 15-30 minutes. Then wash the slide three times with phosphate-buffered saline (PBST) to remove excess water. Add 50-100 μL of secondary antibody working solution and return the slide to the incubation chamber. Incubate at room temperature in the dark for 60 minutes. Remove the slide and wash it three times with PBST to remove excess water. Add 20 μL of anti-fluorescence quenching mounting medium containing DAPI and let it stand at room temperature in the dark for 2-5 minutes. Add a coverslip, wipe off excess liquid, and seal the coverslip with clear nail polish. Use fluorescence scanning to read the immunofluorescence data. S5. Combining the results of S2 and S4, we can provide data references for the clinical detection of circulating tumor cells in peripheral blood.

5. The application according to claim 4, characterized in that, The application method integrates an AI algorithm judgment unit, specifically: the cell morphology data obtained in S2 and the immunofluorescence data obtained in S4 are input into the AI ​​algorithm judgment unit through network or local storage; a three-dimensional coordinate system is established using DAPI nuclear positioning signals; the morphological contour and fluorescence signal are mapped to the same pixel matrix to generate "morphology-fluorescence" data, which is used as AI input; the solidified tensor alignment algorithm is run, and the AI ​​inference output is obtained.

6. The application according to claim 4, characterized in that, The decolorizing solution described in S3 includes methanol, ethanol, denatured alcohol, or hydrochloric acid alcohol.

Images