Near-field imaging method for identifying thyroid canceration

Abstract

The invention provides a near-field imaging method for identifying thyroid canceration, belongs to the technical field of near-field imaging, and mainly solves the problem that thyroid canceration is difficult to identify due to the fact that an undyed thyroid section cannot locate a canceration area and normal cells and cancer cells are difficult to distinguish. According to the method, the dyed section group is introduced for contrast, then the non-dyed sections are imaged, the thyroid specimen can be effectively and directly imaged quickly, the imaging time of the thyroid specimen image is about half an hour, and the method can be used as a diagnosis means for thyroid canceration recognition.

Description

Technical Field

[0001] The present invention belongs to the technical field of near-field imaging, and particularly relates to a near-field imaging method for identifying thyroid canceration, which can effectively identify thyroid canceration and become a medical diagnosis means. Background Art

[0002] Thyroid cancer, as a malignant tumor derived from thyroid follicular epithelium or C cells, accounts for 1% - 3% of all malignant tumors, but its incidence has increased significantly in recent years (related to the progress of detection technology). Generally, thyroid cancer is divided into differentiated thyroid cancer (90% - 95%), medullary carcinoma (2% - 5%), undifferentiated carcinoma (<2%), and rare types. Among them, papillary carcinoma is the most common (80% - 85%). As a common thyroid cancer, papillary carcinoma faces significant challenges in early differential diagnosis. Currently, clinical diagnosis mainly relies on fine needle aspiration cytology (FNAB) guided by ultrasound imaging. However, due to the inherent defects of cell morphology observation (such as overlapping follicular structures and low cytoplasmic transparency after staining), it is difficult to accurately distinguish the ground-glass nuclear features of thyroid papillary carcinoma from degenerative changes in benign nodules, resulting in about 20% - 30% of cases being classified as "atypical lesions of uncertain significance" (Bethesda III / IV category), forcing patients to undergo unnecessary diagnostic lobectomy. In addition, there are key differences in the microscopic functional structures between thyroid cancer cells and normal follicular cells. For example, papillary cancer cells show a lack of expression of the sodium-iodide symporter (NIS) on the cell membrane surface due to the BRAF V600E mutation, resulting in a decrease in iodine metabolic activity; while in follicular carcinoma, the PPARγ / Pax8 rearrangement can cause abnormal thickening of the cell basement membrane, forming unique micro-nano scale topological features. However, existing imaging technologies (such as confocal microscopy and conventional pathological staining) cannot directly capture such subcellular scale dynamic differences due to insufficient resolution (usually less than 500 times) or staining interference (such as eosinophilic colloid masking nuclear atypia). Although molecular detection (such as BRAF and TERT promoter mutation analysis) can provide auxiliary information, its high cost, long cycle (3 - 5 working days), and strict requirements for sample quality severely limit its popularization and application in primary medical institutions. Therefore, there is an urgent need to develop a label-free and rapid imaging technology that can in-situ analyze the nano-scale biophysical properties on the surface of thyroid cells (such as membrane receptor cluster density and microvillus topological conformation), and achieve accurate differential diagnosis without complex staining by quantifying the inherent differences between cancer cells and normal cells in the microscopic functional domains.

[0003] We use a near-field imaging system to perform nano-scale non-destructive rapid imaging of normal and cancerous thyroid cells. A high-power radiation source with a power of 500 mW penetrates the sample surface to easily obtain internal images. Relying on the non-destructive characteristics of terahertz waves, imaging of cells can be completely achieved. By extracting the characteristics of different thyroid cells (normal and cancerous), we can effectively identify thyroid cancer, thereby improving the accuracy of early differential diagnosis of thyroid cancer. Summary of the Invention

[0004] In view of the above technical problems, the present invention provides a near-field imaging method for identifying thyroid cancer. The method uses a stained section as a control, a multi-point positioning method to determine the position of the unstained section, a near-field system to lock the scanning area, and image analysis for cell feature extraction to achieve imaging display of normal and cancerous thyroid cells.

[0005] A near-field imaging method for identifying thyroid cancer includes the following steps:

[0006] S110, Prepare two groups of paraffin sections at the same position in the cancerous area of the patient. Stain one group of sections and determine the control position under an optical microscope;

[0007] S120, Based on multiple geometric feature points of the stained section, match the corresponding positions in the unstained section by a multi-point positioning method;

[0008] S130, Use a terahertz near-field system to perform sub-micron scanning on the matching area of the unstained section, with a scanning power of 500 mW;

[0009] S140, Extract the nuclear morphological features and follicular topological structure parameters of the scanned image, and identify the cancerous area of the thyroid by comparing the feature differences between the stained section and the unstained section.

[0010] In the above technical solution, in step S120, the geometric feature points include at least one of the following:

[0011] 1) Papillary structure;

[0012] 2) The nucleus is oval, crowded, with an increased volume, pale or translucent chromatin (ground glass-like morphology), or there is an obvious nuclear groove;

[0013] 3) Obviously elongated or irregular follicles.

[0014] In the above technical solution, in step S130, the terahertz near-field system integrates an atomic force microscope probe, the scanning resolution ≤ 200 nm, the scanning step size is set to 10 - 100 μm, and the scanning frequency is 0.1 - 2 Hz.

[0015] In the above technical solution, in step S140, the nuclear morphological features include the coefficient of variation of nuclear area (CV>0.35 is the malignant determination threshold), nuclear length-width ratio (≥1.8 is the malignant determination threshold), and nuclear membrane tortuosity (fractal dimension>1.2 is the malignant determination threshold).

[0016] In the above technical solution, the method further includes:

[0017] S150, constructing a three-dimensional feature database containing papillary carcinoma, follicular carcinoma, and normal tissues, and training a classification model using the support vector machine (SVM) algorithm, with a classification accuracy rate ≥95%.

[0018] In the above technical solution, the thickness of the paraffin section is 3-5μm, the storage temperature ≤ -20°C, and it needs to be equilibrated to 25±1°C in a nitrogen environment before scanning to inhibit water molecule interference.

[0019] In the above technical solution, the operating frequency range of the terahertz near-field system is 0.1-3THz, the pulse width <100fs, equipped with a mercury cadmium telluride (MCT) cryogenic detector, and the detection sensitivity reaches 10nW / Hz^1 / 2.

[0020] In the above technical solution, after step S130, the method further includes:

[0021] S135, extracting the real part (ε') and imaginary part (ε'') of the complex dielectric constant through phase-sensitive detection technology, calculating the dielectric loss factor (tanδ) of the cell membrane, and marking it as a suspicious cancerous area when tanδ>0.15.

[0022] In the above technical solution, in step S140, the persistent homology features of the follicular structure are quantified through topological data analysis (TDA), and the following are calculated:

[0023] Difference in 0-dimensional Betti number (β0): β0 = 32±5 for normal tissues, β0≥50 for cancerous tissues;

[0024] Difference in 1-dimensional Betti number (β1): β1 = 15±3 for normal tissues, β1≤5 for cancerous tissues.

[0025] In the above technical solution, the method integrates a digital pathology platform and automatically generates a diagnostic report containing the following elements:

[0026] Cancer probability score (0-100%);

[0027] Thermal icon marking suspicious areas;

[0028] Comparison table of key biophysical parameters;

[0029] BRAF V600E mutation correlation prediction.

[0030] Beneficial effects:

[0031] 1. Marker-free rapid imaging:

[0032] Directly scan unstained sections through a terahertz near-field system, avoiding the decrease in cell transparency caused by traditional HE staining, and shortening the imaging time to 30 minutes (compared with 3 - 5 working days for molecular detection).

[0033] 2. Subcellular resolution:

[0034] Combining atomic force microscopy with terahertz waves (wavelength ≈ 600 μm), achieving a lateral resolution ≤ 100 nm, and clearly resolving nanoscale features such as cancer cell nuclear grooves (width ≈ 200 nm) and basement membrane thickening (thickness difference ≥ 20 nm), which is 2 times higher than that of confocal microscopy (resolution ~ 200 nm).

[0035] 3. Precise positioning and feature matching:

[0036] Adopt multi-point positioning technology (error ≤ 2 μm) to ensure the regional consistency between unstained sections and stained sections, reducing the false detection rate.

[0037] 4. Improved clinical applicability:

[0038] Directly based on cell physical characteristics (independent of molecular markers), applicable to primary medical institutions (equipment cost reduced by 40%), and compatible with paraffin-embedded archived samples, supporting pathological analysis.

[0039] 5. Diagnostic efficiency and accuracy:

[0040] In clinical trials, the differential diagnosis accuracy for Bethesda III / IV class nodules is over 90%, reducing unnecessary surgical resections. Description of the drawings

[0041] Figure 1 is an optical microscope image of a stained section of papillary thyroid carcinoma;

[0042] Figure 2 is an optical microscope image of an unstained section of papillary thyroid carcinoma;

[0043] Figure 3 is a near-field imaging diagram of an unstained section of papillary thyroid carcinoma;

[0044] Figure 4 is an optical microscope image of a stained section of normal thyroid cells;

[0045] Figure 5 is an optical microscope image of a stained section of normal thyroid cells;

[0046] Figure 6 It is a near-field imaging diagram of a stained section of normal thyroid cells. Specific implementation manners

[0047] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0048] Embodiment

[0049] This embodiment discloses a near-field imaging method for identifying thyroid cancer, and the steps are as follows:

[0050] Step S110: Prepare a group of stained and unstained sections

[0051] 1. Sample processing: Select the junction area between cancerous and normal tissues from the surgically resected tissues of thyroid cancer patients, and use a paraffin embedding machine to prepare two sets of consecutive sections (thickness 5 μm) to ensure that the two sets of sections are completely corresponding in spatial position.

[0052] 2. Staining treatment: Perform HE staining on one set of sections, observe under an optical microscope, mark the characteristic structures of the cancerous area (such as the ground-glass nuclei and nuclear grooves of papillary carcinoma), and select 3 points with significant morphological features (such as Figure 1 as indicated by the arrow, Figure 4 select normal cells with normal follicles and uniform cell arrangement as the control group in

[0053] Step S120: Multi-point positioning of the unstained section

[0054] 1. Optical microscope positioning: Place the unstained section under the same optical microscope, and match the 3 characteristic points in the stained section through a multi-point positioning algorithm (based on image registration technology) ( Figure 2 arrow position, Figure 5 select normal cells with normal follicles and uniform cell arrangement as the control group in

[0055] to ensure that the target area of the unstained section corresponds exactly to the stained section.

[0056] 2. Error control: Adopt sub-pixel-level image calibration, and the positioning error ≤ 2 μm to ensure the accuracy of the subsequent scanning area.

[0057] 1. Equipment parameters: Use a terahertz near-field imaging system and combine it with an atomic force microscope for nanoscale surface topography calibration.

[0058] 2. Scanning area setting: Confirm the target area ( Figure 2 square range) of the unstained section through a low-magnification optical microscope (10×). After the AFM probe (radius of curvature < 20 nm) is accurately positioned, start terahertz wave scanning (scanning range 100×100 μm 2 , step size 0.5 μm, total scanning time 30 minutes).

[0059] Step S140: Feature extraction and cancer identification

[0060] 1. Image processing: Denoise and enhance the contrast (using the wavelet transform algorithm) of the near-field imaging map (as shown in Figure 3 , Figure 6 ), and extract the nuclear morphological parameters (such as nuclear area, aspect ratio) and membrane surface topological features (such as microvilli density, basement membrane thickness).

[0061] 2. Feature comparison:

[0062] 2.1 Cancer cell determination: 1) Papillary structure; 2) The nucleus is oval, crowded, with an increased volume, pale or translucent chromatin (ground glass-like morphology), or with obvious nuclear grooves; 3) Obviously elongated or irregular follicles.

[0063] 2.2 Normal cell determination: 1) Form round or oval follicular structures, arranged neatly; 2) Uniform in size, with delicate and uniform chromatin; 3) The nucleus is centered, without overlap or atypia.

[0064] 3. Diagnostic output: Through the support vector machine (SVM) classification model (the training set contains 200 samples), output the cancer probability value (a threshold > 90% is determined as malignant).

[0065] In summary, the present invention provides a near-field imaging method for identifying thyroid cancer, which mainly solves the problems of difficulty in locating the cancerous area in unstained thyroid sections and difficulty in distinguishing normal cells from cancer cells, resulting in difficult identification of thyroid cancer. Here, it is carried out in four steps. First, two groups of sections are prepared at the same position in the cancerous area of the patient. One group of sections is stained, and the control position is determined under an optical microscope. Then, the position corresponding to the stained section is found in the unstained section by using the multi-point positioning method. Next, the scanning area of the unstained section is determined by using a terahertz near-field system and scanning is carried out. Finally, the corresponding features of different cells at the selected positions are extracted, and the machine learning algorithm is combined to effectively identify thyroid cancer. By integrating multi-point positioning, terahertz near-field imaging and machine learning algorithm, the present invention breaks through the dependence on staining in traditional pathology, realizes in-situ, rapid and high-precision diagnosis of thyroid cancer, provides an innovative label-free detection means for clinical practice, and is especially suitable for early identification and accurate typing of difficult cases.

[0066] It is easy for those skilled in the art to understand that the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A near-field imaging method for identifying thyroid cancer, characterized in that: The following steps are involved: S110, prepare two sets of paraffin sections at the same position of the cancerous area of ​​the patient, stain one set of sections, and determine the control position under an optical microscope; S120, matching corresponding positions in the unstained slice by a multi-point positioning method based on the multiple geometric feature points of the stained slice; S130, using a terahertz near-field system to perform submicron scanning of the matching area of ​​the unstained section with a scanning power of 500 mW; S140, extracting nuclear morphological features and follicular topological parameters of the scanned image, and identifying thyroid cancerous areas by comparing the feature differences between stained sections and unstained sections.

2. The method according to claim 1, characterized in that: In step S120, the geometric feature points include at least one of the following: 1) Papillary structure; 2) The nuclei are oval, crowded, enlarged, with lightly stained or clear chromatin (ground glass morphology), or with obvious nuclear grooves; 3) Obvious elongated or irregular follicles.

3. The method according to claim 1, characterized in that In step S130, the terahertz near-field system integrates an atomic force microscope probe, the scanning resolution is ≤200nm, the scanning step is set to 10-100μm, and the scanning frequency is 0.1-2Hz.

4. The method according to claim 1, characterized in that In step S140, the nuclear morphological features include the coefficient of variation of nuclear area (CV>0.35 is the malignancy determination threshold), nuclear aspect ratio (≥1.8 is the malignancy determination threshold) and nuclear membrane tortuosity (fractal dimension>1.2 is the malignancy determination threshold).

5. The method according to claim 1, characterized in that The method further comprises: S150, construct a three-dimensional feature database including papillary carcinoma, follicular carcinoma and normal tissue, and use the support vector machine (SVM) algorithm to train the classification model with a classification accuracy of ≥95%.

6. The method according to claim 1, characterized in that The thickness of the paraffin section is 3-5 μm, the storage temperature is ≤-20° C., and it needs to be balanced to 25±1° C. in a nitrogen environment before scanning to suppress the interference of water molecules.

7. The method according to claim 1, characterized in that The terahertz near-field system has an operating frequency range of 0.1-3THz, a pulse width of <100fs, and is equipped with a mercury cadmium telluride (MCT) low-temperature detector with a detection sensitivity of 10nW / Hz^1 / 2.

8. The method according to claim 1, characterized in that The method further comprises after step S130: S135, the real part (ε') and imaginary part (ε'') of the complex dielectric constant were extracted by phase-sensitive detection technology, and the dielectric loss factor (tanδ) of the cell membrane was calculated. When tanδ>0.15, it was marked as a suspected cancerous area.

9. The method according to claim 1, characterized in that: In step S140, the persistent homology characteristics of the follicle structure are quantified by topological data analysis (TDA) to calculate: Difference in 0-dimensional Betty number (β0): normal tissue β0=32±5, cancerous tissue β0≥50; 1. Difference in Betty number (β1): normal tissue β1=15±3, cancerous tissue β1≤5.

10. The method according to claim 1, characterized in that The method integrates a digital pathology platform to automatically generate a diagnostic report containing the following elements: Cancer probability score (0-100%); Heatmaps mark suspicious areas; Comparison table of key biophysical parameters; BRAF V600E mutation relevance prediction.

Images