A flexible array sensor based intraoperative lung nodule localization system

Abstract

The application belongs to the technical field of medicine, and discloses a lung nodule intraoperative positioning system based on a flexible array sensor, which comprises a flexible array sensor, is arranged at a detection site of a subject, is used for collecting a pressure signal, and obtains a stress distribution; a nodule region detection unit, which is used for determining a nodule region through cluster analysis according to the stress distribution; a morphological optimization unit, which is used for performing morphological optimization on the determined nodule region; a feature extraction unit, which is used for extracting relevant geometric features based on the optimized nodule region; and a visualization unit, which is used for visually displaying the stress distribution, the optimized nodule region and the relevant geometric features. The application can assist doctors to realize accurate positioning of lung nodules and provide quantitative features of the nodules.

Description

Technical Field

[0001] This invention belongs to the field of medical technology and relates to intraoperative positioning systems for lung nodules, and more particularly to an intraoperative positioning system for lung nodules based on a flexible array sensor. Background Technology

[0002] In the early diagnosis and treatment of lung cancer, precise localization of tiny pulmonary nodules detected preoperatively but difficult to palpate intraoperatively is crucial to the success of minimally invasive surgery. To address this challenge, various technical approaches have been developed clinically, including the following localization paradigms:

[0003] The first paradigm is invasive physical markers before surgery, represented by CT-guided hook-wires. This is currently the most widely used clinical "gold standard." Its core idea is to implant physical beacons such as metal hooks into the lung percutaneously under CT guidance before surgery, anchoring them near the nodule to provide path guidance during the operation. The fundamental drawback of this paradigm is its "invasive" nature. The puncture process itself brings significant risks such as pneumothorax, bleeding, infection, and marker dislodgement and displacement, increasing patient suffering and medical costs.

[0004] The second paradigm: real-time intraoperative surface detection, represented by intraoperative ultrasound (IOUS); this paradigm attempts to solve the above problems in a non-invasive way. However, due to the limitations of its physical principles, intraoperative ultrasound technology for lung nodules still has several unsolvable problems: (1) extreme sensitivity to residual gas, which is the core problem of IOUS in lung applications; even if the lung has collapsed, there may still be a small amount of residual gas in the alveoli or bullae. These tiny bubbles are the "enemy" of ultrasound, causing strong acoustic shadowing or comet-tail artifacts, completely obscuring the tissue behind, resulting in a large area becoming a blind spot for detection, making the nodules undetectable; (2) the invisibility of "isoechoic" nodules. Lesions), a considerable number of lung nodules (especially ground-glass nodules) have tissue density very close to that of the surrounding collapsed lung parenchyma, resulting in almost no difference in their acoustic impedance; on ultrasound images, these nodules will appear as "homoecho" with the same brightness as the background tissue, making them "invisible" on ultrasound images and unable to be identified; (3) Depth dependence and resolution limitation, although high-frequency probes have high resolution, their penetration depth is limited; for nodules buried deep (e.g., more than 2-3 cm from the lung surface), the ultrasound signal attenuates severely, the image quality drops sharply, and it is difficult to detect effectively; (4) High dependence on operator experience, the image quality of IOUS is far less clear than that of abdominal ultrasound, and interpreting images full of noise and artifacts requires surgeons to have extremely high ultrasound operation skills and image diagnosis experience, which makes the learning curve of this technology steep, the consistency and reliability of the results poor, and it is difficult to standardize and popularize it in hospitals at all levels.

[0005] The third paradigm involves high-tech internal navigation through natural bronchial orifices, represented by the Electromagnetic Navigation Bronchoscope (ENB). This paradigm represents another sophisticated minimally invasive approach, acting like a "GPS" for the lungs, guiding the catheter through the bronchial tree to the lesion via an electromagnetic field. Its fundamental flaw lies in the "mismatch between static data and dynamic reality." The ENB's navigation path is entirely based on preoperative static CT images. Intraoperative factors such as patient positioning, anesthesia, and lung collapse can cause unpredictable deformation of lung tissue, resulting in what is known as "CT-to-body Divergence" in the navigation system, thus affecting the final positioning accuracy. Furthermore, its equipment and consumables are extremely expensive, and its operation is complex, limiting its widespread application.

[0006] In summary, existing technologies either pose safety risks due to their invasiveness (hook-wire), suffer from insufficient reliability due to limitations in physical principles (IOUS), or are difficult to popularize due to system complexity, high cost, and registration bias (ENB). Currently, the market lacks an intraoperative positioning solution that can simultaneously meet the requirements of being non-invasive (to the lung parenchyma), real-time, easy to operate, low-cost, and based on direct feedback from tissue biomechanical properties. Summary of the Invention

[0007] The purpose of this invention is to address the shortcomings of existing technologies by providing an intraoperative lung nodule localization system based on a flexible array sensor. This system directly senses the difference in hardness between the nodule and normal tissue using the flexible array sensor, and combines cluster analysis and morphological optimization to assist doctors in accurately locating lung nodules while providing quantitative characteristics of the nodules.

[0008] To achieve the above objectives, the present invention adopts the following technical solutions.

[0009] This invention provides an intraoperative localization system for lung nodules based on a flexible array sensor, comprising:

[0010] A flexible array sensor is placed on the part of the subject to be tested to collect pressure signals and obtain stress distribution.

[0011] The nodule region detection unit is used to determine the nodule region based on stress distribution through cluster analysis.

[0012] The morphological optimization unit is used to perform morphological optimization on a defined nodule region.

[0013] The feature extraction unit is used to extract relevant geometric features based on the optimized nodule region;

[0014] The visualization unit is used to visualize the stress distribution, optimized nodal regions, and related geometric features.

[0015] In one implementation, the flexible array sensor comprises multiple pressure sensors arranged in an array on a flexible material; each pressure sensor corresponds to a node. Further, the flexible array sensor is preferably electronic skin.

[0016] In one possible approach, a pressure concentration zone is determined based on the pressure signal acquired by a flexible array of sensors, and the pressure increment of the pressure concentration zone is used as the basis for this determination. The two-dimensional distribution is used as the stress distribution.

[0017] In one possible implementation, the nodule region detection unit uses one of the following clustering analysis methods: K-means algorithm, Gaussian mixture model, or DBSCAN algorithm.

[0018] Specifically, this invention uses a Gaussian mixture model for cluster analysis, automatically classifying stress data into normal and abnormal categories, generating an initial nodule mask, and thus initially determining the nodule region.

[0019] The Gaussian mixture model is represented as follows:

[0020] ;

[0021] in, Representing data points The probability density at that location; This represents the mixture weights of the k-th Gaussian distribution; The mean is Covariance is The Gaussian distribution.

[0022] Furthermore, to further improve the detection accuracy of nodule boundaries, this invention introduces frequency domain analysis to correct the initial clustering results. The original pressure signals corresponding to the pressure sensor nodes near the nodule region boundary are converted to the frequency domain, and the nodule region boundary is corrected based on frequency domain analysis.

[0023] In the specific implementation, based on the nodule boundaries determined by the initial nodule mask obtained from cluster analysis, the original force signals collected by the pressure sensor nodes within a 3-5 pressure sensor node width annular region inside and outside the boundary contour are extracted; the original force signal of each pressure sensor node is represented as follows: ;

[0024] The correction of nodule region boundaries based on frequency domain analysis includes the following steps:

[0025] (1) Perform a Fast Fourier Transform (FFT) on the original force signal to obtain the signal in the frequency domain. :

[0026] ;

[0027] in, Indicates frequency; This represents the pressure amplitude at the nth sampling point; N represents the total number of sampling points.

[0028] (2) Extract the main frequency of each pressure sensor node That is, the frequency component with the highest energy in the power spectral density;

[0029] ;

[0030] (3) Perform global statistics on all pressure sensor nodes and calculate the global average frequency. :

[0031] ;

[0032] Where X and Y represent the number of rows and columns in the array sensor, respectively;

[0033] (4) Calculate the global dominant frequency standard deviation :

[0034] ;

[0035] (5) Set the threshold for determining frequency domain anomalies at the boundary:

[0036] ;

[0037] (6) For each pressure sensor node within the boundary region, if its dominant frequency If the condition is met, the pressure sensor node is marked as an anomaly and adjusted to the nodule mask boundary; otherwise, the boundary is not adjusted, meaning the pressure sensor node is not adjusted and the node mask remains unchanged.

[0038] This helps eliminate misclassifications caused by noise or local variations, improving boundary conditions.

[0039] In one possible implementation, the morphological optimization unit includes:

[0040] The closing operation module is used to perform closing operations on a defined nodule region, filling the small holes inside the nodule region.

[0041] An open operation module is used to eliminate isolated noise points outside the nodule region.

[0042] The aforementioned closed-loop operation module first performs expansion processing on the interior of the nodule region, then performs corrosion processing, connects adjacent abnormal points, and fills in the internal small holes.

[0043] The aforementioned opening operation module first performs erosion processing on the nodule region after the closing operation unit, and then performs expansion processing to eliminate isolated noise points outside the nodule region.

[0044] In one possible implementation, the geometric features extracted by the feature extraction unit include at least one of area, perimeter, equivalent diameter, and circularity. In this invention, connected component analysis can also be used to first filter out valid nodule regions with areas greater than a set area threshold, and then calculate the quantified geometric features of the valid nodule regions.

[0045] In one possible implementation, the aforementioned intraoperative positioning system for lung nodules based on flexible array sensors further includes a physical deformation correction unit for correcting the equivalent diameter.

[0046] Specifically, the physical deformation correction unit corrects the equivalent diameter based on the nodule depth value using a correction function, as follows:

[0047] ;

[0048] in, This represents the correction function; the correction function is determined through multinomial regression or by looking up a table.

[0049] The polynomial regression is a quadratic polynomial regression; the correction function determined by the quadratic polynomial regression is expressed as:

[0050] ;

[0051] in, This represents the corrected true diameter; Indicates the equivalent diameter before correction; Indicates the depth of the nodule; express The regression fitting coefficients of power i and power j of d. It can be calculated using the least squares method.

[0052] The correction function determined by bilinear interpolation, through a lookup table, is expressed as follows:

[0053] ;

[0054] in, Indicates the corrected true diameter; l , , This represents the equivalent diameter calculated so far. This represents the minimum value of a given equivalent diameter. Indicates a given diameter step size. Indicates the current depth of a given nodule. This represents the minimum depth of a given nodule. Indicates a given depth step size, This represents the floor function; (h,l) represents the grid point position determined based on the current equivalent diameter. This represents the experimental value of the true diameter at grid point (l, h) determined by a lookup table, which is represented as follows: , Indicates the grid size of the lookup table. This represents the l-th measured diameter. Indicates the h-th depth; These are relative coordinate parameters, determined by the grid point (l, h). .

[0055] Compared with existing technologies, the intraoperative localization system for lung nodules based on flexible array sensors provided by this invention has the following advantages:

[0056] (1) Fundamentally improves detection reliability and applicability: This invention adopts a brand-new mechanical sensing principle and locates the nodules by measuring the difference in tissue hardness. Hardness is a mechanical property of solids and is almost unaffected by trace amounts of gas. Moreover, most lung nodules (especially malignant lesions) have a significant difference in hardness compared to normal lung parenchyma. Therefore, this invention fundamentally avoids the physical bottleneck of IOUS, significantly improves the success rate of detection in complex intraoperative environments, and expands the applicability to nodule types that are difficult to detect with traditional ultrasound.

[0057] (2) It realizes the objectification and standardization of the detection process and provides rich quantitative information: The implementation of IOUS is highly dependent on the operator's experience. The subjective interpretation of noisy grayscale images leads to poor consistency of results and a steep learning curve. Moreover, this invention transforms the nodule identification process into an automated, mathematical model-based objective process through Gaussian mixture model clustering and morphological optimization. It can stably and repeatedly segment nodule regions from complex stress data, completely eliminating the dependence on the operator's "feel" and "eyesight". More importantly, it not only realizes the localization, but also automatically outputs key quantitative parameters such as the equivalent diameter and roundness of the nodule, providing unprecedented objective data support for doctors' intraoperative decisions (such as assessing the cutting margin and judging the benign or malignant tendency), realizing the leap from "localization" to "quantitative analysis".

[0058] (3) Higher dynamic positioning accuracy and cost-effectiveness: This invention is a real-time detection technology that directly measures the physical characteristics of lung tissue at the current moment and in the current state, and there is no registration deviation problem at all. In addition, this invention can actively calibrate the measurement error caused by tissue pressure through the physical deformation correction unit, and achieves high accuracy that existing surface detection technology cannot achieve. In terms of cost, the system structure of this invention (flexible sensor + standard processing unit) is much simpler than ENB or high-end intraoperative ultrasound equipment, and has extremely high cost-effectiveness, which is conducive to its promotion and popularization in hospitals at all levels.

[0059] (4) Significantly improves safety and greatly simplifies clinical procedures: This invention is completely non-invasive to the lung parenchyma, fundamentally eliminating the safety risks associated with puncture; at the same time, this invention fully integrates the positioning process into the surgical procedure, making it a "one-stop" operation that can be completed independently by the surgeon, which greatly simplifies the clinical workflow, shortens the operation time, and reduces the patient's pain and medical costs. Attached Figure Description

[0060] Figure 1 It is a phantom model.

[0061] Figure 2 This is a schematic diagram of the intraoperative positioning system for lung nodules based on a flexible array sensor provided by the present invention.

[0062] Figure 3 The results of nodule processing using the intraoperative positioning system for lung nodules are shown below; (a) is the stress distribution map, (b) is the cluster analysis result based on the Gaussian mixture model, (c) is the result of correcting the nodule region boundary based on frequency domain analysis, and (d) is the frequency domain projection result of the nodule region after morphological optimization.

[0063] Figure 4 The results show the analysis of the true diameter, equivalent diameter, and corrected diameter of the nodule; among them, (a) is the Bland-Altman analysis result of the corrected diameter and the true diameter, and (b) is a comparison diagram of the true diameter, equivalent diameter, and corrected diameter of the nodule.

[0064] Figure 5 Visualize the results of nodule detection.

[0065] Figure 6 The results are from the trend analysis of nodule diameter. Detailed Implementation

[0066] The technical solutions of various embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] The phantom model used in the following examples is a pig lung model. Gel beads (0.25cm-2.0cm) are embedded in the pig lung model, and the depth is measured (0.5cm-3.0cm) using a cotton swab to make indentations. Figure 1 As shown.

[0068] Example 1

[0069] This embodiment provides an intraoperative localization system for lung nodules based on a flexible array sensor, such as... Figure 2 As shown, it includes a flexible array sensor, a nodule region detection unit, a morphological optimization unit, a feature extraction unit, a physical deformation correction unit, and a visualization unit.

[0070] 1. Flexible array sensor

[0071] Flexible array sensors are placed on the part of the subject to be tested to collect pressure signals and obtain stress distribution.

[0072] In this embodiment, the significant difference in elastic modulus between lung nodules and normal lung parenchyma is utilized. Normal collapsed lung tissue has a lower elastic modulus (Etissue), while lung nodules (especially solid nodules) have a significantly higher elastic modulus (Enodule), i.e., Enodule >> Etissue. Due to the greater stiffness of the nodules, their deformation is much smaller than that of the surrounding normal tissue. This results in the tissue directly above the nodule being unable to effectively unload pressure, thus forming a stress concentration zone on the sensor surface.

[0073] The flexible array sensor consists of multiple pressure sensors arranged in an array on a flexible material. In this embodiment, electronic skin is used as the flexible array sensor. Specifically, the electronic skin uses a humanoid fingertip pressure sensor from Moxian Technology, with an array size of 12×8 (96 sensing units) and a thickness of only 0.025mm, exhibiting high flexibility and biocompatibility. Each pressure sensor corresponds to one node. This flexible array sensor is used to collect pressure signals and display them in real time, such as... Figure 1 As shown.

[0074] The basic physical principle of this invention is to combine the significant difference in elastic modulus between lung nodules and normal lung parenchyma at the macroscopic level with the frequency change caused by the abrupt change in stiffness at the microscopic stress wave level (which is reflected in the increase of high-frequency components due to scattering and reflection when stress waves propagate at the edge of nodules).

[0075] When a flexible array sensor is applied to the lung surface and a relatively uniform back pressure (Papplied) is applied, the tissue beneath the sensor deforms, collecting two-dimensional stress distribution data caused by differences in nodule stiffness, forming an X×Y stress matrix (Praw). Because the nodules are stiffer, their deformation is much smaller than that of the surrounding normal tissue. This results in the tissue directly above the nodule being unable to effectively unload the pressure, thus forming a stress concentration zone on the surface of the flexible array sensor. The pressure value P(x,y) in this zone will be significantly higher than the surrounding area. This pressure increment ΔP(x,y) is expressed as:

[0076] ;

[0077] in, This represents the average pressure in the background area (normal tissue).

[0078] It is a function related to the difference in elastic modulus, the nodule burial depth d, and the nodule size S.

[0079] In this embodiment, the two-dimensional distribution of ΔP(x,y) is used as the stress distribution (e.g., Figure 3As shown in (a), the location, size and shape of nodules can be inferred by accurately measuring and analyzing the two-dimensional distribution of ΔP(x,y).

[0080] 2. Nodule region detection unit

[0081] The nodule region detection unit is used to determine the nodule region based on stress distribution through cluster analysis.

[0082] The nodule region detection unit first performs median filtering on the acquired raw stress matrix Praw to remove potential salt-and-pepper noise, resulting in a preprocessed stress matrix Pproc.

[0083] This invention employs a Gaussian mixture model (GMM) to perform unsupervised clustering of pressure distribution data, automatically classifying stress data into normal and abnormal categories to generate an initial nodule mask. It assumes that the pressure data consists of two distributions: a background distribution representing normal tissue and a high-stress distribution representing nodules. The GMM then uses the overall pressure distribution p( The model is a weighted sum of K = 2 Gaussian components.

[0084] The Gaussian mixture model is represented as follows:

[0085] ;

[0086] in, Representing data points The probability density at that location; This represents the mixture weights of the k-th Gaussian distribution; The k-th mean is Covariance is The Gaussian distribution.

[0087] The parameters are solved iteratively using the Expectation Maximization (EM) algorithm. , , ), calculate the posterior probability that each sensor node belongs to the nodal component. ;like ( If the set posterior probability threshold is met, the sensor node is marked as "1" (node); otherwise, it is marked as "0" (background), thus classifying each data point... Classifying into the class with the highest posterior probability enables automatic segmentation of high-stress areas from the background area (e.g.) Figure 3 As shown in (b), the initial binarized mask region M is generated. GMM .

[0088] To further improve the detection accuracy of nodule boundaries, this invention introduces frequency domain analysis to correct the initial clustering results. The original pressure signals corresponding to the pressure sensor nodes near the nodule region boundary are converted to the frequency domain, and the nodule region boundary is corrected based on frequency domain analysis.

[0089] In the specific implementation, based on the nodule boundaries determined by the initial nodule mask obtained from cluster analysis, the original force signals collected by the pressure sensor nodes within a 3-5 pressure sensor node width annular region inside and outside the boundary contour are extracted; the original force signal of each pressure sensor node is represented as follows: ;

[0090] The correction of nodule region boundaries based on frequency domain analysis includes the following steps:

[0091] (1) Perform a Fast Fourier Transform (FFT) on the original force signal to obtain the signal in the frequency domain. :

[0092] ;

[0093] in, Indicates frequency; This represents the pressure amplitude at the nth sampling point; N represents the total number of sampling points.

[0094] (2) Extract the main frequency of each pressure sensor node That is, the frequency component with the highest energy in the power spectral density;

[0095] ;

[0096] (3) Perform global statistics on all pressure sensor nodes and calculate the global average frequency. :

[0097] ;

[0098] Where X and Y represent the number of rows and columns in the array sensor, respectively; in this embodiment, X=12 and Y=8.

[0099] (4) Calculate the global dominant frequency standard deviation :

[0100] ;

[0101] (5) Set the frequency domain anomaly threshold based on the 3σ criterion:

[0102] ;

[0103] (6) For each pressure sensor node within the boundary region, if its dominant frequency Then mark the pressure sensor node as an abnormal point (e.g. Figure 3 (as shown in (c)) and adjust it to the nodule mask boundary; otherwise, the boundary is not adjusted, that is, the pressure sensor node is not adjusted and the node mask remains unchanged.

[0104] This helps eliminate misclassifications caused by noise or local variations, improving boundary accuracy. For any given sensor node, if The nodule boundary region, and ,but ;otherwise .

[0105] This step generates the initial nodule mask; for each pressure sensor node, if... (It has a higher probability of belonging to the nodule component) and (Increased local high-frequency components) are marked as "1" (nodule), otherwise as "0" (background), generating the final preliminary nodule mask region. .

[0106] 3. Morphological optimization unit

[0107] The morphological optimization unit is used to perform morphological optimization on the defined nodule region. In this embodiment, the nodule mask region... Isolated noise points are removed and internal holes are filled using morphological optimization units.

[0108] The morphological optimization unit includes a closing operation module and an opening operation module.

[0109] The closing operation module is used to perform closing operations on a defined nodule region, filling the small holes inside the nodule region.

[0110] In the specific implementation, the closing operation module first performs expansion processing on the interior of the nodule region, then performs erosion processing, connects adjacent anomalies, and fills in internal small holes. This process can be represented as:

[0111] ;

[0112] This indicates that the initial nodule mask is first applied. An expansion process is performed, followed by an etching process to fill the small pores inside the nodule region. B is a preset structural element.

[0113] .

[0114] An open operation module is used to eliminate isolated noise points outside the nodule region.

[0115] The open operation module first performs erosion on the nodule region after processing by the closed operation unit, followed by dilation, to eliminate isolated noise points outside the nodule region. This process can be represented as:

[0116] ;

[0117] This represents the nodal region after processing the closing operation unit. The process involves corrosion followed by expansion to eliminate isolated noise points outside the nodule region.

[0118] The heatmap of the nodule mask region after morphological optimization, projected into the frequency domain, is as follows: Figure 3 As shown in (d) in the figure.

[0119] 4. Feature Extraction Unit

[0120] The feature extraction unit is used to extract relevant geometric features based on the optimized nodule region. The geometric features extracted by the feature extraction unit include area, perimeter, equivalent diameter, and circularity.

[0121] In this embodiment, based on the final nodule mask First, the connected component area is calculated by marking the connected components and calling the regionprops method. Then, based on the set area threshold, the effective nodule regions with areas greater than the set threshold are selected. Finally, the quantitative geometric features of the effective nodule regions are calculated.

[0122] Area: refers to the total number of pixels within the mask area.

[0123] Perimeter: refers to the length of the outline of the masked area.

[0124] Equivalent Diameter (Measured Diameter): .

[0125] Circularity: The roundness value ranges from 0 to 1, with the value closer to 1 indicating a more rounded shape.

[0126] 5. Physical Deformation Correction Unit

[0127] Because soft tissue deforms under pressure, the depth effect can lead to an inaccurate projection of the nodule size measured by the sensor on the surface. It will systematically be larger than its actual size. Moreover, this deviation is strongly correlated with the burial depth d of the nodule.

[0128]

[0129] To address this problem, the present invention constructs a physical deformation correction unit for correcting the equivalent diameter.

[0130] Data points were systematically collected through in vitro phantom experiments (using silicone implants of different sizes and depths to simulate nodules) or animal experiments. Based on this dataset, multinomial regression or a lookup table can be used. The correction function is constructed in the following way, and is expressed as:

[0131] ;

[0132] in, This represents the correction function.

[0133] Here, the polynomial regression is a quadratic polynomial regression; the correction function determined by the quadratic polynomial regression is expressed as:

[0134] ;

[0135] in, This represents the corrected true diameter; Indicates the equivalent diameter before correction; Indicates the depth of the nodule; express The regression fitting coefficients of power i and power j of d. It can be calculated using the least squares method.

[0136] The lookup table method establishes discrete input-output mappings based on experimental data, achieving fast responses to arbitrary inputs through table lookup and interpolation. This method avoids the assumption limitations of parameterized models, directly establishing corrective relationships based on experimental data.

[0137] Lookup table construction:

[0138] ;

[0139] in, Indicates the grid size of the lookup table. This represents the l-th measured diameter. This represents the h-th depth. The constructed lookup table is shown in Table 1.

[0140] Table 1 Nodule Size Correction Lookup Table

[0141]

[0142] The correction function determined by bilinear interpolation, through a lookup table, is expressed as follows:

[0143] ;

[0144] in, Indicates the corrected true diameter; l , , This represents the equivalent diameter calculated so far. This represents the minimum value of a given equivalent diameter. This indicates a given diameter step size (typically 0.5cm, but precision can be increased as needed). Indicates the current depth of a given nodule. This represents the minimum depth of a given nodule. This indicates a given depth step size (typically 0.5 cm, but precision can be increased as needed). This represents the floor function; (h,l) represents the grid point position determined based on the current equivalent diameter. This represents the experimental value of the true diameter at grid point (l, h) determined by a lookup table, which is represented as follows: These are relative coordinate parameters, determined by the grid point (l, h). .

[0145] Finally, Bland-Altman analysis was used to evaluate the performance of the two correction methods:

[0146] ;

[0147] in, , ; , Let represent the true diameter and the corrected true diameter of the nodule in the q-th sample, respectively; Let represent the depth of the q-th nodule; Q represents the total number of samples.

[0148] Figure 4 Figure (a) shows the diameter correction effect of the correction function constructed using the above-mentioned quadratic polynomial on 0.5cm nodules at different depths, obtained by the method of this embodiment. It can be seen from the figure that most of the scatter points fall within the 0.95 consistency bound, indicating that the diameter correction result is consistent with the actual result. Figure 4 Figure (b) in the figure shows a comparison of the equivalent diameter, the true diameter, and the corrected diameter of the nodule. Overall, the corrected diameter is closer to the true diameter.

[0149] 6. Visualization Unit

[0150] The visualization unit is used to visualize the stress distribution, optimized nodal regions, and related geometric features.

[0151] Stress distribution refers to the two-dimensional and / or three-dimensional distribution of stress data measured in real time, such as... Figure 5As shown.

[0152] The optimized nodule region refers to the nodule mask region after morphological optimization, which can be mapped to the frequency domain to obtain the corresponding heatmap, such as... Figure 5 As shown.

[0153] Relevant geometric features can be used for trend analysis of nodule area, true nodule diameter, etc., such as... Figure 6 As shown.

[0154] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A lung nodule intraoperative localization system based on a flexible array sensor, characterized in that, include: A flexible array sensor is placed on the part of the subject to be tested to collect pressure signals and obtain stress distribution. Specifically, based on the collected pressure signals, pressure concentration zones are determined, and the pressure increment in these zones is used as the basis for analysis. The two-dimensional distribution is used as the stress distribution; The nodule region detection unit is used to determine the nodule region based on the stress distribution through cluster analysis, then convert the original pressure signal corresponding to the pressure sensor node near the nodule region boundary to the frequency domain, and correct the nodule region boundary based on the frequency domain analysis. Specifically, based on the nodule boundaries determined by the initial nodule mask obtained from cluster analysis, the original force signals collected by pressure sensor nodes within a 3-5 pressure sensor node width annular region inside and outside the boundary contour are extracted; the original force signal of each pressure sensor node is represented as follows: ; The correction of nodule region boundaries based on frequency domain analysis includes the following steps: (1) Perform a fast Fourier transform on the original force signal to obtain the signal in the frequency domain. : ； in, Indicates frequency; This represents the pressure amplitude at the nth sampling point; N represents the total number of sampling points. (2) Extract the main frequency of each pressure sensor node ; ； (3) Perform global statistics on all pressure sensor nodes and calculate the global average frequency. : ； Where X and Y represent the number of rows and columns in the array sensor, respectively; (4) Calculate the global dominant frequency standard deviation : ； (5) Set the threshold for determining frequency domain anomalies at the boundary: ； (6) For each pressure sensor node within the boundary region, if its dominant frequency If the condition is met, the pressure sensor node is marked as an anomaly and adjusted to the nodule mask boundary; otherwise, the boundary is not adjusted. The morphological optimization unit is used to perform morphological optimization on a defined nodule region. The feature extraction unit is used to extract relevant geometric features based on the optimized nodule region; The visualization unit is used to visualize the stress distribution, optimized nodal regions, and related geometric features.

2. The intraoperative localization system for lung nodules based on a flexible array sensor according to claim 1, characterized in that, The flexible array sensor consists of multiple pressure sensors arranged in an array on a flexible material.

3. The intraoperative localization system for lung nodules based on a flexible array sensor according to claim 1, characterized in that, The nodule region detection unit uses one of the following clustering algorithms: K-means, Gaussian mixture model, or DBSCAN.

4. The intraoperative localization system for lung nodules based on a flexible array sensor according to claim 3, characterized in that, Cluster analysis uses a Gaussian mixture model, expressed as: ； in, Representing data points The probability density at that location; This represents the mixture weights of the k-th Gaussian distribution; The mean is Covariance is The Gaussian distribution.

5. The intraoperative localization system for lung nodules based on a flexible array sensor according to claim 1, characterized in that, The morphological optimization unit includes: The closing operation module is used to perform closing operations on a defined nodule region, filling the small holes inside the nodule region. An open operation module is used to eliminate isolated noise points outside the nodule region.

6. The intraoperative localization system for lung nodules based on a flexible array sensor according to claim 1, characterized in that, The feature extraction unit first filters out the effective nodule regions based on a set area threshold, and then extracts the geometric features of the effective nodule regions. Geometric features include at least one of area, perimeter, equivalent diameter, and circularity.

7. The intraoperative localization system for lung nodules based on a flexible array sensor according to claim 6, characterized in that, It also includes a physical deformation correction unit for correcting the equivalent diameter.

8. The intraoperative localization system for lung nodules based on a flexible array sensor according to claim 7, characterized in that, The physical deformation correction unit corrects the equivalent diameter based on the nodule depth value using a correction function, expressed as follows: ； in, This represents the correction function; the correction function is determined through multinomial regression or by looking up a table.

9. The intraoperative localization system for lung nodules based on a flexible array sensor according to claim 8, characterized in that, The polynomial regression is a quadratic polynomial regression; the correction function determined by the quadratic polynomial regression is expressed as: ； in, This represents the corrected true diameter; Indicates the equivalent diameter before correction; Indicates the depth of the nodule; express The regression fit coefficients of d raised to the power of i and d raised to the power of j; The correction function determined by the lookup table is expressed as follows: ； in, Indicates the corrected true diameter; l , , This represents the equivalent diameter calculated so far. This represents the minimum value of a given equivalent diameter. Indicates a given diameter step size. Indicates the current depth of a given nodule. This represents the minimum depth of a given nodule. Indicates a given depth step size, This represents the floor function; (h,l) represents the grid point position determined based on the current equivalent diameter. This represents the experimental value of the true diameter at grid point (l, h) determined by a lookup table, which is represented as follows: , Indicates the grid size of the lookup table. This represents the l-th measured diameter. Indicates the h-th depth; These are relative coordinate parameters, determined by the grid point (l, h). .

Images