WO2026001663A1 - 激光功率确定模型构建方法及激光功率自动调节方法 - Google Patents

激光功率确定模型构建方法及激光功率自动调节方法

Info

Publication number
WO2026001663A1
WO2026001663A1 PCT/CN2025/100098 CN2025100098W WO2026001663A1 WO 2026001663 A1 WO2026001663 A1 WO 2026001663A1 CN 2025100098 W CN2025100098 W CN 2025100098W WO 2026001663 A1 WO2026001663 A1 WO 2026001663A1
Authority
WO
WIPO (PCT)
Prior art keywords
laser power
fiber core
solution
brightness
average
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
PCT/CN2025/100098
Other languages
English (en)
French (fr)
Inventor
段西尧
冯宇
马骁萧
丁莽
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Viestar Hubei Medical Technology Co Ltd
Original Assignee
Viestar Hubei Medical Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from CN202410852246.5A external-priority patent/CN118736350A/zh
Priority claimed from CN202410852114.2A external-priority patent/CN118717002A/zh
Application filed by Viestar Hubei Medical Technology Co Ltd filed Critical Viestar Hubei Medical Technology Co Ltd
Publication of WO2026001663A1 publication Critical patent/WO2026001663A1/zh
Pending legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting

Definitions

  • This application relates to the field of endoscopic imaging, and more specifically, to a method for constructing a laser power determination model and a method for automatically adjusting laser power.
  • Confocal endoscopy is a medical device that can be inserted into the human body through channels such as gastroscopes and colonoscopes to obtain local histological images, enabling precise diagnosis of tiny lesions, gastrointestinal diseases, and early gastrointestinal cancers.
  • the basic principle of confocal endoscopy is to illuminate tissue with a laser and then detect the fluorescence reflected from the tissue. Under the condition of a certain concentration of fluorescent dye, different laser powers will produce images of varying quality. Excessively high or low laser power will result in images that are too bright or too dark, making it impossible to distinguish details and lacking sufficient contrast, thus reducing image quality. Therefore, appropriate laser power is crucial for image quality.
  • this application proposes a method for constructing a laser power determination model, the method comprising:
  • N frames of image data of the aforementioned endoscope probe in the air are acquired to construct an air environment image dataset
  • N frames of image data were acquired of the endoscope probe immersed in a sodium fluorescein solution of a reference concentration to construct a solution environment image dataset.
  • the aforementioned laser power determination model includes a first correspondence and a second correspondence.
  • the first correspondence is the correspondence between air environment image data and the aforementioned K different laser powers
  • the second correspondence is the correspondence between solution environment image data and the aforementioned K different laser powers.
  • the above-mentioned data fitting based on the above-mentioned air environment image dataset, the above-mentioned solution environment image dataset, the laser power set, and the preset curve fitting method to obtain the above-mentioned laser power determination model includes:
  • the average brightness signal values at all fiber core locations under each laser power are used to construct a set of average fiber core brightness signals in the air.
  • the average brightness signal values at all fiber core locations under each laser power are used to construct a set of average fiber core brightness signals in solution.
  • Data fitting is performed based on the above-mentioned average value set of fiber core brightness signals in air, the above-mentioned laser power set, and the above-mentioned preset curve fitting method to construct the above-mentioned first correspondence relationship;
  • the data is fitted using the average set of fiber core brightness signals in the solution, the laser power set, and the preset curve fitting method to construct the second correspondence.
  • the laser power determination model is created.
  • it further includes:
  • the fiber core is located using the aforementioned air environment image dataset or the aforementioned solution environment image dataset to obtain the fiber core position.
  • the method for obtaining the average value of the brightness signal at the aforementioned fiber core location and the average value of the fiber core brightness signal in the solution is specifically as follows:
  • the average of N frames of image data acquired at each laser power is calculated pixel by pixel to obtain the average image set in air.
  • Image set of mean values in solution
  • the average value set of fiber core brightness signal in the solution is extracted from the fiber core location located on the mean image set in the above solution.
  • the aforementioned reference concentration is the average value of the fiber core brightness signal in the reference solution.
  • the average brightness signal of the fiber core in the aforementioned reference solution is the average brightness signal of all fiber core positions extracted from the located fiber core position on the average image set in the reference solution.
  • the aforementioned average image set in the reference solution is obtained by averaging multiple frames of image data obtained by immersing the endoscope probe in the aforementioned reference concentration of fluorescein sodium solution at maximum laser power.
  • the brightness value in the preset brightness value range is the maximum brightness value that the hardware can achieve multiplied by a coefficient greater than or equal to 0.5 and less than or equal to 1.
  • this application also proposes a method for automatic adjustment of laser power in a confocal endoscope, including:
  • a model is constructed based on the current concentration, laser power, and fiber core brightness to determine the laser power corresponding to the current concentration of sodium fluorescein solution when the fiber core brightness signal just reaches the maximum brightness value achievable by the hardware.
  • the above-mentioned model for constructing the current concentration, laser power, and fiber core brightness to determine the laser power corresponding to the current concentration of sodium fluorescein solution when the fiber core brightness signal just reaches the maximum brightness value achievable by the hardware specifically includes:
  • the brightness signal value adjustment range is determined based on the maximum brightness value that the hardware can achieve, the first adjustment coefficient, and the second adjustment coefficient; the laser power adjustment range is determined based on the maximum laser power and the minimum laser power of the laser.
  • the laser power determination model is constructed based on the concentration of sodium fluorescein solution, the brightness signal value of the fiber core, and the laser power.
  • the laser power at the point where the maximum brightness value that the hardware can achieve is calculated based on the current concentration of sodium fluorescein solution.
  • the laser power at which the brightness signal reaches its maximum value is determined as the adjusted target laser power.
  • the current concentration of the sodium fluorescein solution corresponding to the current laser power is calculated based on the average of the larger brightness values, including:
  • the current concentration of sodium fluorescein solution corresponding to the current laser power is calculated.
  • the method for obtaining a valid image set includes:
  • the cached images in the cached image set are filtered to obtain the effective image set.
  • the laser power determination model construction method of this application embodiment includes: selecting K different laser powers to construct a laser power set; acquiring N frames of image data of the endoscope probe in air at each laser power to construct an air environment image dataset; acquiring N frames of image data of the endoscope probe immersed in a reference concentration of sodium fluorescein solution at each laser power to construct a solution environment image dataset; and performing data fitting based on the air environment image dataset, the solution environment image dataset, the laser power set, and a preset curve fitting method to obtain the laser power determination model.
  • the laser power determination model includes a first correspondence and a second correspondence. The first correspondence is the correspondence between the air environment image data and the K different laser powers, and the second correspondence is the correspondence between the solution environment image data and the K different laser powers.
  • the method proposed in this application embodiment by precisely controlling the laser power, automatically adjusts the laser power according to different environments (air and fluorescent dye solution), ensuring optimal image quality under various conditions.
  • This method can effectively cope with individual differences and changes in fluorescent dye concentration, ensuring moderate brightness and contrast in the image, thereby making it easier to identify details and lesions.
  • Automated laser power adjustment reduces the burden on physicians operating confocal endoscopes. Physicians do not need in-depth knowledge of how to adjust laser power based on images, reducing training requirements and minimizing human error during operation.
  • the model proposed in this application enables automated laser power adjustment, shortening equipment setup time and accelerating the diagnostic process.
  • the model in this application is adaptable to different types of confocal endoscopes and diverse clinical environments.
  • This application's embodiment establishes a laser power determination model based on actual image data, achieving automated laser power adjustment of confocal endoscopes in clinical use. This effectively improves image quality and operational efficiency while reducing the complexity of physician operations and training requirements, providing clinicians with a more efficient and accurate diagnostic tool.
  • the automatic laser power adjustment method for confocal endoscopes includes: determining a brightness signal value adjustment range based on the maximum brightness value achievable by the hardware, a first adjustment coefficient, and a second adjustment coefficient; determining a laser power adjustment range based on the maximum laser power of the laser and the maximum laser power; acquiring an effective image set using the maximum laser power as the initial laser power; acquiring a larger brightness value for each frame in the effective image set; and determining the adjusted target laser power based on the average of the larger brightness values and the brightness signal value adjustment range.
  • the method proposed in this application achieves automatic laser power adjustment, enabling the endoscope to automatically adjust to the optimal laser output during tissue imaging, thereby improving image quality, simplifying the operation process, and reducing the burden on doctors during operation.
  • Figure 1 is a flowchart illustrating a laser power determination model construction method provided in an embodiment of this application.
  • Figure 2 is a flowchart illustrating an automatic laser power adjustment method for a confocal endoscope according to an embodiment of this application.
  • FIG. 3 is a schematic diagram of an operator structure provided in an embodiment of this application.
  • FIG. 4 is a schematic diagram of another operator structure provided in an embodiment of this application.
  • FIG. 5 is a schematic diagram of another operator structure provided in an embodiment of this application.
  • Figure 6 is a schematic diagram of another operator structure provided in an embodiment of this application.
  • Figure 1 is a flowchart illustrating a laser power determination model construction method provided in an embodiment of this application. Specifically, it may include:
  • selected Different laser powers are used to construct a laser power set. .
  • N frames of image data were captured in an air environment to form an air environment image dataset.
  • step S120 Similar to step S120, but this time the probe is immersed in a sodium fluorescein solution of a reference concentration. Similarly, at K different laser powers... To capture N frames of images, a solution environment image dataset is created. .
  • the above-mentioned solution environment image dataset Based on the above-mentioned air environment image dataset, the above-mentioned solution environment image dataset, laser power set and preset curve fitting method, perform data fitting to obtain the above-mentioned laser power determination model, wherein the above-mentioned laser power determination model includes a first correspondence relationship and a second correspondence relationship, the above-mentioned first correspondence relationship is the correspondence relationship between air environment image data and the above-mentioned K different laser powers, and the above-mentioned second correspondence relationship is the correspondence relationship between solution environment image data and the above-mentioned K different laser powers.
  • a pre-defined curve fitting method is used for data analysis and fitting.
  • a laser power determination model is established, which includes two corresponding relationships:
  • the first correspondence f1 the relationship between air environment image data and different laser powers.
  • the second correspondence f2 the relationship between solution environment image data and different laser powers.
  • the method proposed in this application by precisely controlling laser power and automatically adjusting it according to different environments (air and fluorescent dye solution), ensures optimal image quality under various conditions.
  • This method effectively addresses individual differences and variations in fluorescent dye concentration, ensuring moderate brightness and contrast in the image, thus making it easier to identify details and lesions.
  • Automated laser power adjustment reduces the burden on physicians operating confocal endoscopes. Physicians do not need in-depth knowledge of how to adjust laser power based on images, reducing training requirements and minimizing human error during operation.
  • the model proposed in this application can be used to achieve automated laser power adjustment, shortening equipment adjustment time and accelerating the diagnostic process.
  • the model in this application can adapt to different types of confocal endoscopes and diverse clinical environments.
  • the model's versatility allows for easy integration into different hardware and software configurations, providing a wider range of applications.
  • this application achieves automatic laser power adjustment of confocal endoscopes in clinical use, effectively improving image quality and operational efficiency while reducing the complexity of physician operation and training requirements, providing a more efficient and accurate diagnostic tool for clinicians.
  • the above-mentioned data fitting based on the above-mentioned air environment image dataset, the above-mentioned solution environment image dataset, the laser power set, and the preset curve fitting method is used to obtain the above-mentioned laser power determination model, including:
  • the average brightness signal values at all fiber core locations under each laser power are used to construct a set of average fiber core brightness signals in the air.
  • the average brightness signal values at all fiber core locations under each laser power are used to construct a set of average fiber core brightness signals in solution.
  • Data fitting is performed based on the above-mentioned average value set of fiber core brightness signals in air, the above-mentioned laser power set, and the above-mentioned preset curve fitting method to construct the above-mentioned first correspondence relationship;
  • the data is fitted using the average set of fiber core brightness signals in the solution, the laser power set, and the preset curve fitting method to construct the second correspondence.
  • the laser power determination model is created.
  • the average image set of air at all laser powers can be obtained by taking the average value of pixel regions from the above air environment image dataset. and in The average value of the brightness signal at all fiber core locations is extracted and denoted as the set of average brightness signal values of the fiber core in air. ,in .
  • the mean image set of sodium fluorescein solution at all laser powers was obtained by taking the mean value of the solution environment image dataset. ;Depend on Extract the average value of the brightness signal at all fiber core locations to form a set of average brightness signal values of the fiber core in the solution. ,in .
  • f1 is the first correspondence.
  • f2 represents the second correspondence.
  • the laser power determination model is a model that includes the first correspondence and the second correspondence.
  • it also includes:
  • the fiber core is positioned to obtain its location.
  • the fiber core is located using the aforementioned air environment image dataset or the aforementioned solution environment image dataset.
  • an effective image set can be obtained using an air environment image dataset or the aforementioned solution environment image dataset. Then, the median value is calculated for these effective images pixel by pixel.
  • An effective image set refers to a collection of clear images selected from a large amount of imaging data through some filtering mechanism. This filtering aims to ensure the quality of the selected images, excluding blurry or damaged images. The purpose of calculating the median image is to reduce or eliminate random noise, accidental errors, or other one-off anomalies. Median filtering is a non-linear process that is particularly effective in preserving image edge and structural details. Based on the effective image set, a statistical method is used to calculate the cumulative probability of the fiber core's location.
  • the possible locations of the fiber core in each frame are scored or their probabilities evaluated, and these evaluation values are then accumulated to form an overall estimate of the actual location of the fiber core. This improves the accuracy and reliability of fiber core location identification.
  • an exclusion operation is performed to identify connected components in the image, i.e., regions formed by adjacent pixels. This includes removing pixels with probabilities below a certain threshold, retaining only high-probability regions. Identifying connected components effectively eliminates isolated noise points or non-target regions, focusing on the region most likely to be the fiber core.
  • one or more candidate fiber core locations are determined. This process involves analyzing the shape, size, and location of each connected region to determine which area most likely corresponds to the actual fiber core location.
  • the potential fiber core location area is narrowed down to more specific candidate locations, providing an accurate reference for the final fiber core localization.
  • Each candidate fiber core location undergoes further evaluation and processing, including assessing its accuracy and reliability, to ultimately determine the precise location of the fiber core.
  • Detailed evaluation of candidate locations ensures that the selected fiber core location is as accurate as possible.
  • image processing techniques such as edge detection and contrast enhancement, can also be used to improve the visual contrast between the fiber core and other parts of the fiber.
  • Image analysis algorithms such as pattern recognition or machine learning models, are used to identify and locate the precise location of the fiber core.
  • the methods for obtaining the average value of the brightness signal at the aforementioned fiber core location and the average value of the fiber core brightness signal in the solution are as follows:
  • the average of N frames of image data acquired at each laser power is calculated pixel by pixel to obtain the average image set in air.
  • Image set of mean values in solution
  • the average value set of fiber core brightness signal in the solution is extracted from the fiber core location located on the mean image set in the above solution.
  • the average value of the signal at all fiber cores is extracted using the above-mentioned fiber core positioning results, and denoted as .
  • the above reference concentration is the average core brightness signal in the reference solution.
  • the average brightness signal of the fiber core in the aforementioned reference solution is the average brightness signal of all fiber core positions extracted from the located fiber core position on the average image set in the reference solution.
  • the aforementioned average image set in the reference solution is obtained by averaging multiple frames of image data obtained by immersing the endoscope probe in the aforementioned reference concentration of fluorescein sodium solution at maximum laser power.
  • a sodium fluorescein solution of a certain concentration is prepared, and this concentration is denoted as...
  • Sodium fluorescein is a commonly used fluorescent dye that fluoresces under specific laser irradiation.
  • the endoscope probe is immersed in this solution and the laser power is adjusted to maximum to ensure that the acquired brightness signal value is within a preset range.
  • This preset range is set according to experimental or clinical needs to ensure image quality and safety.
  • the preset range can be... .
  • the brightness values in the preset range of the above brightness values are the maximum brightness values that the hardware can achieve, multiplied by a coefficient greater than or equal to 0.5 and less than or equal to 1.
  • the upper limit of the signal value is denoted as The ideal signal value is
  • the acceptable signal value range is Maximum laser power Minimum laser power , This represents the current laser power.
  • the maximum brightness value that can be achieved is determined by hardware characteristics (such as the number of bits in the AD sampling chip).
  • Typical values are as follows: Take 0.9, Take 0.85, Take 0.95.
  • the aforementioned preset curve fitting methods include one or more of the following: straight line fitting, piecewise straight line fitting, quadratic equation, multivariate equation, and log curve.
  • pre-defined curve fitting methods are a key step in establishing a model for determining laser power. These methods model the relationship between image data and laser power, thereby providing optimal laser power settings for the endoscope.
  • curve fitting methods include linear fitting, piecewise linear fitting, quadratic equations, polynomial equations, and logarithmic curves.
  • the brightness signal value of the fiber core containing the current concentration of sodium fluorescein solution is first measured using a confocal endoscope.
  • This brightness signal value reflects the luminescence intensity of sodium fluorescein under specific laser excitation.
  • a previously established laser power determination model is used, which is based on standard data of sodium fluorescein solutions at different concentrations. The model consists of two main parts: a first correspondence (the relationship between sodium fluorescein concentration and brightness) and a second correspondence (the relationship between brightness and laser power).
  • a model is constructed that describes the relationship between core brightness and laser power at the current concentration.
  • This model allows for dynamic adjustment of the laser power based on the actually measured brightness signal value.
  • the constructed model is used to determine the laser power corresponding to the current sodium fluorescein solution concentration when the core brightness signal just reaches the maximum brightness value achievable by the hardware. This automatic adjustment ensures optimal imaging results at different concentrations, preventing excessive or insufficient laser output.
  • the method proposed in this application maximizes image quality and improves diagnostic accuracy and efficiency by precisely adjusting the laser power. It prevents sample damage or equipment failure due to excessive laser power, while also avoiding image information loss due to insufficient brightness.
  • This automatic adjustment method reduces the operator's workload, making the operation of confocal endoscopes simpler and safer, especially in complex or lengthy medical examinations.
  • the construction of a model of the current concentration, laser power, and fiber core brightness to determine the laser power corresponding to the current concentration of sodium fluorescein solution when the fiber core brightness signal just reaches the maximum brightness value achievable by the hardware specifically includes:
  • the brightness signal value adjustment range is determined based on the maximum brightness value that the hardware can achieve, the first adjustment coefficient, and the second adjustment coefficient; the laser power adjustment range is determined based on the maximum laser power and the minimum laser power of the laser.
  • a brightness signal value adjustment range is determined based on a set upper limit for the brightness signal value and the first and second adjustment coefficients. This range serves as a reference for adjusting the laser power in subsequent steps, ensuring that the image brightness signal intensity is within an ideal range, neither too bright nor too dark.
  • the upper limit of the signal value is denoted as The ideal signal value is .
  • the acceptable signal value range is Maximum laser power Minimum laser power , This represents the current laser power.
  • the maximum brightness value that can be achieved is determined by hardware characteristics (such as the number of bits in the AD sampling chip). and It is determined by the laser itself, specifically the maximum and minimum laser power that the laser can provide.
  • the effective image set is obtained by using the maximum laser power as the initial laser power
  • image acquisition is performed using the endoscope's current maximum laser power to obtain a set of buffered images, with a frame rate of [number missing].
  • the cached image set be . These images will be used for subsequent analysis and processing to evaluate image quality at maximum power and provide preliminary data for power adjustment.
  • the initially obtained cached image set is filtered to remove images of low quality or those that do not meet analytical requirements. Valid images are then selected from the cached images. Due to the inherent characteristics of confocal endoscopy—small field of view, high magnification, and short working distance—a portion of the images produced are invalid, such as completely black images when there is no tissue contact, completely white images when contact is too tight, and motion artifacts. Ignoring these images when calculating the laser power to be adjusted results in more accurate values, closer to the expected value.
  • the filtered image set contains more representative and analytically valuable images for further brightness analysis.
  • the current concentration of the sodium fluorescein solution corresponding to the current laser power is calculated based on the average of the larger brightness values.
  • an adjusted target laser power is calculated based on the larger brightness values extracted from the selected image set and a previously set brightness signal value adjustment range. This new laser power keeps the image brightness within an ideal range, improving the image's usability and diagnostic value.
  • the larger value can be obtained by extracting all fiber core brightness signal values from the image, calculating the histogram, and taking a certain percentage of values accumulated from high to low in the histogram as the larger brightness value, denoted as . A certain percentage can be taken as 5%. The average of the larger brightness values for each frame in the selected image set is then calculated. .
  • the laser power determination model is constructed based on the concentration of sodium fluorescein solution, the brightness signal value of the fiber core, and the laser power.
  • the laser power at which the maximum brightness value that the hardware can just reach is calculated based on the current concentration of sodium fluorescein solution.
  • the laser power needs to be increased. If the laser power has not reached its maximum value, it can be increased. First, the current concentration of the sodium fluorescein solution needs to be determined. Then, using either of the laser power determination models mentioned above or the current concentration of the sodium fluorescein solution, the laser power required to reach the maximum brightness signal can be calculated.
  • the laser power at which the brightness signal reaches its maximum value is determined as the adjusted target laser power.
  • the current concentration of the sodium fluorescein solution corresponding to the current laser power is calculated based on the average of the larger brightness values mentioned above, including:
  • the current concentration of sodium fluorescein solution corresponding to the current laser power is calculated.
  • the concentration of sodium fluorescein was calculated using the established model. Just reached the maximum value of the brightness signal laser power at time : ;
  • the methods described above for obtaining a valid set of images include:
  • the cached images in the cached image set are filtered to obtain the effective image set.
  • effective image screening can be performed based on the following methods:
  • Filtering operations can specifically include:
  • the method for calculating the larger brightness value is as follows: extract all fiber core brightness signal values from the image, calculate the histogram, and take the values that account for the largest percentage from high to low in the histogram as the larger brightness value, denoted as . .
  • the percentages from highest to lowest are as follows:
  • the proportion of values between 9 and 8 is less than 5%, while the proportion of values between 9 and 7 is greater than 5%. Therefore, the largest brightness signal value with a cumulative proportion of 5% from high to low is 7.
  • the method for calculating larger gradient values is as follows: convolve the image using the following four 3x3 operators respectively, and obtain... , , , .make
  • the histogram is used to select the values that represent the largest gradient, with the cumulative percentage from high to low in the histogram being 5%.
  • the operators used can be any four of those shown in Figures 3 to 6.
  • the operators used can also be symmetrically flipped along the row, column, and diagonal where 0 is located.
  • the laser power determination model construction method provided in this application through precise control of laser power, automatically adjusts the laser power according to different environments (air and fluorescent dye solution) to ensure optimal image quality under various conditions.
  • This method can effectively cope with individual differences and variations in fluorescent dye concentration, ensuring moderate brightness and contrast in the image, thus making it easier to identify details and lesions.
  • Another embodiment of this model construction method implements an automatic laser power adjustment method, which can automatically adjust to the optimal laser output, thereby improving image quality, simplifying the operation process, and reducing the burden on doctors during operation. Therefore, the laser power determination model construction method and the automatic laser power adjustment method provided in the various embodiments of this application have industrial applicability.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Endoscopes (AREA)

Abstract

本申请公开了一种激光功率确定模型构建方法及激光功率自动调节方法，涉及内窥镜成像领域，该激光功率确定模型构建方法包括基于空气环境图像数据集、溶液环境图像数据集、激光功率集和预设曲线拟合方法进行数据拟合，以获取激光功率确定模型。还公开了包括通过该激光功率确定模型实现的激光功率自动调节方法。

Description

激光功率确定模型构建方法及激光功率自动调节方法 
本申请要求申请人于2024 年06月28日提交的中国专利局申请号为202410852246.5、发明名称为“一种激光功率确定模型构建方法、调节方法及存储介质”的中国专利申请及2024年06月28日提交的中国专利局申请号为202410852114.2、发明名称为“一种共聚焦内窥镜激光功率自动调节方法”的优先权，其全部内容通过引用结合在本申请，本申请所引用并包含的其公开的所有相关申请是本申请不可分割的一部分。
技术领域
本申请涉及内窥镜成像领域，更具体地说，本申请涉及一种激光功率确定模型构建方法及激光功率自动调节方法。
 背景技术
共聚焦内窥镜是一种可以借助胃镜、结肠镜等通道伸入人体，获取局部组织学图像来实现微小病灶、胃肠道病变及早期胃肠道癌变的精准诊断的医疗设备。共聚焦内窥镜的基本原理是是用激光照亮组织，然后检测从组织反射的荧光。荧光染料浓度一定的条件下，不同的激光功率会得到质量不同的图像，过高或过低的激光功率会导致图像过亮或过暗，致使所成图像无法区分细节，缺乏足够的对比度，即降低了图像质量，因而合适的激光功率对于成像质量至关重要。
共聚焦内窥镜应用于临床检查时，给予被检查者相同剂量的荧光染料，也会因为被检查个体差异，导致组织中的荧光染料浓度不同。为了得到质量好的图像，通常需要调节激光功率到合适水平。如果让医师手动调节激光功率，会增加操作复杂度和工作量，而且事先要对医师进行培训，使其能够根据当时的图像合理调节激光功率，这本身并非易事。因此，如何简化共聚焦内窥镜的激光功率调节工作，成为了亟需解决的问题。
 发明内容
在发明内容部分中引入了一系列简化形式的概念，这将在具体实施方式部分中进一步详细说明。本申请的发明内容部分并不意味着要试图限定出所要求保护的技术方案的关键特征和必要技术特征，更不意味着试图确定所要求保护的技术方案的保护范围。
第一方面，本申请提出一种激光功率确定模型构建方法，上述方法包括：
选定K个不同激光功率，以构建激光功率集；
在每个激光功率下各自获取上述内窥镜探头在空气中的N帧图像数据，以构建空气环境图像数据集；
在每个激光功率下各自获取上述内窥镜探头浸入参考浓度的荧光素钠溶液中的N帧图像数据，以构建溶液环境图像数据集；
基于上述空气环境图像数据集、上述溶液环境图像数据集、激光功率集和预设曲线拟合方法进行数据拟合，以获取上述激光功率确定模型，其中，上述激光功率确定模型包括第一对应关系和第二对应关系，上述第一对应关系为空气环境图像数据与上述K个不同激光功率的对应关系，上述第二对应关系为溶液环境图像数据与上述K个不同激光功率的对应关系。
在一种可行的实施方式中，上述基于上述空气环境图像数据集、上述溶液环境图像数据集、激光功率集和预设曲线拟合方法进行数据拟合，以获取上述激光功率确定模型，包括：
基于上述空气环境图像数据集在每个激光功率下所有纤芯位置的亮度信号值的平均值，以构建空气中纤芯亮度信号平均值集 ；
基于上述溶液环境图像数据集在每个激光功率下所有纤芯位置的亮度信号值的平均值，以构建溶液中纤芯亮度信号平均值集 ；
根据上述空气中纤芯亮度信号平均值集、上述激光功率集和上述预设曲线拟合方法进行数据拟合，以构建上述第一对应关系；
根据上述溶液中纤芯亮度信号平均值集、上述激光功率集和上述预设曲线拟合方法进行数据拟合，以构建上述第二对应关系；
根据上述第一对应关系和上述第二对应关系，创建上述激光功率确定模型。
在一种可行的实施方式中，还包括：
利用上述空气环境图像数据集或上述溶液环境图像数据集对上述纤芯进行定位，以获得纤芯位置。
在一种可行的实施方式中，上述纤芯位置的亮度信号值的平均值和上述溶液中纤芯亮度信号平均值的获取方法具体为：
分别在上述空气环境图像数据集和上述溶液环境图像数据集中，对上述每个激光功率下采集的N帧图像数据按像素取平均，以获得空气中均值图像集 和溶液中均值图像集 ；
在上述空气中均值图像集上的定位的纤芯位置上提取上述空气中纤芯亮度信号平均值集；
在上述溶液中均值图像集上的定位的纤芯位置上提取溶液中纤芯亮度信号平均值集。
在一种可行的实施方式中，上述参考浓度为参考溶液中纤芯亮度信号平均值 在亮度值预设区间内对应的浓度，上述参考溶液中纤芯亮度信号平均值是在参考溶液中均值图像集上在定位的纤芯位置上提取所有纤芯位置的亮度信号的平均值，上述参考溶液中均值图像集是内窥镜探头以最大激光功率浸入上述参考浓度的荧光素钠溶液中获的取多帧图像数据按像素取平均获取得到的。
在一种可行的实施方式中，上述亮度值预设区间中的亮度值为硬件所能达到的最大亮度值乘以大于等于0.5、小于等于1的系数。
第二方面、本申请还提出一种共聚焦内窥镜激光功率自动调节方法，包括：
基于当前浓度的荧光素钠溶液获得的纤芯亮度信号值、通过第一方面任一项上述的激光功率确定模型构建方法获得的参考荧光素钠溶液浓度、第一对应关系和第二对应关系；
构建当前浓度、激光功率及纤芯亮度的模型，从而确定当前浓度的荧光素钠溶液在纤芯亮度信号刚达到硬件所能达到的最大亮度值时对应的激光功率。
在一种可行的实施方式中，上述构建当前浓度、激光功率及纤芯亮度的模型，从而确定当前浓度的荧光素钠溶液在纤芯亮度信号刚达到硬件所能达到的最大亮度值时对应的激光功率具体包括：
根据硬件所能达到的最大亮度值上限、第一调节系数和第二调节系数确定亮度信号值调节范围，根据激光器的最大激光功率及最小激光功率确定激光功率调节范围；
以最大激光功率为初始激光功率，获取有效图像集； 
对上述有效图像集中的每帧图像获取亮度较大值；
基于上述亮度较大值的平均值计算上述当前激光功率对应的荧光素钠溶液当前浓度；
根据荧光素钠溶液浓度、纤芯亮度信号值及激光功率构建的激光功率确定模型和上述荧光素钠溶液当前浓度计算刚达到硬件所能达到的最大亮度值时的激光功率；
将上述达到亮度信号最大值时的激光功率确定为上述调节后的目标激光功率。
在一种可行的实施方式中，基于上述亮度较大值的平均值计算上述当前激光功率对应的荧光素钠溶液的当前浓度，包括：
根据上述亮度较大值的平均值、空气中纤芯亮度信号集和激光功率集拟合构建的第一对应关系、溶液中纤芯亮度信号集和激光功率集拟合构建的第二对应关系及上述第二对应关系中所使用的荧光素钠的参考浓度，计算上述当前激光功率对应的荧光素钠溶液的当前浓度。
在一种可行的实施方式中，上述获取有效图像集的方法包括：
获取缓存图像集；
对上述缓存图像集中的每帧图像计算亮度较大值和梯度较大值；
基于上述亮度较大值和上述梯度较大值对上述缓存图像集中的缓存图像进行筛选，以获取上述有效图像集。
综上，本申请实施例的激光功率确定模型构建方法包括：选定K个不同激光功率，以构建激光功率集；在每个激光功率下各自获取上述内窥镜探头在空气中的N帧图像数据，以构建空气环境图像数据集；在每个激光功率下各自获取上述内窥镜探头浸入参考浓度的荧光素钠溶液中的N帧图像数据，以构建溶液环境图像数据集；基于上述空气环境图像数据集、上述溶液环境图像数据集、激光功率集和预设曲线拟合方法进行数据拟合，以获取上述激光功率确定模型，其中，上述激光功率确定模型包括第一对应关系和第二对应关系，上述第一对应关系为空气环境图像数据与上述K个不同激光功率的对应关系，上述第二对应关系为溶液环境图像数据与上述K个不同激光功率的对应关系。本申请实施例提出的方法，通过精确控制激光功率，根据不同的环境（空气和荧光染料溶液）自动调整激光功率，确保在各种条件下都能获得最佳的图像质量。这种方法能够有效应对个体差异和荧光染料浓度的变化，保证图像的明暗和对比度适中，从而更容易识别细节和病变。自动化的激光功率调节减少了医师在操作共聚焦内窥镜时的负担。医师不需要深入了解如何根据图像调整激光功率，降低了对医师的培训要求，同时减少了操作过程中的人为错误。通过本申请提出的模型，可以应用本模型实现自动化调节激光功率，可以缩短设备的调整时间，加快诊断流程。本申请中的模型可以适应不同类型的共聚焦内窥镜和多样的临床环境。模型的通用性使其可以轻松地整合到不同的硬件和软件配置中，提供更广泛的应用范围。本申请实施例通过建立一个基于实际图像数据的激光功率确定模型，实现了共聚焦内窥镜在临床使用中的自动激光功率调节，有效提升了图像质量和操作效率，同时降低了医师操作的复杂性和培训要求，为临床提供了一种更加高效、精确的诊断工具。
本申请实施例的共聚焦内窥镜激光功率自动调节方法包括：根据硬件所能达到的最大亮度值上限、第一调节系数和第二调节系数确定亮度信号值调节范围，根据激光器的最大激光功率及最大激光功率确定激光功率调节范围；以最大激光功率为初始激光功率，获取有效图像集；对上述有效图像集中的每帧图像获取亮度较大值；基于上述亮度较大值的平均值和上述亮度信号值调节范围，确定调节后的目标激光功率。本申请实施例提出的方法，实现了激光功率的自动调节，使得内窥镜在进行组织成像时，可以自动调整到最佳的激光输出，从而提高图像质量，简化操作流程，并减少医生在操作中的负担。
本申请提出的激光功率确定模型构建方法及激光功率自动调节方法，本申请的其它优点、目标和特征将部分通过下面的说明体现，部分还将通过对本申请的研究和实践而为本领域的技术人员所理解。
附图说明
通过阅读下文优选实施方式的详细描述，各种其他的优点和益处对于本领域普通技术人员将变得清楚明了。附图仅用于示出优选实施方式的目的，而并不认为是对本说明书的限制。而且在整个附图中，用相同的参考符号表示相同的部件。在附图中：
图1为本申请实施例提供的一种激光功率确定模型构建方法流程性示意图；
图2为本申请实施例提供的一种共聚焦内窥镜激光功率自动调节方法流程性示意图；
图3为本申请实施例提供的一种算子结构示意图；
图4为本申请实施例提供的另一种算子结构示意图；
图5为本申请实施例提供的又一种算子结构示意图；
图6为本申请实施例提供的再一种算子结构示意图。
具体实施方式
本申请的说明书和权利要求书及上述附图中的术语“第一”、“第二”、“第三”、“第四”等（如果存在）是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换，以便这里描述的实施例能够以除了在这里图示或描述的内容以外的顺序实施。此外，术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，例如，包含了一系列步骤或单元的过程、方法、系统、产品或设备不必限于清楚地列出的那些步骤或单元，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或单元。下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本申请一部分实施例，而不是全部的实施例。
请参阅图1，为本申请实施例提供的一种激光功率确定模型构建方法流程性示意图，具体可以包括：
S110、选定K个不同激光功率，以构建激光功率集；
示例性的，选定 个不同激光功率，以构建激光功率集 。
S120、在每个激光功率下各自获取上述内窥镜探头在空气中的N帧图像数据，以构建空气环境图像数据集；
示例性的，内窥镜探头在 个不同的激光功率设置下，分别在空气环境中捕获N帧图像数据，以形成空气环境图像数据集 。
S130、在每个激光功率下各自获取上述内窥镜探头浸入参考浓度的荧光素钠溶液中的N帧图像数据，以构建溶液环境图像数据集；
示例性的，与步骤 S120 类似，但是此时探头浸入参考浓度的荧光素钠溶液中。同样，在K个不同的激光功率 设置下捕获N帧图像，建立一个溶液环境图像数据集 。
S140、基于上述空气环境图像数据集、上述溶液环境图像数据集、激光功率集和预设曲线拟合方法进行数据拟合，以获取上述激光功率确定模型，其中，上述激光功率确定模型包括第一对应关系和第二对应关系，上述第一对应关系为空气环境图像数据与上述K个不同激光功率的对应关系，上述第二对应关系为溶液环境图像数据与上述K个不同激光功率的对应关系。
示例性的，基于前两个数据集（空气环境和溶液环境图像数据集）以及不同的激光功率，使用预设的曲线拟合方法进行数据分析和拟合。通过这种方法，建立一个激光功率确定模型，该模型包括两种对应关系：
第一对应关系f 1：空气环境图像数据与不同激光功率之间的关系。
第二对应关系f 2：溶液环境图像数据与不同激光功率之间的关系。
综上，本申请实施例提出的方法，通过精确控制激光功率，根据不同的环境（空气和荧光染料溶液）自动调整激光功率，确保在各种条件下都能获得最佳的图像质量。这种方法能够有效应对个体差异和荧光染料浓度的变化，保证图像的明暗和对比度适中，从而更容易识别细节和病变。自动化的激光功率调节减少了医师在操作共聚焦内窥镜时的负担。医师不需要深入了解如何根据图像调整激光功率，降低了对医师的培训要求，同时减少了操作过程中的人为错误。通过本申请提出的模型，可以应用本模型实现自动化调节激光功率，可以缩短设备的调整时间，加快诊断流程。本申请中的模型可以适应不同类型的共聚焦内窥镜和多样的临床环境。模型的通用性使其可以轻松地整合到不同的硬件和软件配置中，提供更广泛的应用范围。本申请实施例通过建立一个基于实际图像数据的激光功率确定模型，实现了共聚焦内窥镜在临床使用中的自动激光功率调节，有效提升了图像质量和操作效率，同时降低了医师操作的复杂性和培训要求，为临床提供了一种更加高效、精确的诊断工具。
在一些示例中，上述基于上述空气环境图像数据集、上述溶液环境图像数据集、激光功率集和预设曲线拟合方法进行数据拟合，以获取上述激光功率确定模型，包括：
基于上述空气环境图像数据集在每个激光功率下所有纤芯位置的亮度信号值的平均值，以构建空气中纤芯亮度信号平均值集 ；
基于上述溶液环境图像数据集在每个激光功率下所有纤芯位置的亮度信号值的平均值，以构建溶液中纤芯亮度信号平均值集 ；
根据上述空气中纤芯亮度信号平均值集、上述激光功率集和上述预设曲线拟合方法进行数据拟合，以构建上述第一对应关系；
根据上述溶液中纤芯亮度信号平均值集、上述激光功率集和上述预设曲线拟合方法进行数据拟合，以构建上述第二对应关系；
根据上述第一对应关系和上述第二对应关系，创建上述激光功率确定模型。
示例性的，对上述空气环境图像数据集根据像素区取均值得到所有激光功率下空气中均值图像集 ，并在 上提取所有纤芯处亮度信号值的平均值，记为空气中纤芯亮度信号平均值集 ，其中 。
溶液环境图像数据集取均值得到所有激光功率下荧光素钠溶液中均值图像集 ；由 提取所有纤芯处亮度信号值的平均值，形成溶液中纤芯亮度信号平均值集 ，其中 。
使用激光功率集 和空气中纤芯亮度信号平均值集 进行拟合，得到 ，f 1为第一对应关系。
使用激光功率集 和溶液中纤芯亮度信号平均值集 进行拟合，得到 ，f 2为第二对应关系。
激光功率确定模型是包括第一对应关系和第二对应关系的模型。
在一些示例中，还包括：
对上述纤芯进行定位，以获得纤芯位置。
在一些示例中，利用上述空气环境图像数据集或上述溶液环境图像数据集对上述纤芯进行定位。
示例性的，可以通过空气环境图像数据集或上述溶液环境图像数据集，获取有效图像集。然后对这些有效图像按像素计算中值。有效图像集是指通过某种筛选机制从大量成像数据中选出的清晰图像集合。这种筛选旨在确保所选图像的质量，排除模糊或损坏的图像。计算中值图像的目的是减少或消除随机噪声、偶然误差或其他一次性异常。中值滤波是一种非线性过程，对维持图像边缘和结构细节方面特别有效。基于有效图像集，采用统计方法计算纤芯的位置累计概率。对每一帧图像中纤芯可能的位置进行评分或概率评估，然后将这些评估值累积起来，形成对纤芯实际位置的整体估计。从而提高纤芯位置识别的准确性和可靠性。根据累积概率，执行排除操作，从而辨识出图像中的连通域，即相邻像素点形成的区域。包括去除那些概率低于某一阈值的像素点，仅保留那些高概率区域。识别连通域，可以有效地排除孤立的噪声点或非目标区域，聚焦于最有可能是纤芯的区域。基于上一步骤获得的连通域信息，确定一个或多个候选纤芯位置。包括对每个连通域的形状、大小、位置等特征进行分析，以判断哪个区域最可能对应于真实的纤芯位置。将潜在的纤芯位置区域缩小到更具体的候选位置，为最终的纤芯定位提供准确的参考。对每个候选纤芯位置进行进一步的评估和处理，包括对其准确性、可靠性进行评估，最终确定纤芯的准确位置。通过对候选位置的详细评估，确保选定的纤芯位置尽可能准确无误。需要说明的是，也可以利用其他的图像处理技术，如边缘检测、对比度增强等，以提高纤芯与光纤其他部分的可视对比度。通过图像分析算法，如模式识别或机器学习模型，识别并定位纤芯的准确位置。
在一些示例中，上述纤芯位置的亮度信号值的平均值和上述溶液中纤芯亮度信号平均值的获取方法具体为：
分别在上述空气环境图像数据集和上述溶液环境图像数据集中，对上述每个激光功率下采集的N帧图像数据按像素取平均，以获得空气中均值图像集 和溶液中均值图像集 ；
在上述空气中均值图像集上的定位的纤芯位置上提取上述空气中纤芯亮度信号平均值集；
在上述溶液中均值图像集上的定位的纤芯位置上提取溶液中纤芯亮度信号平均值集。
示例性的，对每个激光功率下采集的 帧数据按像素取均值。得到所有激光功率下空气中均值图像集 和荧光素钠溶液中均值图像集 ；
利用 ，使用纤芯定位方法定位所有纤芯。
在 上用上述定位纤芯的结果提取所有纤芯处信号值的平均值，记为 。所有激光功率下空气中纤芯处信号平均值集 ；由 提取所有纤芯处信号值的平均值 ，所有激光功率下空气中纤芯处信号平均值集 。
在一些示例中，上述参考浓度为参考溶液中纤芯亮度信号平均值 在亮度值预设区间内对应的浓度，上述参考溶液中纤芯亮度信号平均值是在参考溶液中均值图像集上在定位的纤芯位置上提取所有纤芯位置的亮度信号的平均值，上述参考溶液中均值图像集是内窥镜探头以最大激光功率浸入上述参考浓度的荧光素钠溶液中获的取多帧图像数据按像素取平均获取得到的。
示例性的，配置一定浓度的荧光素钠溶液记该浓度为 。荧光素钠是一种常用的荧光染料，可以在特定激光照射下发出荧光。在这个步骤中，内窥镜探头被浸入这种溶液中，并调整到最大激光功率，确保在这种条件下获取的亮度信号值在一个预设的区间内。这个预设区间是根据实验或临床需求设定的，以确保图像的质量和安全性。预设区间可以为 。
在一些示例中，上述亮度值预设区间中的亮度值为硬件所能达到的最大亮度值乘以大于等于0.5、小于等于1的系数。
示例性的，记信号值上限为 ，理想信号值大小为 ,可接受信号值大小区间为 ，最大激光功率 ，最小激光功率 ， 为当前激光功率。
是由硬件特性（比如AD采样芯片位数）决定的亮度值能达到的最大值。
和 是由激光器本身决定的，该激光器能提供的激光功率最大值和最小值。
， ， ，
典型取值如， 取0.9， 取0.85， 取0.95。
在一些示例中，上述预设曲线拟合方法包括直线拟合、分段直线拟合、二次方程、多次方程和log曲线中一种或多种。
示例性的，预设曲线拟合方法是用来建立激光功率确定模型的关键步骤。这些方法通过分析图像数据与激光功率之间的关系来建模，从而能够为内窥镜提供最佳的激光功率设置。这些曲线拟合方法包括直线拟合、分段直线拟合、二次方程、多项方程和对数曲线等。通过选择合适的曲线拟合方法，可以根据实验或临床数据的具体表现来决定使用哪种数学模型，从而准确地建立激光功率与图像质量之间的关系。
第二方面、如图2所示，本申请还提出一种共聚焦内窥镜激光功率自动调节方法，包括：
S210、基于当前浓度的荧光素钠溶液获得的纤芯亮度信号值、通过第一方面任一项上述的激光功率确定模型构建方法获得的参考荧光素钠溶液浓度、第一对应关系和第二对应关系；
示例性的，首先通过共聚焦内窥镜测量含有当前浓度荧光素钠溶液的纤芯亮度信号值。这个亮度信号值反映了荧光素钠在特定激光激发下的发光强度。使用先前建立的激光功率确定模型，这个模型基于不同浓度的荧光素钠溶液的标准数据。模型包括两个主要部分：第一对应关系（荧光素钠浓度与亮度之间的关系）和第二对应关系（亮度与激光功率之间的关系）。
S220、构建当前浓度、激光功率及纤芯亮度的模型，从而确定当前浓度的荧光素钠溶液在纤芯亮度信号刚达到硬件所能达到的最大亮度值时对应的激光功率。
示例性的，根据从S210步骤获得的实时数据（纤芯亮度）和参考模型（荧光素钠浓度对应的亮度和激光功率），构建一个能够描述当前浓度下纤芯亮度与激光功率关系的模型。这个模型使得可以根据实际测得的亮度信号值动态调整激光功率。利用构建的模型确定在当前荧光素钠溶液浓度下，当纤芯亮度信号刚好达到硬件所能达到的最大亮度值时对应的激光功率。这种自动调节确保在不同浓度下都能获得最佳的成像效果，防止过度或不足的激光输出。
综上，本申请实施例提出的方法通过精确调整激光功率，可以最大化成像质量，提高诊断的准确性和效率。防止因激光功率过高而导致的样本损伤或设备损坏，同时避免由于亮度不足导致的图像信息缺失。这种自动调节方法减少了操作者的负担，使得共聚焦内窥镜的操作更加简便和安全，特别是在复杂或长时间的医学检查中。
在一些示例中，所述构建当前浓度、激光功率及纤芯亮度的模型，从而确定当前浓度的荧光素钠溶液在纤芯亮度信号刚达到硬件所能达到的最大亮度值时对应的激光功率具体包括：
根据硬件所能达到的最大亮度值上限、第一调节系数和第二调节系数确定亮度信号值调节范围，根据激光器的最大激光功率及最小激光功率确定激光功率调节范围；
示例性的，根据设定的亮度信号值上限以及第一和第二调节系数来确定一个亮度信号值调节范围。这个范围用于后续步骤中调整激光功率的参考依据，确保图像亮度信号强度在一个理想的区间内，既不过亮也不过暗。
具体的，记信号值上限为 ，理想信号值大小为为 ,可接受信号值大小区间为 ，最大激光功率 ，最小激光功率 ， 为当前激光功率。
是由硬件特性（比如AD采样芯片位数）决定的亮度值能达到的最大值。 和 是由激光器本身决定的，该激光器能提供的激光功率最大值和最小值。
以最大激光功率为初始激光功率，获取有效图像集；
示例性的，以内窥镜当前的最大激光功率进行图像采集，获得一组缓存图像集，图像帧数为 ，记缓存的图像集为 。这些图像将用于后续的分析和处理，目的是评估最大功率下的图像质量，并为功率调整提供初步数据。对初始获得的缓存图像集进行筛选操作，目的是剔除质量不高或不符合分析要求的图像。从缓存图像中筛选有效图像。共聚焦内窥镜因为其视野小、放大倍率高、工作距离短的固有特性，其所成图像中有一部分是无效图像，比如没有接触到组织时的全黑图像、接触过紧时的全白图像、运动伪影等。在计算要调节的激光功率时，不考虑这部分图像会使得计算值更准确，即更接近预期值。筛选后的图像集（筛选图像集）包含了更具代表性和分析价值的图像，用于进一步的亮度分析。
对所述有效图像集中的每帧图像获取亮度较大值；
示例性的，在有效图像集中进一步分析，从每帧图像中提取亮度较大的值。这一步骤关键在于评估在最大功率下图像的亮度水平，为调整激光功率到更适宜的水平提供依据。
基于所述亮度较大值的平均值计算所述当前激光功率对应的荧光素钠溶液当前浓度；
示例性的，基于从筛选图像集中提取的亮度较大值及先前设定的亮度信号值调节范围，计算出一个调节后的目标激光功率。这个新的激光功率使图像的亮度保持在理想的范围内，提升图像的可用性和诊断价值。
其中，较大值可以从图像上提取所有纤芯亮度信号值，统计直方图，取直方图中从高到低累计一定占比的值作为亮度较大值，记为  ，一定占比可以取5%。对筛选图像集中的每帧图像亮度较大值取平均值，得到 。
根据荧光素钠溶液浓度、纤芯亮度信号值及激光功率构建的激光功率确定模型和所述荧光素钠溶液当前浓度计算刚达到硬件所能达到的最大亮度值时的激光功率；
示例性的，如果 且 ，亮度偏低需要调大激光功率，激光功率没有达到最大值，可以调大。首先需要确定当前的荧光素钠溶液浓度。通过一方面任一项上述的激光功率确定模型和上述荧光素钠溶液当前浓度计算达到亮度信号最大值时的激光功率。
将所述达到亮度信号最大值时的激光功率确定为所述调节后的目标激光功率。
在一些示例中，基于上述亮度较大值的平均值计算上述当前激光功率对应的荧光素钠溶液的当前浓度，包括：
根据上述亮度较大值的平均值、空气中纤芯亮度信号集和激光功率集拟合构建的第一对应关系、溶液中纤芯亮度信号集和激光功率集拟合构建的第二对应关系及上述第二对应关系中所使用的荧光素钠的参考浓度，计算上述当前激光功率对应的荧光素钠溶液的当前浓度。
示例性的，计算 对应的荧光素钠浓度 ： ； 
通过建立的模型计算荧光素钠浓度 刚达到亮度信号最大值 时的激光功率 ： ；
调节激光功率到 。
在一些示例中，上述获取有效图像集的方法包括：
获取缓存图像集；
对上述缓存图像集中的每帧图像计算亮度较大值和梯度较大值；
基于上述亮度较大值和上述梯度较大值对上述缓存图像集中的缓存图像进行筛选，以获取上述有效图像集。
示例性的，具体的，在进行有效图像筛选时可以基于下述方式进行：
筛选操作具体可以包括：
（1）对缓存的每帧图像，计算亮度较大值和梯度较大值。
计算亮度较大值的方法为：从图像上提取所有纤芯亮度信号值，统计直方图，取直方图中从高到低累计占比5%的值作为亮度较大值，记为 。
举例说明，假如统计直方图结果如下：
亮度信号值 0 1 2 3 4 5 6 7 8 9
数量 3 12 16 18 15 13 11 8 2 2
总数为3+5+6+……+2=100.从高到低依次占比为：
9: 2/100 = 0.02
9~8: (2+2)/100 = 0.04
9~7: (2+2+8)/100 = 0.12
9~8的占比小于5%，而9~7的占比大于5%，因而从高到低累计占比5%的较大亮度信号值是7.
计算梯度较大值的方法为：对图像分别使用以下4个3x3大小的算子进行卷积，得到 ， ， ， 。令
， 分别为像素行坐标和列坐标。其中round函数为就近取整，max函数位取最大值，abs函数为取绝对值。统计 的直方图，取直方图中从高到低累计占比5%的值作为梯度较大值。 
使用的算子可以用图3至6中任意4种。使用的算子还可以按0所在行、列、对角线对称翻转。
（2）按亮度较大值和梯度较大值对缓存的图像进行筛选。
将所有缓存图像的亮度较大值 按从大到小顺序排序，取前一半缓存图像，记为 。将所有缓存图像的梯度较大值 按从大到小顺序排序，取前一半缓存图像，记为 。筛选出的图像是 与 的并集，即 。以上，以上实施例仅用以说明本申请的技术方案，而非对其限制；尽管参照前述实施例对本申请进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本申请各实施例技术方案的精神和范围。
工业实用性
本申请实施例提供的激光功率确定模型构建方法，通过精确控制激光功率，根据不同的环境（空气和荧光染料溶液）自动调整激光功率，确保在各种条件下都能获得最佳的图像质量。这种方法能够有效应对个体差异和荧光染料浓度的变化，保证图像的明暗和对比度适中，从而更容易识别细节和病变。另一实施例提供的该模型构建方法实现的激光功率自动调节方法，可以自动调整到最佳的激光输出，从而提高图像质量，简化操作流程，并减少医生在操作中的负担。因此，本申请各实施例提供的激光功率确定模型构建方法及激光功率自动调节方法具有工业实用性。

Claims (10)

  1. 一种激光功率确定模型构建方法，其特征在于，包括：
    选定K个不同激光功率，以构建激光功率集；
    在每个激光功率下各自获取所述内窥镜探头在空气中的N帧图像数据，以构建空气环境图像数据集；
    在每个激光功率下各自获取内窥镜探头浸入参考浓度的荧光素钠溶液中的N帧图像数据，以构建溶液环境图像数据集；
    基于所述空气环境图像数据集、所述溶液环境图像数据集、激光功率集和预设曲线拟合方法进行数据拟合，以获取所述激光功率确定模型，其中，所述激光功率确定模型包括第一对应关系和第二对应关系，所述第一对应关系为空气环境图像数据与所述K个不同激光功率的对应关系，所述第二对应关系为溶液环境图像数据与所述K个不同激光功率的对应关系。
  2. 根据权利要求1所述的激光功率确定模型构建方法，其特征在于，所述基于所述空气环境图像数据集、所述溶液环境图像数据集、激光功率集和预设曲线拟合方法进行数据拟合，以获取所述激光功率确定模型，包括：
    基于所述空气环境图像数据集在每个激光功率下所有纤芯位置的亮度信号值的平均值，以构建空气中纤芯亮度信号平均值集 ；
    基于所述溶液环境图像数据集在每个激光功率下所有纤芯位置的亮度信号值的平均值，以构建溶液中纤芯亮度信号平均值集 ；
    根据所述空气中纤芯亮度信号平均值集、所述激光功率集和所述预设曲线拟合方法进行数据拟合，以构建所述第一对应关系；
    根据所述溶液中纤芯亮度信号平均值集、所述激光功率集和所述预设曲线拟合方法进行数据拟合，以构建所述第二对应关系；
    根据所述第一对应关系和所述第二对应关系，创建所述激光功率确定模型。
  3. 根据权利要求2所述的激光功率确定模型构建方法，其特征在于，还包括：
    利用所述空气环境图像数据集或所述溶液环境图像数据集对所述纤芯进行定位，以获得纤芯位置。
  4. 根据权利要求3所述的激光功率确定模型构建方法，其特征在于，所述纤芯位置的亮度信号值的平均值和所述溶液中纤芯亮度信号平均值的获取方法具体为：
    分别在所述空气环境图像数据集和所述溶液环境图像数据集中，对所述每个激光功率下采集的N帧图像数据按像素取平均，以获得空气中均值图像集 和溶液中均值图像集 ；
    在所述空气中均值图像集上的定位的纤芯位置上提取所述空气中纤芯亮度信号平均值集；
    在所述溶液中均值图像集上的定位的纤芯位置上提取溶液中纤芯亮度信号平均值集。
  5. 根据权利要求4所述的激光功率确定模型构建方法，其特征在于，所述参考浓度为参考溶液中纤芯亮度信号平均值 在亮度值预设区间内对应的浓度，所述参考溶液中纤芯亮度信号平均值是在参考溶液中均值图像集上在定位的纤芯位置上提取所有纤芯位置的亮度信号的平均值，所述参考溶液中均值图像集是内窥镜探头以最大激光功率浸入所述参考浓度的荧光素钠溶液中获的取多帧图像数据按像素取平均获取得到的。
  6. 根据权利要求5所述的激光功率确定模型构建方法，其特征在于，所述亮度值预设区间中的亮度值为硬件所能达到的最大亮度值乘以大于等于0.5且小于等于1的系数。
  7. 一种共聚焦内窥镜激光功率自动调节方法，其特征在于，
    基于当前浓度的荧光素钠溶液获得的纤芯亮度信号值、通过权利要求1-6任一项所述的激光功率确定模型构建方法获得的参考荧光素钠溶液浓度、第一对应关系和第二对应关系，
    构建当前浓度、激光功率及纤芯亮度的模型，从而确定当前浓度的荧光素钠溶液在纤芯亮度信号刚达到硬件所能达到的最大亮度值时对应的激光功率。
  8. 根据权利要求7所述的共聚焦内窥镜激光功率自动调节方法，其特征在于，所述构建当前浓度、激光功率及纤芯亮度的模型，从而确定当前浓度的荧光素钠溶液在纤芯亮度信号刚达到硬件所能达到的最大亮度值时对应的激光功率具体包括：
    根据硬件所能达到的最大亮度值上限、第一调节系数和第二调节系数确定亮度信号值调节范围，根据激光器的最大激光功率及最小激光功率确定激光功率调节范围；
    以最大激光功率为初始激光功率，获取有效图像集； 
    对所述有效图像集中的每帧图像获取亮度较大值；
    基于所述亮度较大值的平均值计算所述当前激光功率对应的荧光素钠溶液当前浓度；
    根据荧光素钠溶液浓度、纤芯亮度信号值及激光功率构建的激光功率确定模型和所述荧光素钠溶液当前浓度计算刚达到硬件所能达到的最大亮度值时的激光功率；
    将所述达到亮度信号最大值时的激光功率确定为所述调节后的目标激光功率。
  9. 根据权利要求8所述的共聚焦内窥镜激光功率自动调节方法，其特征在于，基于所述亮度较大值的平均值计算所述当前激光功率对应的荧光素钠溶液的当前浓度，包括：
    根据所述亮度较大值的平均值、空气中纤芯亮度信号集和激光功率集拟合构建的第一对应关系、溶液中纤芯亮度信号集和激光功率集拟合构建的第二对应关系及所述第二对应关系中所使用的荧光素钠的参考浓度，计算所述当前激光功率对应的荧光素钠溶液的当前浓度。
  10. 根据权利要求8所述的共聚焦内窥镜激光功率自动调节方法，其特征在于，所述获取有效图像集的方法包括：
    获取缓存图像集；
    对所述缓存图像集中的每帧图像计算亮度较大值和梯度较大值；
    基于所述亮度较大值和所述梯度较大值对所述缓存图像集中的缓存图像进行筛选，以获取所述有效图像集。
PCT/CN2025/100098 2024-06-28 2025-06-10 激光功率确定模型构建方法及激光功率自动调节方法 Pending WO2026001663A1 (zh)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
CN202410852246.5 2024-06-28
CN202410852246.5A CN118736350A (zh) 2024-06-28 2024-06-28 一种激光功率确定模型构建方法、调节方法及存储介质
CN202410852114.2A CN118717002A (zh) 2024-06-28 2024-06-28 一种共聚焦内窥镜激光功率自动调节方法
CN202410852114.2 2024-06-28

Publications (1)

Publication Number Publication Date
WO2026001663A1 true WO2026001663A1 (zh) 2026-01-02

Family

ID=98220793

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2025/100098 Pending WO2026001663A1 (zh) 2024-06-28 2025-06-10 激光功率确定模型构建方法及激光功率自动调节方法

Country Status (1)

Country Link
WO (1) WO2026001663A1 (zh)

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107040745A (zh) * 2017-03-16 2017-08-11 精微视达医疗科技（武汉）有限公司 一种提高探头式共焦显微内窥镜帧频方法
US20170315056A1 (en) * 2014-12-08 2017-11-02 Sony Corporation Information processing device, image acquisition system, information processing method, image information acquisition method, and program
CN111678913A (zh) * 2020-06-11 2020-09-18 浙江工业大学 一种基于图像识别实现溶液浓度定量测定的实验方法
CN112906834A (zh) * 2019-12-03 2021-06-04 精微视达医疗科技（武汉）有限公司 一种内窥成像系统视场掩模的获得方法及装置
CN116528056A (zh) * 2023-01-09 2023-08-01 之江实验室 内窥镜光源亮度调节方法、装置、电子装置和存储介质
CN118196144A (zh) * 2024-04-28 2024-06-14 精微视达医疗科技(苏州)有限公司 一种显微内窥镜纤芯校正参数确定方法及纤芯跟踪方法
CN118736350A (zh) * 2024-06-28 2024-10-01 精微视达医疗科技(湖北)有限公司 一种激光功率确定模型构建方法、调节方法及存储介质

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170315056A1 (en) * 2014-12-08 2017-11-02 Sony Corporation Information processing device, image acquisition system, information processing method, image information acquisition method, and program
CN107040745A (zh) * 2017-03-16 2017-08-11 精微视达医疗科技（武汉）有限公司 一种提高探头式共焦显微内窥镜帧频方法
CN112906834A (zh) * 2019-12-03 2021-06-04 精微视达医疗科技（武汉）有限公司 一种内窥成像系统视场掩模的获得方法及装置
CN111678913A (zh) * 2020-06-11 2020-09-18 浙江工业大学 一种基于图像识别实现溶液浓度定量测定的实验方法
CN116528056A (zh) * 2023-01-09 2023-08-01 之江实验室 内窥镜光源亮度调节方法、装置、电子装置和存储介质
CN118196144A (zh) * 2024-04-28 2024-06-14 精微视达医疗科技(苏州)有限公司 一种显微内窥镜纤芯校正参数确定方法及纤芯跟踪方法
CN118736350A (zh) * 2024-06-28 2024-10-01 精微视达医疗科技(湖北)有限公司 一种激光功率确定模型构建方法、调节方法及存储介质

Similar Documents

Publication Publication Date Title
CN101404923B (zh) 医疗用图像处理装置
CN116758069B (zh) 用于肠道内镜的医学图像增强方法
CN118736350A (zh) 一种激光功率确定模型构建方法、调节方法及存储介质
WO2011129176A1 (ja) 医用画像処理装置及び医用画像処理方法
DE112015002614T5 (de) Bildbearbeitungsvorrichtung, Bildverarbeitungsverfahren und Bildverarbeitungsprogramm
CN116919639A (zh) 一种可视清洗方法、系统及其可视清洗器
CN115035086B (zh) 一种基于深度学习的结核皮试智能筛查分析方法和装置
CN113017687A (zh) 一种腹腔积液b超图像自动识别方法
CN112597981B (zh) 基于深度神经网络的肠镜退镜质量智能监控系统及方法
CN117442190B (zh) 一种基于目标检测的创面自动测量方法及系统
CN117012344A (zh) 一种4cmos相机采集的图像分析方法
WO2026021267A1 (zh) 一种共聚焦显微内窥镜的图像质量控制方法
US12569128B2 (en) Apparatus and method for detecting cervical cancer
CN118674715B (zh) 基于人工智能的消化内镜检查质量控制系统及方法
CN111028219B (zh) 一种结肠图像识别方法、装置以及相关设备
CN113270163A (zh) 基于人工智能的li-rads诊断报告生成系统及方法
CN117045178B (zh) 一种用于荧光腔镜成像的图像品质智能监测方法及系统
CN119693280A (zh) 一种神经内镜的图像去雾方法、装置及系统
CN119274787A (zh) 一种精神疾病筛选的系统、芯片、装置和计算机可读存储介质
CN112037217A (zh) 一种基于荧光成像的术中血流成像方法
CN119112298B (zh) 一种气管镜下切割器和定位系统
CN118717002A (zh) 一种共聚焦内窥镜激光功率自动调节方法
CN119896492B (zh) 冠状动脉ct血管成像质量优化方法及系统
CN118982640B (zh) 一种可视喉镜摄像的成像方法及系统
CN119924755B (zh) 医用内窥镜荧光摄像系统

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 25825048

Country of ref document: EP

Kind code of ref document: A1