Super-resolution compensation method based on scanning probe technology combined with electrochemiluminescence
By combining electrochemiluminescence combination in scanning probe technology, SECM-ECL three-dimensional array and deep learning network are constructed to distinguish between regions of interest and non-interest, image preprocessing and feature extraction are carried out, and scanning probes are solved in solution The damping interference problem during movement in the environment achieves high-precision and efficient time resolution detection.
Patent Information
- Application Number
- CN202210536924.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-05-17
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2042-05-17
AI Technical Summary
In the prior art, the reality of the image is disturbed due to damping when the scanning probe moves in a solution environment. Especially at high scanning speed and high sampling frequency, it is difficult to maintain high accuracy, resulting in a prolonged experimental time.
Using a super-resolution compensation method based on scanning probe technology and electrochemiluminescence, we can distinguish between regions of interest and non-interest by constructing a SECM-ECL three-dimensional array, and use interpolation algorithm and deep learning network for image preprocessing and feature extraction. , realize background fusion and motion compensation for non-interested areas, and generate virtual detection images to improve temporal resolution.
The time resolution and signal-to-noise ratio of scanning probe technology in biological long-term detection is improved, the experimental time is reduced, while maintaining high accuracy, and solving the impact of damping interference on the image.
Smart Images

Figure CN114965448B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of super-resolution compensation, and in particular to a super-resolution compensation method based on the combination of scanning probe technology and electrochemical luminescence. Background Art
[0002] For SECM experiments, the most common detection mode is the positive and negative feedback mode. The feedback working mode is mainly used to determine the height of the probe relative to the substrate being measured. It is divided into positive feedback mode and negative feedback mode, which are mainly distinguished by whether the current increases or decreases when the probe approaches the substrate being measured.
[0003] In the prior art, when the probe enters a solution environment, the surrounding solution will have a certain damping effect on the movement of the working probe. If the scanning speed and the sampling frequency are too high, the solution will interfere with the authenticity of the image. The general practice is to reduce the scanning speed as much as possible, increase the dwell time of the scanning point, and improve the step accuracy, but at the same time it also prolongs the experimental running time. Therefore, it is necessary to invent a super-resolution compensation method based on scanning probe technology combined with electrochemiluminescence to solve the above problems. Summary of the invention
[0004] 1. Technical issues to be resolved
[0005] In view of the shortcomings of the prior art, the present invention provides a super-resolution compensation method based on scanning probe technology combined with electrochemiluminescence, which realizes mutual detail enhancement based on time and channels through regions of interest and non-regions of interest, and has sufficiently good time, spatial resolution and signal-to-noise ratio for the super-resolution of the overall detection data based on samples through super-resolution. The reaction area is predicted by temporal multi-potential acquisition images, and the background of non-regions of interest is fused to generate a predicted image to improve the time resolution, thereby providing possibility for long-term detection of biological organisms with scanning probe technology.
[0006] (II) Technical solution
[0007] In order to achieve the above-mentioned improvement in time resolution, the present invention provides the following technical solutions:
[0008] A super-resolution compensation method based on scanning probe technology combined with electrochemiluminescence comprises the following steps:
[0009] S1. Arrange the SECM-ECL image matrix according to each potential, add the time dimension, and form a SECM-ECL three-dimensional array [(X, Y, current, potential, time].
[0010] Specifically, by adding different detection solutions, different targets are detected at different potentials, and each potential (including electrochemical excitation potential) is used to detect the substrate in different channels.
[0011] S2. Sort the regions of interest in the SECM-ECL three-dimensional array and take the intersection to obtain the maximum region of interest to distinguish the overall region of interest from the overall non-region of interest.
[0012] Specifically, each potential is regarded as a different color channel, the sampling rate is each scanning imaging time, the imaging interval of the front part is large, and the imaging interval of the back part is short. The front scanning area is large (interested and non-interested) and the imaging area of the back part is small and mostly concentrated in the area of interest.
[0013] S3. Use an interpolation algorithm to pre-enlarge each region of interest to enhance the edge and ensure that all input images have the same input size. Specifically, it is based on simulated interpolation, or based on theoretical models and deep learning to obtain error functions from samples.
[0014] S4. The SECM-ECL output group processed in step S3 is used as the input of the deep learning network, and features are extracted process by process and channel by channel by constructing a temporal convolutional deep network (similar to the common convolutional network combined with a recurrent network).
[0015] Specifically, different convolution sub-networks are used to perform inverse convolution to achieve super-resolution of the image, and the interference error function is obtained in a sample-based manner.
[0016] S5. Use the peak signal-to-noise ratio as an evaluation index to obtain the interference error function (convolution function) of each channel, and achieve super-resolution for each channel of each frame through upsampling.
[0017] Specifically, machine learning is used to extract overall features channel by channel and time by time, and the overall resolution and signal-to-noise ratio of non-interest areas are comprehensively improved.
[0018] S6. Construct a virtual detection background based on the fusion of all non-interested regions.
[0019] Specifically, the probe needs to be stationary for a certain period of time after it moves, so that the scanning probe is in a stable state. Different potentials have different regions of interest, but the whole has a common non-sensory region of interest.
[0020] S7, extracting the region of interest in the acquired super-resolution array, establishing the migration vector of the four adjacent regions of interest at each potential, and generating the virtual region of interest in the middle of the four input images according to the vector.
[0021] Specifically, the collected data is super-resolved through the nonlinear learning ability of machine learning, and the temporal resolution is further improved by generating virtual detection images through motion compensation.
[0022] S8. Merge the virtual background with the virtual region of interest to improve the overall temporal resolution.
[0023] Specifically, multiple input channels are established by means of motion compensation to predict the reaction migration of the region of interest, and a virtual detection image between two imaging regions is generated by prediction.
[0024] (III) Beneficial effects
[0025] Compared with the prior art, the present invention provides a super-resolution compensation method based on scanning probe technology combined with electrochemiluminescence, which has the following beneficial effects:
[0026] 1. This super-resolution compensation method based on scanning probe technology combined with electrochemiluminescence realizes time- and channel-based mutual detail enhancement through regions of interest and non-regions of interest, and has sufficiently good resolution and signal-to-noise ratio for the overall detection data based on samples through super-resolution.
[0027] 2. This super-resolution compensation method based on the combination of scanning probe technology and electrochemiluminescence predicts the reaction area by acquiring images with multiple potentials in time, and fuses the background of non-interested areas to generate predicted images to improve the temporal resolution, making it possible for long-term detection of biological organisms using scanning probe technology. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] Figure 1 Schematic diagram of a process of a super-resolution compensation method according to an embodiment of the present invention;
[0029] Figure 2 Schematic diagram of data acquisition principle of the super-resolution compensation method according to an embodiment of the present invention. DETAILED DESCRIPTION
[0030] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0031] Example:
[0032] See also Figure 1-2 The super-resolution compensation method based on scanning probe technology combined with electrochemiluminescence provided in an embodiment of the present invention comprises the following steps:
[0033] S1. Arrange the SECM-ECL image matrix according to each potential, add the time dimension, and form a SECM-ECL three-dimensional array [(X, Y, current, potential, time].
[0034] S2. Sort the regions of interest in the SECM-ECL three-dimensional array, take the intersection to obtain the largest region of interest, so as to distinguish the overall region of interest from the overall region of no interest, regard each potential as a different color channel, and the sampling rate is the imaging time of each scan. The imaging interval of the front part is large, and the imaging interval of the back part is short. The front scanning area is large (including the region of interest and the region of no interest), and the imaging area of the back part is small and mostly concentrated in the region of interest.
[0035] S3. Use interpolation algorithms to pre-enlarge each region of interest to enhance the edges and ensure that all input images have the same input size, based on simulated interpolation, or obtain error functions from samples based on theoretical models and deep learning.
[0036] S4. The SECM-ECL output group processed in step S3 is used as the input of the deep learning network. By constructing a time-dependent convolutional deep network (similar to the common convolutional network combined with a recurrent network), features are extracted process by process and channel by channel. Inverse convolution is performed with different convolution sub-networks to achieve super-resolution of the image, and the interference error function is obtained in a sample-based manner.
[0037] S5. Use the peak signal-to-noise ratio as the evaluation index to obtain the interference error function (convolution function) of each channel, achieve super-resolution for each frame and channel through upsampling, extract overall features channel by channel and time by machine learning, and comprehensively improve the overall resolution and signal-to-noise ratio of non-interested areas.
[0038] S6. Based on the fusion of all non-interested regions, a virtual detection background is constructed. After the probe moves, it needs to be still for a certain period of time so that the scanning probe is in a stable state. Different potentials have different regions of interest, but the whole has a common non-interested region.
[0039] S7. Extract the region of interest from the acquired super-resolution array, establish the migration vector of the four adjacent regions of interest under each potential, and generate a virtual region of interest in the middle of the four input images based on the vector. Super-resolve the collected data through the nonlinear learning ability of machine learning, and generate a virtual detection image through motion compensation to further improve the temporal resolution.
[0040] S8. The virtual background is merged with the virtual region of interest to improve the overall temporal resolution. Multiple input channels are established through motion compensation to predict the reaction migration of the region of interest, and a virtual detection image between two imaging regions is predicted and generated.
[0041] The super-resolution compensation method based on the combination of scanning probe technology and electrochemiluminescence provided by the present invention realizes mutual detail enhancement based on time and channels through the region of interest and the non-region of interest, and has sufficiently good time, spatial resolution and signal-to-noise ratio for the super-resolution of the overall detection data based on the sample through super-resolution. The reaction area is predicted by acquiring images with multiple potentials in time, and the background of the non-region of interest is fused to generate a predicted image to improve the time resolution, thereby providing the possibility for long-term detection of organisms with the scanning probe technology.
[0042] The present invention provides a super-resolution compensation method based on scanning probe technology combined with electrochemiluminescence, which organizes the SECM-ECL image matrix according to each potential, adds a time dimension, forms a SECM-ECL three-dimensional array, organizes the regions of interest in the SECM-ECL three-dimensional array, takes the intersection to obtain the maximum region of interest, distinguishes the overall region of interest from the overall non-region of interest, uses an interpolation algorithm to pre-amplify each region of interest to improve the edge, and ensures that all input images have the same input size, and uses the processed SECM-ECL output group as a deep learning network input. The super-resolution compensation method based on scanning probe technology combined with electrochemiluminescence realizes time- and channel-based mutual detail enhancement through the region of interest and the non-region of interest, and has sufficiently good time, spatial resolution and signal-to-noise ratio for the super-resolution of the overall detection data based on the sample through super-resolution.
[0043] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. A super-resolution compensation method based on scanning probe technology combined with electrochemiluminescence, characterized in that: The following steps are involved: S1, sorting the SECM-ECL image matrix according to each potential, adding the time dimension, and forming a SECM-ECL three-dimensional array; S2, sorting the regions of interest in the SECM-ECL three-dimensional array, taking the intersection to obtain the maximum region of interest, so as to distinguish the overall region of interest from the overall region of non-interest; S3, using an interpolation algorithm to pre-enlarge each region of interest to enhance the edge and ensure that all input images have the same input size; S4, using the SECM-ECL output group processed in step S3 as the input of the deep learning network, extracting features process by process and channel by channel by constructing a time-dependent convolutional deep network, performing inverse convolution with different convolution sub-networks to achieve super-resolution of the image, and obtaining the interference error function in a sample-based manner; S5, using the peak signal-to-noise ratio as an evaluation index to obtain the interference error function of each channel, and achieving super-resolution for each channel of each frame through upsampling; S6, constructing a virtual detection background based on fusion of all non-interested regions; S7, extracting the region of interest in the acquired super-resolution array, establishing the migration vectors of the four adjacent regions of interest at each potential, and generating the virtual region of interest in the middle of the four input images according to the vectors; S8. The virtual background is merged with the virtual region of interest to improve the overall temporal resolution. Multiple input channels are established through motion compensation to predict the reaction migration of the region of interest, and a virtual detection image between two imaging regions is predicted and generated.
2. The super-resolution compensation method based on scanning probe technology combined with electrochemiluminescence according to claim 1, characterized in that: In the step S1, different detection solutions are added to detect different targets at different potentials, and the substrate is detected at different channels at various potentials including the electrochemical excitation potential.
3. The super-resolution compensation method based on scanning probe technology combined with electrochemiluminescence according to claim 1, characterized in that: In step S2, each potential is regarded as a different color channel, the sampling rate is the imaging time of each scan, the imaging interval of the front part is large, and the imaging interval of the back part is short, the front scanning area is large and the back imaging area is small and mostly concentrated in the area of interest.
4. The super-resolution compensation method based on scanning probe technology combined with electrochemiluminescence according to claim 1, characterized in that: In step S3, an error function is obtained from the sample based on simulation interpolation, or based on a theoretical model and deep learning.
5. The super-resolution compensation method based on scanning probe technology combined with electrochemiluminescence according to claim 1, characterized in that: In step S5, overall feature extraction is performed channel by channel and time by time through machine learning, and the overall resolution and signal-to-noise ratio of non-interested areas are comprehensively improved.
6. The super-resolution compensation method based on scanning probe technology combined with electrochemiluminescence according to claim 1, characterized in that: In step S6, the probe needs to be stationary for a certain period of time after moving, so that the scanning probe is in a stable state, and different potentials have different regions of interest, but the whole has a common non-interest region.
7. The super-resolution compensation method based on scanning probe technology combined with electrochemiluminescence according to claim 1, characterized in that: In step S7, the collected data is super-resolved by using the nonlinear learning ability of machine learning, and a virtual detection image is generated by motion compensation to further improve the temporal resolution.
