METHOD FOR IMPROVING TISSUE AND CELL IMAGES

Abstract

Today, medical imaging is performed using high-end and expensive microscopes. These microscopes are difficult to access and use due to their cost, precision, and the need for large installation areas. To overcome these challenges, new devices have been developed to facilitate micro-scale imaging: Mobile Microscopes. The aim of these mobile devices is to obtain images at the micro-scale that are close to high-end microscope images. However, due to hardware limitations, image quality deteriorates at magnification ratios of 100x (and above), resulting in noisy, blurry, color-aberrant, low-contrast, and low-quality images. This project aims to improve these types of biomedical images using signal and image processing methods.The developed method improves tissue and cell images obtained from mobile microscopes using principal component analysis, good light and saturation mapping, histogram matching, and various filtering methods. This algorithm was successfully tested on tissue images obtained from a mobile microscope at the Izmir Biomedical and Genome Center. The proposed image enhancement outputs approach a high-level microscope image quality, and the algorithm is adaptable to all kinds of low-cost and low-resolution microscopic imaging devices with simple parameter changes. Furthermore, statistical comparisons were made with different algorithms found in the literature. The tests and analyses show that the developed unique method is more successful than existing algorithms in the literature in providing lower-cost and easier medical imaging.

Description

1 TARIFF METHOD FOR IMPROVING TISSUE AND CELL IMAGES Technical Area 5 This invention enables low-cost imaging devices and micro-scale, high-budget solutions. a tissue sample used to obtain images close to a microscope image and It relates to methods for improving cell images. Previous Technique Technological advancements in mobile phone cameras in recent years Improvements have led to enhanced high resolution and zoom capabilities. This has led to the development of mobile cameras with their small size, low weight, and low price. Their presence has led to their increasing use in medical applications. The use of phone cameras for microscopic examination was first introduced in 2008. It began with an article published by Yair Granot and his colleagues in [year]. Since 2008, the number of low-cost devices produced with mobile cameras has reached 20. It has increased considerably and has gained a significant market share. For example, in 2016 an Italian company that was founded and rapidly produces microscope systems for mobile phones. The share price of the growing SmartMicroOptics (SMO) is $362,000. However, mobile Phone cameras, by their very nature, need to be customized for specific applications, and This has brought about the necessity of making the necessary arrangements. This 25 Customizations can be at the physical or software level. For example, in 2018 Rivenson et al. developed a low-cost portable mobile microscope. Deep neural network learning method to improve acquired images They have used this method, a low-cost imaging example, from a phone. To reduce noise in the captured image, improve image quality, and enhance color. The aim was to correct the change. In addition, Gupta and Purkayastha simply 2 Improving images obtained from a normal microscope using image processing methods. and aimed to regulate the color distribution. Systems existing in the current state of the art have started with similar goals. There are multiple approaches, but most research in this area focuses on deep neural pathways. It uses networks. This invention, which is the subject of the application, collects millions of data. Deep nerve tissue, which is essential and cannot perform the same in different tissues. Instead of network-based approaches, the phone's camera goes beyond its intended purpose. successfully removing the noise generated by its use, 10 An algorithm has been developed to provide this. The application concerns a method for improving tissue and cell images. low cost which has a very important market stemming from the needs of industry The problems that limited the use of microscopes have been largely solved, 15 It has been brought closer to the level of professional-grade, high-cost microscopes. The invention in question involves a method for improving tissue and cell images, and tissue... and a contribution has been made to the cell image enhancement literature. The invention uses red, green, and blue color channels transmit information to each other, and thus The concept of recreating color channels at the end of the process has been used in this field for more than 20 years. It is one of the elements that has not been used before and that needs to be preserved. The technical specifications of the invention subject to the application and the application itself are as follows: The text does not include a statement regarding the technical effects provided by the invention. In current applications, a tissue and cell with similar technical characteristics 25 No methods for improving their images have been found. Purposes of the Invention 3 The aim of this invention is to utilize the developed algorithm to obtain data from low-cost microscopes. a tissue and cell image that enables the enhancement of tissue images The goal is to implement an improvement method. Another aim of this invention is to create a low-cost 5 with the developed algorithm. By improving tissue images obtained from microscopes, laboratory and a tissue that has many uses in places such as research centers and to perform a method of improving cell images. Another advantage of this invention is that the algorithm created is not embedded in the system. all tissue sample printouts except those taken with low-cost microscopes Tissue and cell images that can be improved with easily made changes. The goal is to implement an improvement method. Another purpose of this invention is to use the improved algorithm to interact with phone cameras and various other devices. images embedded within low-budget systems constructed with lenses a tissue and cell that can make the healing process much more effective The method is to improve their images. Another purpose of this invention is to improve the cameras used in professional microscopes. Unlike improving image outputs with simple image enhancement techniques, this The algorithm, which will be embedded in the systems, can also be used in professional microscopes. an enhancement of tissue and cell images that enables it to improve its outputs The method is to implement it. 25 Another objective of this invention is to integrate the improved algorithm with machine learning and data to be used in data processing methods such as deep learning networks because it will improve and standardize, the benefits obtained from such methods a method of improving tissue and cell images that greatly contributes to the results 30 to accomplish. 4 Brief Description of the Invention The initial requirement and other requirements implemented to achieve the purpose of this invention... A method for improving tissue and cell images described in the requests, noise 5 elimination, histogram matching, map generation, filtering, and finally five main color channels as the reconstruction of red, green and blue color channels. It consists of sections. The application concerns tissue and cell images in the invention. The algorithm of the improvement method is taken from a low-budget mobile microscope. low-quality tissue images and the same 10 taken from the same section of the same tissue input of high-quality microscope images at varying zoom levels is obtained. The invention in question involves obtaining high-quality tissue samples from the same tissue section. Microscope images were used as a reference. The invention in question is... For example, in the algorithm of the method for improving tissue and cell images When the resulting output is visually examined, the result is much clearer and 15 compared to the input. This ensures that an output close to the reference is produced. Detailed Description of the Invention Tissue and cell images taken to achieve the goal of this discovery 20 The improvement method is shown in the attached figures; Figure 1. Schematic of the method for enhancing tissue and cell images. It is the appearance. The parts in the figures are individually numbered, and these numbers correspond to: It is given below. 100. Method for improving tissue and cell images. 101. Low-quality tissue images obtained from a low-budget mobile microscope and High zoom level at the same rate taken from the same section of tissue. input of high-quality microscope images 102. Detection in the input histogram using a Non-Local Means (NLM) filter. Elimination of the resulting Gaussian noise type 5 103. Histogram of the color input image compared to the histogram of the reference image. matching and then matching the three color channels with each other to obtain data. ensuring transfer 104. For the nine outputs derived from histogram matching between color channels, saturation, good lighting (WE), and the top 10 features that highlight different characteristics of the image Extraction of component analysis (PCA) maps 105. As a result of the mapmaking process, one of each of the three map types was produced. After a total of twenty-seven different maps were created, the maps... filtering 106. Maps created using green (G), red (R) and blue (B) color channels 15 reconstruction using 107. Creating a new output that is close to the reference image. Imaging devices provide high-budget microscope images at the micro-scale. Tissue and cell images used to obtain close-up views 20 Improvement method (100) in its most basic form, - low-quality tissue images taken from a low-budget mobile microscope and the same High quality image taken from the same section of tissue at the same zoom level. Input of microscope images (101), - 25 detected in the input histogram using a Non-Local Means (NLM) filter. Elimination of Gaussian noise type (102), - histogram of the color input image compared with the histogram of the reference image. matching and then matching the three color channels with each other to obtain data. ensuring transfer (103), 6 - for nine outputs derived from histogram matching between color channels, Saturation, good lighting (WE), and main features that highlight different characteristics of the image Extraction of component analysis (PCA) maps (104), - as a result of the mapmaking process, one of each of the three map types After creating a total of twenty-seven different maps, 5 of the maps filtering (105), - Maps created from green (G), red (R) and blue (B) color channels reconstruction using (106), - the steps of creating a new output close to the reference image (107) It includes. 10 Method of improving tissue and cell images which are the subject of the application (100), Imaging devices provide high-budget microscope images at the micro-scale. It is used to obtain close-up images. Tissue and cell images. Improvement method (100), low cost with the developed algorithm 15 It helps to improve tissue images obtained from microscopes. Tissue and Method of improving cell images (100), with the developed algorithm, low by improving tissue images obtained from expensive microscopes It has many uses in places such as laboratories and research centers. It has. Method of improving tissue and cell images (100). Tissue and cell 20 Method of improving images (100), the algorithm created is embedded in the system all tissue samples taken except with low-cost microscopes because of this The outputs can be easily improved by making changes to the parameters. Method of improving tissue and cell images (100), phone camera and various The image is embedded in low-budget systems built with lenses. The healing process can be made much more effective. Tissue and cell method of improving images (100), improved algorithm of machine will be used in data processing methods such as learning and deep learning networks. Because it will improve and standardize the data, the benefits obtained from such methods It has made a significant contribution to the results. 30 7 Improving tissue and cell images, which is included in one application of the invention. In the method (100), two different types of input were used: low-budget mobile Low-quality tissue images obtained through a microscope, and images of the same tissue. High-quality microscope with the same magnification level obtained from the same section. images. Improving tissue and cell images in the invention that is the subject of the application 5 In the algorithm of method (100), high quality microscope images are referenced. It has been used as. Tissue and cells involved in one application of the invention. In the method of improving their images (100), firstly low-budget mobile Low-quality tissue images obtained through a microscope, and images of the same tissue. High-quality microscope with the same magnification level as the section 10 The process of inputting images (101) is carried out. Reference and Non-Local after performing inputs (101) of low imagery Gaussian detected in the input histogram using the Means (NLM) filter Noise type elimination (102) is performed. NLM filter, image It is one of the noise reduction filters frequently used in the processing field, and there are 15 similar ones. It is preferred because it provides less sound softening compared to filters. Noise histogram of the color input image after removal (102) process step matching with the reference image histogram (103) process step is performed. In histogram matching (103), the color input is first The histogram of the image is matched with the histogram of the reference image. 20 Thus, color distortions in the image obtained from the low-cost microscope are significant. It has been resolved to some extent. Afterwards, the color channels are paired with each other and each color Three outputs are generated for each channel. The purpose of this process is to obtain color channel information. transferring color channels to other channels so that maps can be generated later. It is possible to recreate it. As a result of this process, there will be 25, three from each channel. A total of nine outputs were generated; for example: GR green channel coupled with red channel. While showing the result, BG shows the blue result paired with the green channel. Then there's saturation, which highlights different features of the image, and good lighting (WE). and extraction of principal component analysis (PCA) maps (104) process step 30 is carried out. In the step of map extraction (104), the image is 8 Three types of maps were used, highlighting different characteristics. These are: Saturation The maps are Good Light (WE) Maps and Principal Component Analysis (PCA) Maps. The general purpose of PCA Maps is to assign higher weight to dominant pixels. The aim is to provide this type of map. These types of maps are particularly useful in the field of image processing, especially with Multi Exposure Fusion. It is frequently used in this field. PCA weight maps are the 5 mentioned in the second step. The outputs are calculated accordingly. For example: Outputs paired with channel R are RR, GR, and BR. These are nxm matrices. These three images are converted into vectors, nm x 3. Scores were plotted on the columns of a matrix of a certain size and observed using PCA. calculated. Afterwards, these found PCA score vectors are in the range [0 1]. It has been linearly normalized and converted back to its original dimensions (nxm). This process is 10 This was repeated for outputs paired with green and blue (Result naming). Examples: PRR, PGR, and PBR). WE maps were first proposed by Mertens et al. and later by Ulucan and It has been adapted by the researchers. WE, meaning good lighting, is best at 15. It identifies illuminated areas. The purpose of WE maps is to show areas with low or moderate illumination. The goal is to eliminate the overly illuminated areas and highlight the best-illuminated ones. Considering the types of images we use, the image Since the background is white, the resulting weight maps are normalized between [0 1] It is subtracted from 1. The operation is given as a formula below (1), in this formula Y 20 µ indicates the brightness channel of any of the 9 outputs of the second stage. σ represents the arithmetic mean of Y and the standard deviation of σ. (Result) Examples of naming conventions: ERR, EGR, and EBR) Saturation maps use a weighting system to highlight image brightness. It is a type of saturation map. To arrive at the saturation map, each channel is matched. Outputs, for example: If the outputs paired with channel R are considered as RR, GR and BR, then the color Standard deviation values ​​are calculated for each pixel based on the average of its channels. 30 𝐸𝑌 = 1 − 𝑒𝑥𝑝(− (𝑌−(1−µ))2 2σ2 ) (1) 9 and is used. Maps that are produced taking our background into consideration [0 1] It was normalized and subtracted from 1, and used in this way. (Result) Naming examples: SRR, SGR, and SBR) Extraction of maps included in an application of the invention (104) 5 of the process steps then, as a result of the mapmaking process, one of each of the three map types After a total of twenty-seven different maps were created, the maps... Filtering (105) is provided. After these three types of maps are multiplied combined, multiplied by 255, brought back to the range [0 255], so that the final Weights were found. Finally, these weights were filtered using a Laplacian filter to obtain 10 weights. It has been sharpened. The Laplacian filter is the most frequently used filter in image processing. It is one of the sharpening filters and creates a second-order derivative of the image. It uses. This process is given in the following formula (2) (Result naming Examples: WRR, WGR, and WBR). After the filtering (105) process step, green (G), red (R) and blue (B) colors The process step of reconstructing the channels (106) is carried out. In this step, all three channels were recreated. The outputs paired with channel R are RR, 20 GR and BR were multiplied and added together with their corresponding weights, and then this... The total weights were normalized by dividing by the sum of the weights. Thus, R, G, and B channels R', B', G' are recreated (3). After this process, the following The results were combined and created as a three-channel (RGB) image. Then this... outputs and corresponding high-cost reference images, YCbCr 25 The outputs are transferred to the coordinate system. The Cb and Cr color channels are referenced. The images were histogram-matched with Cb and Cr channels, thus determining color balance. It has been recovered. 𝑊𝑌 = 𝑳𝒂𝒑𝒍𝒂𝒄𝒊𝒂𝒏(𝑃𝑌 ∗ 𝐸𝑌 ∗ 𝑆𝑌) (2) After the color channels are recreated (106), the reference image is shown. A new output is formed in the near future (107). When this output is examined visually The result appears to be much clearer and closer to the reference than the input. 5

Claims

11 REQUESTS 1. Imaging devices provide high-budget microscope images at the micro-scale. Used to obtain close-up images, in its most basic form, - low-quality tissue images taken from a low-budget mobile microscope and 5 High zoom level at the same rate taken from the same section of tissue. input of quality microscope images (101), - Detection in the input histogram using a Non-Local Means (NLM) filter. Elimination of the Gaussian noise type (102), - histogram of the color input image compared with the histogram of the reference image and 10 matching color channel histograms with each other and thus nine output creation (103), - saturation, good lighting (WE), and other features that highlight different aspects of the image. Extraction of principal component analysis maps (104), • In the creation of maps (104) the image is more or less illuminated 15 by eliminating areas and highlighting the best-lit areas, good lighting. (WE) mapping, • In the extraction of maps (104) a higher value to dominant pixels highlighting different features of the image in order to give it weight saturation maps, well-exposure (WE) maps, and principal component analysis 20 Three types of maps will be used, including (PCA) maps. - as a result of the mapmaking process, one of each of the three map types after a total of twenty-seven different maps were created filtering maps (105), - Reconstruction of green (G), red (R) and blue (B) color channels 25 (106), - characterized by the formation of a new output close to the reference image (107) Method of improving tissue and cell images (100).

Images