A microscopic medical image recognition system and its analysis method

Abstract

The present invention relates to the technical field of image recognition, and specifically to a microscopic medical image recognition system and its analysis method. The beneficial effects of the present invention are as follows: By using the conveying unit, automatic conveyance of the slide can be achieved, enabling the slide to automatically move to the image recognition position of the microorganism image recognition instrument body, reducing manual intervention, and initially positioning the position of the slide. By driving the rotating rod to rotate through the motor, the conveyor belt can be driven to drive the slide to move on the top of the microorganism image recognition instrument body, so that the slide moves to the image recognition position of the microorganism image recognition instrument body. By using the positioning unit, the tissue sample on the carrier slide can be accurately aligned with the center of the image recognition position, further increasing the integrity of image recognition. By restricting the moving distance of the slide through the fixed plate and guiding the movement of the slide through the guide plate, the shooting of the tissue sample can be made more complete.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and particularly to a microscopic medical image recognition system and an analysis method thereof. Background Art

[0002] Microscopic medical image recognition devices have broad application prospects in the fields of medical diagnosis, disease research, and medical education. They can help doctors identify microbial infections more quickly and accurately, improve the accuracy and efficiency of diagnosis. At the same time, they can also provide a large amount of sample data for disease research, promoting the progress of medical science.

[0003] When a microscopic medical image recognition device identifies microorganisms, a glass slide containing a tissue sample is usually placed below the image recognition device, and the glass slide is irradiated by a lamp tube at the bottom for supplementary lighting treatment. The tissue sample is recognized and photographed by the image recognition device, and the photographed image is subjected to filtering and denoising treatment to complete the acquisition of the microorganism image. When docking the glass slide with the image recognition device, manual docking is usually adopted. Through the manual field of view, the glass slide is placed on top of the supplementary light lamp, and the position of the glass slide is manually fine-tuned, resulting in an increase in the adjustment time. Using manual adjustment reduces the integrity of the shooting.

[0004] Combining the above problems, we will find that the existing ones on the market usually adopt manual docking. Through the manual field of view, the glass slide is placed on top of the supplementary light lamp, and the position of the glass slide is manually fine-tuned, resulting in an increase in the adjustment time. Using manual adjustment reduces the integrity of the shooting. When in use, it is very difficult to avoid the above-mentioned problems at the same time, and even if it can be solved, it needs to be solved with the cooperation of external tools, thus unable to achieve the desired effect. Therefore, we propose a microscopic medical image recognition system. Summary of the Invention

[0005] The purpose of the present invention is to provide a microscopic medical image recognition system and an analysis method thereof to solve the problems raised in the above background art.

[0006] To achieve the above purpose, the present invention provides the following technical solution: A microscopic medical image recognition system, a friction pad is fixedly installed at the bottom of the main body of the microorganism image recognition instrument, and a transmission and positioning mechanism is arranged on one side of the main body of the microorganism image recognition instrument;

[0007] The transmission and positioning mechanism includes a transmission unit, the transmission unit is arranged on one side of the main body of the microorganism image recognition instrument, and the transmission unit is used for transmitting the glass slide;

[0008] The transfer and positioning mechanism further includes a positioning unit, which is arranged on the top of the microbial image recognition device body, and the positioning unit is used for precise positioning of the slide;

[0009] A collection mechanism is arranged on one side of the microbial image recognition device body. The collection mechanism is used for fixing the collection position of the slide, and the collection mechanism is used in cooperation with the transfer unit.

[0010] Preferably, the transfer unit includes a collection box, which is fixedly installed on one side of the microbial image recognition device body. A moving groove is formed in the top of the microbial image recognition device body. The number of the moving grooves is two. A placement groove is formed in the inner wall of the moving groove. A motor is fixedly installed on the inner wall of the placement groove. The output end of the motor is fixedly installed with a rotating rod. One end of the rotating rod penetrates to the inner wall of the moving groove. A driven rod is rotatably connected to the inner wall of the collection box. A conveyor belt is sleeved on the surface of the rotating rod. The surfaces of the rotating rod and the driven rod are rotationally connected through the conveyor belt. The conveyor belt is made of rubber.

[0011] Preferably, the positioning mechanism includes a sliding groove. A fixing plate is slidably connected to the inner cavity of the sliding groove. A connecting hole is formed in the inner wall of the sliding groove. A support plate is fixedly installed on one side of the fixing plate. A positioning rod is fixedly installed on the top of the support plate. One end of the positioning rod penetrates to the inner cavity of the connecting hole. A positioning block is threadedly connected to the surface of the positioning rod. The bottom of the positioning block is in close contact with the top of the microbial image recognition device body. A guiding plate is fixedly installed on the top of the microbial image recognition device body.

[0012] Preferably, the collection mechanism includes a shielding plate, which is slidably connected to the inner cavity of the collection box. A limiting plate is fixedly installed on one side of the shielding plate. A support frame is fixedly installed at the bottom of the limiting plate. A clamping groove is formed in one side of the limiting plate. A connecting rod is slidably connected to the inner cavity of the clamping groove. A blocking rod is fixedly installed at the top of the connecting rod.

[0013] Preferably, blocking blocks are fixedly installed on the surfaces of the rotating rod and the driven rod, and the surfaces of the blocking blocks are in contact with the surface of the conveyor belt.

[0014] Preferably, bearings are fixedly installed on the inner walls of the moving groove and the collection box. The inner walls of the moving groove and the collection box are fixedly connected to the outer sides of the bearing outer rings, and the surfaces of the rotating rod and the driven rod are fixedly connected to the inner sides of the bearing inner rings.

[0015] Preferably, a connecting rod is fixedly installed in the inner cavity of the sliding groove. The inner wall of the fixing plate is slidably connected to the surface of the connecting rod. A decompression pad is fixedly installed on the surface of the fixing plate. The decompression pad is made of rubber.

[0016] Preferably, a connection block is fixedly installed at the top of the collection box, an elastic clip is fixedly installed inside the connection block, a support rod is fixedly installed at the top of the baffle plate, and the surface of the support rod is in contact with the inside of the elastic clip.

[0017] Preferably, it includes the following steps:

[0018] A1: Preprocessing of training data, adjusting the size of the image to meet the input requirements of the model, using random flipping operations with a probability of 50%, and increasing the data diversity by horizontally or vertically flipping the image;

[0019] A2: Construction of the object detection model, using the Faster R-CNN network with ROIAlign to construct the object detection model. The specific steps of step A2 include that Faster R-CNN is a network structure based on the FPN model for object region detection, and using the Region Proposal Network (RPN) to predict the candidate rectangular boxes of the object detection regions;

[0020] A3: Construction of the semantic segmentation network. After the object detection process is completed, the rectangular boxes corresponding to each target substance will be detected, the image regions corresponding to each rectangular box will be extracted, and using the semantic segmentation algorithm, the image regions containing only the target substance part will be segmented.

[0021] A method for using a microscopic medical image recognition system includes the following steps:

[0022] S1: Place the glass slide on the baffle plate, guide the installation of the glass slide through the limit plate, support the bottom of the glass slide through the support frame, and be able to push the stacked glass slides through the sliding connection between the stop rod and the card slot, so that adjacent glass slides can be accurately aligned;

[0023] S2: Make the fixed plate slide in the inner cavity of the sliding groove, adjust the distance between the fixed plate and the accurate shooting position according to the position of the tissue sample on the glass slide, so that the tissue sample on the glass slide can accurately enter the recognition range;

[0024] S3: Start the motor, drive the glass slide through the conveyor belt. The conveyor belt is located at both ends of the glass slide, guide the movement of the glass slide through the guide plate, so that the glass slide moves horizontally to the bottom of the camera of the microorganism image recognition instrument body and contacts the surface of the fixed plate, so that the glass slide can be accurately aligned with the image recognition position of the microorganism image recognition instrument body.

[0025] Compared with the prior art, the beneficial effects of the present invention are:

[0026] 1. The present invention utilizes a conveying unit to achieve automatic conveyance of a glass slide, enabling the glass slide to automatically move to the image recognition position of the microbial image recognition instrument body, reducing manual intervention, and initially positioning the glass slide. By driving a rotating rod to rotate with a motor, it can drive a conveyor belt to drive the glass slide to move on the top of the microbial image recognition instrument body, so that the glass slide moves to the image recognition position of the microbial image recognition instrument body.

[0027] 2. The present invention utilizes a positioning unit to enable the tissue sample on the carrier slide to be precisely aligned with the center of the image recognition position, further increasing the integrity of image recognition. By restricting the moving distance of the glass slide with a fixing plate and guiding the movement of the glass slide with a guiding plate, precise positioning of the tissue sample on the glass slide is achieved, enabling more complete shooting of the tissue sample.

[0028] 3. The present invention utilizes a collection mechanism to keep the glass slides collected in the inner cavity of the collection box stacked vertically. By guiding the stacking of the glass slides with a limiting plate, the adjacent two glass slides are vertically butted. In case the stacked glass slides are skewed, the glass slides are pushed by the sliding connection between the card slots and the blocking rods to keep the adjacent two glass slides vertical, further increasing the accuracy of the position of the glass slides when driving and conveying the glass slides. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 is a schematic diagram of the overall structure of the present invention;

[0030] Figure 2 is a schematic diagram of the structure of the moving groove and the guiding plate of the present invention;

[0031] Figure 3 is of the present invention Figure 2 is an enlarged view of part A in;

[0032] Figure 4 is a schematic diagram of the structure of the motor, rotating rod, driven rod and conveyor belt of the present invention;

[0033] Figure 5 is a schematic diagram of the structure of the collection box of the present invention;

[0034] Figure 6 is a schematic diagram of the structure of the fixing plate and the connecting rod of the present invention;

[0035] Figure 7 is a schematic diagram of the structure of the shielding plate, limiting plate, support frame and blocking rod of the present invention;

[0036] Figure 8 is a schematic diagram of the structure of the connecting block and the elastic clip of the present invention;

[0037] Figure 9 is a recognition diagram of the image recognition system of the present invention;

[0038] Figure 10 This is the flowchart of the image recognition system of the present invention.

[0039] In the figure: 1. Main body of the microbial image recognition instrument; 11. Friction pad; 2. Conveyor positioning mechanism; 21. Conveyor unit; 2101. Collection box; 2102. Moving groove; 2103. Placing groove; 2104. Motor; 2105. Rotating rod; 2106. Driven rod; 2107. Conveyor belt; 2108. Stopper; 2109. Bearing; 22. Positioning unit; 2201. Sliding groove; 2202. Fixed plate; 2203. Connecting hole; 2204. Support plate; 2205. Positioning rod; 2206. Positioning block; 2207. Guide plate; 2208. Connecting rod; 2209. Decompression pad; 3. Collection mechanism; 31. Baffle; 32. Limiting plate; 33. Support frame; 34. Card slot; 35. Shift lever; 36. Connecting block; 37. Elastic clip; 38. Support rod. Specific embodiments

[0040] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0041] Embodiment 1

[0042] Please refer to Figures 1-10 , the present invention provides a technical solution: a microscopic medical image recognition system. The present invention makes corresponding improvements to the technical problems mentioned in the background technology, including a main body 1 of a microbial image recognition instrument, characterized in that: a friction pad 11 is fixedly installed at the bottom of the main body 1 of the microbial image recognition instrument, and a conveyor positioning mechanism 2 is arranged on one side of the main body 1 of the microbial image recognition instrument;

[0043] The conveyor positioning mechanism 2 includes a conveyor unit 21, and the conveyor unit 21 is arranged on one side of the main body 1 of the microbial image recognition instrument, and the conveyor unit 21 is used for conveying the glass slide.

[0044] The conveying unit 21 includes a collection box 2101, which is fixedly installed on one side of the microbial image recognition device body 1. There are two moving grooves 2102 opened at the top of the microbial image recognition device body 1. A placement groove 2103 is opened on the inner wall of the moving groove 2102. A motor 2104 is fixedly installed on the inner wall of the placement groove 2103. The output end of the motor 2104 is fixedly installed with a rotating rod 2105. One end of the rotating rod 2105 penetrates to the inner wall of the moving groove 2102. A driven rod 2106 is rotatably connected to the inner wall of the collection box 2101. A conveyor belt 2107 is sleeved on the surface of the rotating rod 2105. The surfaces of the rotating rod 2105 and the driven rod 2106 are rotationally connected through the conveyor belt 2107. The conveyor belt 2107 is made of rubber;

[0045] Blocks 2108 are fixedly installed on the surfaces of the rotating rod 2105 and the driven rod 2106. The surfaces of the blocks 2108 are in contact with the surface of the conveyor belt 2107;

[0046] Bearings 2109 are fixedly installed on the inner walls of the moving groove 2102 and the inner wall of the collection box 2101. The inner walls of the moving groove 2102 and the collection box 2101 are fixedly connected to the outer sides of the outer rings of the bearings 2109. The surfaces of the rotating rod 2105 and the driven rod 2106 are fixedly connected to the inner sides of the inner rings of the bearings 2109.

[0047] The specific implementation manner of this embodiment is as follows: A plurality of glass slides are placed vertically in the inner cavity of the collection box 2101. The lowermost glass slide is in contact with the top of the conveyor belt 2107. The motor 2104 is started through an external power supply. The motor 2104 drives the rotating rod 2105 to rotate, which can drive the rotation of the conveyor belt 2107. The rotation of the conveyor belt 2107 can drive the rotation of the driven rod 2106, achieving the purpose of conveying the glass slides. During the conveying process, the bearings 2109 reduce the friction between the driven rod 2106 and the rotating rod 2105 and the inner wall of the moving groove 2102, making the rotation of the rotating rod 2105 and the driven rod 2106 smoother. Through the blocks 2108, the conveyor belt 2107 is prevented from moving on the surfaces of the rotating rod 2105 and the driven rod 2106, making the transmission of the conveyor belt to the glass slides more stable, making the movement of the glass slides more stable, and increasing the detection accuracy.

[0048] Embodiment 2

[0049] Please refer to Figures 1-10 , the present invention provides a technical solution: A microscopic medical image recognition system. The present invention makes corresponding improvements to the technical problems mentioned in the background technology. The conveying and positioning mechanism 2 further includes a positioning unit 22. The positioning unit 22 is arranged on the top of the microbial image recognition device body 1. The positioning unit 22 is used for precise positioning of the glass slides.

[0050] The positioning unit 22 includes a sliding groove 2201. A fixing plate 2202 is slidably connected to the inner cavity of the sliding groove 2201. A connecting hole 2203 is formed in the inner wall of the sliding groove 2201. A support plate 2204 is fixedly installed on one side of the fixing plate 2202. A positioning rod 2205 is fixedly installed on the top of the support plate 2204. One end of the positioning rod 2205 penetrates into the inner cavity of the connecting hole 2203. A positioning block 2206 is threadedly connected to the surface of the positioning rod 2205. The bottom of the positioning block 2206 is in close contact with the top of the microorganism image recognition device body 1. A guiding plate 2207 is fixedly installed on the top of the microorganism image recognition device body 1;

[0051] A connecting rod 2208 is fixedly installed in the inner cavity of the sliding groove 2201. The inner wall of the fixing plate 2202 is slidably connected to the surface of the connecting rod 2208. A pressure reducing pad 2209 is fixedly installed on the surface of the fixing plate 2202. The pressure reducing pad 2209 is made of rubber.

[0052] The specific implementation manner of this embodiment is as follows: During the transmission of the glass slide, the two sides of the glass slide are blocked by the guiding plate 2207, which can prevent the glass slide from shifting during movement. The fixing plate 2202 moves in the inner cavity of the sliding groove 2201, and the docking position of the glass slide can be adjusted according to the position of the tissue sample on the glass slide. During the movement of the fixing plate 2202, the movement of the fixing plate 2202 is guided by the connecting rod 2208 to prevent the connecting rod 2208 from shifting during movement. The support plate 2204 slides in the inner cavity of the connecting hole 2203 to drive the positioning rod 2205 to move. The position of the fixing plate 2202 is fixed by the threaded connection between the positioning block 2206 and the positioning rod 2205. When the glass slide contacts the pressure reducing pad 2209, precise positioning of the glass slide is achieved, so that the tissue sample inside the glass slide can be accurately located at the shooting position, making the photographed tissue sample more accurate, avoiding manual adjustment of the position of the glass slide, achieving precise positioning of the glass slide, and achieving the effect of automatic positioning.

[0053] Embodiment III

[0054] Please refer to Figures 1-10 , the present invention provides a technical solution: A microscopic medical image recognition system. The present invention makes corresponding improvements to the technical problems mentioned in the background art. The collection mechanism 3 includes a shielding plate 31. The shielding plate 31 is slidably connected to the inner cavity of the collection box 2101. A limiting plate 32 is fixedly installed on one side of the shielding plate 31. A support frame 33 is fixedly installed at the bottom of the limiting plate 32. A clamping groove 34 is formed on one side of the limiting plate 32. A connecting rod 2208 is slidably connected to the inner cavity of the clamping groove 34. A blocking rod 35 is fixedly installed at the top of the connecting rod 2208.

[0055] A connecting block 36 is fixedly installed at the top of the collection box 2101, an elastic clip 37 is fixedly installed inside the connecting block 36, a support rod 38 is fixedly installed at the top of the shielding plate 31, and the surface of the support rod 38 is in contact with the inside of the elastic clip 37.

[0056] The specific implementation manner of this embodiment is as follows: When collecting the glass slides, move the support rod 38 out of the inside of the elastic clip 37 to separate the shielding plate 31 from the collection box 2101. Guide the installation of the glass slides through the limiting plate 32, support the bottom of the glass slides through the support frame 33, and be able to push the stacked glass slides through the sliding connection between the blocking rod 35 and the card slot 34, so that adjacent glass slides can be accurately aligned, further increasing the accuracy of the glass slides during transmission and making the tissue samples on the glass slides photographed more accurately.

[0057] Embodiment Four

[0058] Please refer to Figures 1-10 , the present invention provides a technical solution: a microscopic medical image recognition system. The present invention makes corresponding improvements to the technical problems mentioned in the background art, including the following steps:

[0059] A1: Preprocessing of training data, adjusting the size of the image to meet the input requirements of the model, and using random flipping operations with a probability of 50%. The data diversity can be increased by flipping the image horizontally or vertically.

[0060] A2: Construction of the target detection model. The Faster R-CNN network using ROIAlign is used to construct the target detection model. The specific steps of step A2 include that Faster R-CNN is a network structure based on the FPN model for target area detection, and uses the Region Proposal Network (RPN) to predict the candidate rectangular frames of the target detection areas.

[0061] A3: Construction of the semantic segmentation network. After the target detection process is completed, the rectangular frames corresponding to each target substance will be detected, the image areas corresponding to each rectangular frame will be extracted, and the image areas containing only the target substance part will be segmented using the semantic segmentation algorithm.

[0062] The specific implementation manner is as follows:

[0063] Preprocessing of training data: For all pictures, we first adjust the size of the image to meet the input requirements of the model, which is to ensure that all input images have the same size, facilitating the training and inference of the model. For training pictures, we use random flipping operations with a probability of 50%. The data diversity can be increased by flipping the image horizontally or vertically.

[0064] Construction of the target detection model: Faster-rcnn is a network structure based on the FPN model for target region detection. The model uses the Region Proposal Network (RPN) to predict the candidate rectangular boxes (bounding boxes) of the target detection regions. Compared with the pooling operation (ROIPooling) of the traditional Faster-rcnn network, ROIAlign uses bilinear interpolation to accurately sample each position on the feature map, rather than simply taking the nearest integer coordinates, which can avoid the quantization loss of information and improve the spatial accuracy of features;

[0065] When calculating the approximation degree between the candidate rectangular box and the gt box (ground-truth), that is, the accuracy of the prediction, Intersection over Union (IoU) is adopted, and its specific calculation method is as follows.

[0066]

[0067] Among them, A and B correspond to the predicted candidate rectangular box region and the gt box region respectively;

[0068] In traditional target detection methods, classification scores are usually used to measure the confidence of predictions. However, such classification scores often cannot accurately reflect the quality of the prediction boxes, resulting in the exclusion of some candidate boxes with accurate predictions but low classification scores during non-maximum suppression (NMS); by adopting the Faster-rcnn network, the quality prediction of the target position is combined with the classification score, and the IoU-related classification score (IACS, IoU-aware classification score) is used as the detection score;

[0069] The loss function used for network training includes two parts: classification loss and regression loss, and its expression is as follows:

[0070]

[0071] p i represents the probability that the i-th anchor predicted by the grid is a target. represents the corresponding ground truth. If the iou between the i-th anchor and a certain real target is greater than 0.7, then if the iou is less than 0.3, then the anchors in other cases do not participate in the training. t i represents the offset between the predicted box and the anchor box. represents the corresponding ground truth, representing the offset between the GT box and the anchor box. Ncls is the batch size, N reg is the number of anchor positions, and λ is used to balance the two loss functions;

[0072] In Faster R-CNN, the bounding box is used to represent the target region. The bounding box is usually represented by the coordinates of the upper left corner and the lower right corner, i.e., (x_min, y_min, x_max, y_max). This representation can accurately define the position and size of a rectangular box;

[0073] The candidate rectangular boxes in Faster R-CNN are generated by the RPN (Region Proposal Network). It uses anchor boxes to represent the candidate target regions. The anchor boxes are a series of predefined rectangular boxes with different sizes and aspect ratios, fixed at multiple positions on the image. The RPN will perform classification prediction and regression on each anchor box to generate candidate rectangular boxes;

[0074] For the update of the bounding box, the model will predict the position offset between it and the corresponding ground truth box. The specific update is as follows:

[0075]

[0076] where [px, py, pw, ph] represents the coordinates of the original anchor, [dx, dy, dw, dh] represents the coordinate offset predicted by the RPN network, and [gx, gy, gw, gh] represents the coordinates of the corrected anchor;

[0077] Construction of the semantic segmentation network: After the object detection process is completed, the rectangular box corresponding to each target substance will be detected. In the present invention, the image region corresponding to each rectangular box is extracted, and using the semantic segmentation algorithm, the image region containing only the target substance part is segmented, thus eliminating the operation of the doctor manually cropping the image and saving the doctor's time and energy;

[0078] The FCN algorithm is a classic semantic segmentation algorithm that can accurately segment the objects in the picture. Compared with the traditional convolutional neural network (CNN), the FCN realizes end-to-end pixel-level prediction by using fully convolutional layers. The main modules include convolution and deconvolution, that is, first perform convolution and pooling on the image to continuously reduce the size of its feature map; then perform deconvolution operation, that is, perform interpolation operation to continuously increase its feature map, and finally classify each pixel value to generate a semantic segmentation result with the same size as the input image;

[0079] The Binary Cross Entropy loss function is as follows

[0080]

[0081] Among them, y is a binary label 0 or 1, and p(y) is the probability that the output belongs to the y label. As a loss function, binary cross-entropy is used to evaluate the goodness of the prediction results of a binary classification model. That is, for the case where the label y is 1, if the predicted value p(y) approaches 1, the value of the loss function should approach 0. On the contrary, if the predicted value p(y) approaches 0 at this time, the value of the loss function is very large.

[0082] Working principle: A plurality of glass slides are vertically placed in the inner cavity of the collection box 2101. The bottom glass slide is in contact with the top of the conveyor belt 2107. The motor 2104 is started through an external power supply. The motor 2104 drives the rotating rod 2105 to rotate, which can drive the rotation of the conveyor belt 2107. The rotation of the conveyor belt 2107 can drive the driven rod 2106 to rotate, achieving the purpose of conveying the glass slide. During the conveying process, the friction between the driven rod 2106, the rotating rod 2105 and the inner wall of the moving groove 2102 is reduced through the bearing 2109, making the rotation of the rotating rod 2105 and the driven rod 2106 smoother. Through the stopper 2108, the conveyor belt 2107 is prevented from moving on the surfaces of the rotating rod 2105 and the driven rod 2106, making the conveyance of the glass slide by the conveyor belt more stable and the movement of the glass slide carrier more stable, increasing the detection accuracy. During the conveyance of the glass slide, the two sides of the glass slide are blocked by the guide plate 2207, which can prevent the glass slide from shifting during movement. The fixing plate 2202 moves in the inner cavity of the sliding groove 2201, and the docking position of the glass slide can be adjusted according to the position of the tissue sample on the glass slide. During the movement of the fixing plate 2202, the movement of the fixing plate 2202 is guided through the connecting rod 2208 to prevent the connecting rod 2208 from shifting during movement. The support plate 2204 slides in the inner cavity of the connecting hole 2203, driving the positioning rod 2205 to move. Through the threaded connection between the positioning block 2206 and the positioning rod 2205, the position of the fixing plate 2202 is fixed. When the glass slide contacts the pressure relief pad 2209, precise positioning of the glass slide is achieved, enabling the tissue sample inside the glass slide to be accurately positioned at the shooting position, making the photographed tissue sample more accurate, avoiding manual adjustment of the position of the glass slide, achieving precise positioning of the glass slide, and achieving the effect of automatic positioning. When collecting the glass slide, the support rod 38 is moved out of the inside of the elastic clip 37, and the shielding plate 31 is separated from the collection box 2101. The installation of the glass slide is guided through the limit plate 32, and the bottom of the glass slide is supported by the support frame 33. Through the sliding connection between the blocking rod 35 and the card slot 34, the stacked glass slides can be pushed, enabling adjacent glass slides to be accurately aligned, further increasing the accuracy during the conveyance of the glass slide and making the photographing of the tissue sample on the glass slide more accurate.

[0083] It should be noted that in this text, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variation thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device.

[0084] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A microbiological medical image recognition system, comprising a microbiological image recognition instrument body (1), characterized in that: A friction pad (11) is fixedly mounted on the bottom of the microorganism image recognition instrument body (1), and a transmission positioning mechanism (2) is provided on one side of the microorganism image recognition instrument body (1); The conveying and positioning mechanism (2) comprises a conveying unit (21), the conveying unit (21) is arranged on one side of the microorganism image recognition instrument body (1), and the conveying unit (21) is used to convey the slide glass; The transmission positioning mechanism (2) further comprises a positioning unit (22), wherein the positioning unit (22) is arranged on the top of the microorganism image recognition instrument body (1), and the positioning unit (22) is used for accurately positioning the glass slide; A collecting mechanism (3) is provided on one side of the microorganism image recognition instrument body (1), and the collecting mechanism (3) is used to fix the collection position of the slide, and the collecting mechanism (3) is used in conjunction with a transmission unit (21); The transmission unit (21) comprises a collection box (2101), the collection box (2101) is fixedly mounted on one side of the microorganism image recognition instrument body (1), the top of the microorganism image recognition instrument body (1) is provided with a moving groove (2102), the number of the moving grooves (2102) is two, the inner wall of the moving groove (2102) is provided with a placement groove (2103), the inner wall of the placement groove (2103) is fixedly mounted with a motor (2104), the motor (2104) 104) is fixedly mounted with a rotating rod (2105), one end of the rotating rod (2105) penetrates the inner wall of the moving groove (2102), the inner wall of the collecting box (2101) is rotatably connected with a driven rod (2106), the surface of the rotating rod (2105) is sleeved with a conveyor belt (2107), the rotating rod (2105) and the surface of the driven rod (2106) are rotatably connected through the conveyor belt (2107), and the conveyor belt (2107) is made of rubber; The positioning unit (22) comprises a sliding groove (2201), the inner cavity of the sliding groove (2201) is slidably connected to a fixing plate (2202), the inner wall of the sliding groove (2201) is provided with a connecting hole (2203), a support plate (2204) is fixedly installed on one side of the fixing plate (2202), a positioning rod (2205) is fixedly installed on the top of the supporting plate (2204), one end of the positioning rod (2205) passes through the inner cavity of the connecting hole (2203), a positioning block (2206) is threadedly connected to the surface of the positioning rod (2205), the bottom of the positioning block (2206) is in close contact with the top of the microorganism image recognition instrument body (1), and a guide plate (2207) is fixedly installed on the top of the microorganism image recognition instrument body (1); The collecting mechanism (3) comprises a shielding plate (31), the shielding plate (31) is slidably connected to the inner cavity of the collecting box (2101), a limiting plate (32) is fixedly installed on one side of the shielding plate (31), a supporting frame (33) is fixedly installed on the bottom of the limiting plate (32), a slot (34) is opened on one side of the limiting plate (32), a connecting rod (2208) is slidably connected to the inner cavity of the slot (34), and a shift rod (35) is fixedly installed on the top of the connecting rod (2208).

2. A microbiological medical image recognition system according to claim 1, characterized in that: A stopper (2108) is fixedly mounted on the surface of the rotating rod (2105) and the driven rod (2106), and the surface of the stopper (2108) is in contact with the surface of the conveyor belt (2107).

3. A microbiological medical image recognition system according to claim 1, characterized in that: The inner wall of the movable groove (2102) and the inner wall of the collecting box (2101) are both fixedly mounted with bearings (2109); the inner walls of the movable groove (2102) and the collecting box (2101) are both fixedly connected to the outer side of the outer ring of the bearing (2109); and the surfaces of the rotating rod (2105) and the driven rod (2106) are both fixedly connected to the inner side of the inner ring of the bearing (2109).

4. The microbiological medical image recognition system according to claim 1, characterized in that: A connecting rod (2208) is fixedly installed in the inner cavity of the sliding groove (2201), the inner wall of the fixing plate (2202) is slidably connected to the surface of the connecting rod (2208), and a decompression pad (2209) is fixedly installed on the surface of the fixing plate (2202), and the decompression pad (2209) is made of rubber.

5. The microbiological medical image recognition system according to claim 1, characterized in that: A connecting block (36) is fixedly mounted on the top of the collection box (2101), an elastic clip (37) is fixedly mounted on the inner side of the connecting block (36), and a supporting rod (38) is fixedly mounted on the top of the shielding plate (31), the surface of the supporting rod (38) is in contact with the inner side of the elastic clip (37).

6. The analysis method of a microbiological medical image recognition system according to claim 1, characterized in that: The following steps are involved: A1: Preprocessing training data, resizing the image to fit the model’s input requirements, using a random flipping operation with a probability of 50% to increase data diversity by flipping the image horizontally or vertically; A2: Construction of target detection model: Using ROIAlign's Fast R-CNN network to build a target detection model. The specific steps of step A2 include: Fast R-CNN is a network structure based on the FPN model for target region detection, and a candidate rectangular frame of the target detection region is predicted by using Region Proposal Network (RPN); A3: Construction of semantic segmentation network. After the target detection process is completed, the rectangular box corresponding to each target substance will be detected, and the image area corresponding to each rectangular box will be extracted. Using the semantic segmentation algorithm, the image area containing only the target substance part will be segmented out.

7. A method for using a microbiological medical image recognition system according to any one of claims 1 to 5, characterized in that: The following steps are involved: S1: placing a glass slide on the shielding plate (31), guiding the installation of the glass slide through the limit plate (32), supporting the bottom of the glass slide through the support frame (33), and pushing the stacked glass slides through the sliding connection between the gear lever (35) and the card slot (34), so that adjacent glass slides can be accurately aligned; S2: sliding the fixing plate (2202) in the inner cavity of the sliding groove (2201), adjusting the distance between the fixing plate (2202) and the precise position of the photographing according to the position of the tissue sample on the glass slide, so that the tissue sample on the glass slide can accurately enter the identification range; S3: Start the motor (2104) to convey the glass slide by driving the transmission belt. The transmission belt (2107) is located at both ends of the glass slide. The movement of the glass slide is guided by the guide plate (2207), so that the glass slide moves horizontally to the bottom of the camera of the microbial image recognition instrument body (1) and contacts the surface of the fixed plate (2202), so that the glass slide can be accurately aligned with the image recognition position of the microbial image recognition instrument body (1).

Images