Bacteriological testing equipment

Abstract

This invention provides a bacterial testing device that achieves the goal of saving manual labor, while offering high-precision detection and automatic separation of overlapping colonies, all while providing the technical benefits of robust system and interference resistance. [Solution] The device consists of an identity recognition unit 1, an adjustable light source unit 2, an image capture unit 3, a bacterial inspection dish information recognition unit 4, a system unit 5, and a database unit 6. The bacterial inspection device 100 according to this invention eliminates the problems of complex coding and operation by combining radio frequency identification (RFID) labels with the bacterial inspection procedure, and also achieves the objective of reducing human labor consumption. Furthermore, the system can effectively shorten working time through automated shooting, archiving, and identification functions, thereby improving the overall efficiency of bacterial inspection.

Description

Technical Field

[0001] The present invention relates to a bacteria testing apparatus, in particular to a bacteria testing procedure incorporating Radio Frequency Identification (RFID) labels, more particularly capable of solving the problems of complicated coding and operation, achieving the object of saving manual labor consumption, and obtaining the technical effects of system robustness and anti-interference performance while having the characteristics of high-precision detection and automatic separation of overlapping colonies. Background Art

[0002] In the existing biotechnology industry, bacteria testing work is mostly performed manually, and the process is complicated and time-consuming, which requires a large amount of human resources and reduces work efficiency. However, since bacteria testing is an indispensable part of the biotechnology industry, it is necessary to ensure a safe and sterile condition for consumables, raw materials, products, environments and personnel.

[0003] The existing manual bacteria testing process requires complicated bacteria testing coding and operation steps, which further increases manual labor consumption and working time. In addition to manual processes, there are conventional bacteria detection devices and automatic colony counting devices applied to bacteria testing work, most of which use conventional image processing technologies (such as binarization, edge detection or morphological processing), so there are the following technical problems in practical use.

[0004] 1. Due to insufficient environmental adaptability, counting errors of colonies are easily caused by reflection from a medium light source, shadow at the edge of a culture dish or interference from impurities.

[0005] 2. Because it is difficult to divide overlapping colonies, when colonies grow in clusters and their edges overlap, conventional algorithms can only divide them according to the features of a low-order grayscale gradient. Furthermore, if the boundary contrast is insufficient, even if there are discernible contour cues in the image, the fixed threshold algorithm is not effectively utilized, resulting in many colonies being misidentified as a single large colony and the count being underestimated.

[0006] 3. Because parameter adjustment is cumbersome, operators must repeatedly and artificially calibrate the threshold parameters for different types of microorganisms or different color characteristics of culture media, making it impossible to use it automatically and widely.

[0007] Furthermore, when applying conventional convolutional neural networks to colony detection, the inference process is cumbersome and the post-processing complexity increases because the anchor frame design must be defined in advance, and non-maximal suppression must be performed as a necessary post-processing step.

[0008] As described above, existing bacterial testing technologies, such as artificial processes, conventional bacterial detection devices, and automated colony counting devices, all suffer from problems such as insufficient environmental adaptability, difficulty in colony segmentation, and complicated parameter adjustments. Deep learning methods are limited due to the complexity of their inference processes. Because these technical bottlenecks make it difficult to achieve both efficiency and accuracy in bacterial testing, existing technologies cannot resolve these problems. Therefore, there is a need for a novel technological solution that possesses high environmental adaptability, effectively segments overlapping colonies, and reduces parameter dependence, thereby improving the reliability and automation level of bacterial testing in the biotechnology industry.

[0009] To overcome the aforementioned shortcomings, the inventor has conducted careful research and utilized scientific principles to propose this invention, which effectively eliminates the aforementioned drawbacks and has a rational design. [Disclosure of the Invention] [Problems that the invention aims to solve]

[0010] The main objective of this invention is to provide a bacterial testing device that can solve the problems of conventional methods by performing a bacterial testing procedure that incorporates RFID labels, thereby reducing manual labor while eliminating complex coding and operational problems. Furthermore, it can effectively shorten working time through automated imaging, archiving, and identification functions, thereby improving the overall efficiency of bacterial testing.

[0011] Another objective of this invention is to provide a bacterial testing device that features high-precision detection and automatic overlap separation, and that can still effectively utilize image contour cues to precisely identify overlapping colonies in a crowded environment, even when the lighting is not uniform or the contrast of colony boundaries is relatively low, thereby reducing underestimation of counts.

[0012] Another objective of this invention is to provide a bacterial testing device that is robust and interference-resistant, can automatically ignore scratches on bacterial testing dishes, scratches on culture media, bubbles, and uneven light interference, significantly reducing the need for human parameter adjustment and enabling a highly automated detection process. [Means for solving the problem]

[0013] This invention is a bacterial testing apparatus for achieving the above objectives, comprising: an identity recognition unit provided in the case that identifies and records the identity of the operator each time, thereby allowing investigation of the operator inputting each record; an adjustable light source unit provided in the case that adjusts different light sources based on the size of different bacterial testing dishes; an image capture unit provided in the case that automatically captures video material of colonies in each bacterial testing dish; a bacterial testing dish information recognition unit provided in the case near the bottom or sides of the bacterial testing dishes, each bacterial testing dish having its own identity recognition unit, each identity recognition unit identifying the contents of each bacterial testing dish and recording them in the corresponding bacterial testing record; and an AI recognition unit provided in the case, which is the core of information transmission and control of each unit, in which the identity recognition unit, the adjustable light source unit, the image capture unit, and the bacterial testing dish information recognition unit are electrically connected, and which includes an AI model training module. The system unit includes a training module which collects video material of multiple colonies captured by the image capture unit and labels the colonies based on their contours and brightness distribution to obtain labeled video material of multiple colonies, and a deep learning object detection model which is trained based on the labeled video material of the colonies to generate a colony discrimination model, and an AI inference module which performs AI inference based on the colony discrimination model, and an image capture unit which captures video material of colonies that are the content of the required bacterial examination, and a system unit which transmits the video material of the colonies that are required for bacterial examination to the colony discrimination model, and can effectively distinguish between true microbial colonies and background noise based on global context-related features of the image acquired by the self-attention mechanism, and can infer colony localization and counting by integrating multiscale features of different sensory fields of view, and then a feedback mechanism which is established to update and correct the results of the bacterial examination identified by the AI.

[0014] According to an embodiment of the present invention, the adjustable light source unit is comprised of a spectral light source board and a uniform light panel installed between the spectral light source board and the bacterial inspection dish.

[0015] According to an embodiment of the present invention, the identity recognition unit has an identity and encoding module that provides information on the operator of the bacterial test and the operator who collects samples for the bacterial test, records the identity of the operator at each stage, and has a single coding that binds to the content of the bacterial test, thereby ensuring that the content of each bacterial test does not overlap or have typographical errors, so that the content of the bacterial test matches the record of the corresponding bacterial test, and thus has traceability corresponding to the content of the bacterial test.

[0016] According to an embodiment of the present invention, the system unit is further connected to a database unit that stores the results of the bacterial test, manages the operator information recorded by the identity and encoding module and the contents of the bacterial test, and supports subsequent questions, verification, and tracking.

[0017] According to the embodiment of the present invention, the identity recognition unit is a card reader.

[0018] According to an embodiment of the present invention, the identity authentication unit is one or a combination of a QR code (registered trademark) (QR Code), a radio frequency identification (RFID) label, a near-field communication (NFC) label, and a beacon.

[0019] According to the embodiment of the present invention, the image capture unit is autofocus.

[0020] According to the embodiment of the present invention, the AI ​​recognition unit is either built into the system unit or connected externally.

[0021] According to an embodiment of the present invention, the AI model training module is further configured to re-annotate and update images with identification errors based on user feedback, then send the images to the deep learning object detection model for retraining to optimize the model.

[0022] According to an embodiment of the present invention, the AI inference module directly outputs a prediction set including colony positions and confidence levels through an ensemble prediction mechanism.

[0023] According to an embodiment of the present invention, the background noise includes scratches on bacterial test dishes, medium cracks, bubbles, and uneven illumination.

[0024] According to an embodiment of the present invention, the bacterial test dish information recognition unit has a bacterial test recording interface, and is provided with a bacterial test recording module that groups different contents requiring bacterial testing and provides special coding recorded in identification data to correspond to bacterial test records.

[0025] According to an embodiment of the present invention, the feedback mechanism for updating and correcting AI-identified bacterial test results is provided in the bacterial test recording module.

[0026] Hereinafter, the features and technical content of the present invention will be described in detail with reference to the drawings. However, these drawings are for reference and explanation only, and the present invention is not limited thereto. Best Mode for Carrying Out the Invention

[0027] Figures 1 to 10 are respectively a three-dimensional conceptual diagram of the bacterial inspection apparatus according to the present invention, a three-dimensional conceptual diagram of the bacterial inspection apparatus according to the present invention from another angle, an enlarged conceptual diagram of partial elements of the bacterial inspection apparatus according to the present invention, a conceptual diagram of the block architecture of the first embodiment of the bacterial inspection apparatus according to the present invention, a conceptual diagram of the block architecture of the second embodiment of the bacterial inspection apparatus according to the present invention, a conceptual diagram of the bacterial inspection process of the bacterial inspection apparatus according to the present invention, a conceptual diagram of the recording interface for bacterial inspection of the present invention, a conceptual diagram of when selecting the type of bacterial inspection and the manufacturing / effective date from the interface of the present invention, a photograph when performing model optimization through AI label feedback according to the present invention, and a photograph of the bacterial inspection result of AI inference according to the present invention. As shown in the figure, the bacterial inspection apparatus 100 according to the present invention is composed of an identity recognition unit 1, an adjustable light source unit 2, an image capture unit 3, a bacterial culture dish information recognition unit 4, a system unit 5, and a database unit 6.

[0028] The above-mentioned identity recognition unit 1 is provided on a casing 10, and identifies and records the identity of an operator each time an operation is performed, so that the input operator of each record can be traced.

[0029] The adjustable light source unit 2 is provided on the casing 10, and adjusts different photographing light sources based on sizes of different bacterial culture dishes 7. The adjustable light source unit 2 is composed of a spectral light source board 21 and a uniform light panel 22, wherein the uniform light panel 22 is arranged between the spectral light source board 21 and the bacterial culture dish 7.

[0030] The image capture unit 3 is provided on the casing 10, and automatically captures video materials of colonies on each bacterial culture dish 7.

[0031] The bacterial culture dish information recognition unit 4 is provided on the casing 10 so as to be located near the bottom or side of the bacterial culture dish 7, each bacterial culture dish 7 is provided with an identity authentication unit 71, and each identity authentication unit 71 identifies the content of each bacterial culture dish 7 and records the content in the corresponding bacterial inspection record.

[0032] The system unit 5, which is the core of the bacterial inspection device 100, is installed in the case 10 and is electrically connected to the identity recognition unit 1, the adjustable light source unit 2, the image capture unit 3, and the bacterial inspection dish information recognition unit 4, and is the core for information transmission and control of each unit. The system unit 5 includes an AI recognition unit 51 which contains an AI model training module 511 and an AI inference module 512. The AI ​​model training module 511 collects video material of multiple colonies captured by the image capture unit 3 and labels the colonies based on their contours and brightness distribution, thereby obtaining video material of multiple labeled colonies. Based on this video material of labeled colonies, a deep learning object detection model is trained. First, based on user feedback, images with identification errors are relabeled and updated, then sent back to the deep learning object detection model for retraining and model optimization, finally generating a colony discrimination model. The AI ​​inference module 512 is configured to perform AI inference based on this colony discrimination model, and the image capture unit 3 detects bacteria. The system unit 5 captures video footage of colonies requiring inspection and transmits this video footage to a colony discrimination model. The colony discrimination model uses global feature extraction and attention mechanisms to effectively distinguish between true microbial colonies and background noise (e.g., scratches on the bacterial inspection dish, scratches on the culture medium, bubbles, and uneven lighting) based on global context-related features of the image obtained from the attention mechanism, thereby reducing the false positive rate. Furthermore, the colony discrimination model incorporates multiscale features and sensory field optimization, enabling it to integrate multiscale features from different sensory fields, as there may be size imbalances or highly crowded colonies in the target image. This allows for precise localization and counting of minute colonies in high-density collection environments, and establishes a feedback mechanism that allows for updating and correcting the results of bacterial inspections identified by AI.

[0033] The database unit 6 is electrically connected to the system unit 5, stores the results of the bacterial test, manages the operator information and bacterial test details recorded by the identity and encoding module, and supports subsequent inquiries and tracking. The above elements constitute the new bacterial testing device.

[0034] According to a better specific embodiment of the present invention, the identity recognition unit 1 is provided with an identity and encoding module 11 that records the identity of the operator at each stage, and is bound to the required bacterial test content, ensuring that the content of each bacterial test does not overlap or have typographical errors, and that the content of the bacterial test matches the record of the corresponding bacterial test, thus having a unique coding that provides traceability corresponding to the content of the bacterial test.

[0035] According to a better specific embodiment of the present invention, the bacterial inspection dish information recognition unit 4 is provided with a bacterial inspection recording module 41 that provides a bacterial inspection recording interface, groups different contents requiring bacterial inspection, provides special coding, and records them in identification data, thereby enabling the bacterial inspection dish information recognition unit 4 to correspond to bacterial inspection recordings.

[0036] According to a better specific embodiment of the present invention, the feedback mechanism is provided in the bacterial test recording module 41, which updates or corrects the results of the AI-identified bacterial test.

[0037] According to a better specific embodiment of the present invention, the bacterial testing device 100 has an end-to-end detection structure and does not utilize conventional convolutional neural networks that require prior design to define an anchor frame design. Instead, the AI ​​inference module 512 utilizes an ensemble prediction mechanism to directly output a prediction set that includes colony locations and confidence levels, and does not rely on non-maximal suppression, which is a necessary post-processing step. This simplifies the inference process and reduces the complexity of post-processing.

[0038] According to a better specific embodiment of the present invention, the AI ​​recognition unit 51 is built into the system unit 5, as shown in Figure 4.

[0039] According to a better specific embodiment of the present invention, the AI ​​recognition unit 51 is connected from outside the system unit 5, as shown in Figure 5.

[0040] According to this invention, the identity recognition unit 1 is a card reader, the image capture unit 3 is an autofocus, and the identity authentication unit 71 is one or a combination of a QR code (registered trademark), a radio frequency identification (RFID) label, a near-field communication (NFC) label, and a beacon.

[0041] The following examples merely illustrate the details and content of the present invention, but the scope of the claims for registration of the present invention is not limited thereto.

[0042] According to this invention, the bacterial testing process is as follows: Step S11 is preparatory work before bacterial testing, as shown in Figure 6, and steps S12 to S16 constitute the main process of bacterial testing. In step S111, the operator reads an identification card (e.g., employee ID) with the card reader of the bacterial testing device 100 and enters the bacterial testing record interface, where the time of bacterial testing and the bacteriologist are recorded, as shown in Figure 7. In step S112, the type of bacterial testing and the manufacturing date / expiration date are selected at the interface. Also, as shown in Figure 8, the type of bacterial testing dish and collection method are pre-set based on the type of bacterial testing dish, so only the information needs to be confirmed. In step S113, the sample is divided into dishes according to the bacterial testing operation process, and a cross mark is displayed on the surface of the bacterial testing dish, which is advantageous for follow-up observation and interpretation of colony distribution. The above dish division is generally distinguished based on the sample date, and depending on the situation, dish division is performed based on the cell source. Then, in step S114, the RFID reader head reads the information of the bottles that are to be grouped for bacterial testing, and the read group information is written to the RFID of the corresponding bacterial testing dish. By binding the sample group and the bacterial testing dish, the system supports the management and tracking of subsequent bacterial testing records. In step S115, after reading the sample, the operator executes the action of writing to the RFID on the interface. In step S116, the system determines whether the writing was successful, and in step S117, if the system indicates failure, it returns to step S115 and performs the writing again. If the writing is successful, in step S118, the time of the bacterial testing is recorded, success is indicated on the interface, the corresponding item text is changed to blue, and green is displayed in the lower right corner of the interface, allowing the operator to confirm the result of the operation. In step S119, after the operator has successfully written the information, they perform smearing on the dish using a glass rod or a straw to draw lines and perform culture and observation of the colony distribution.In step S120, after the smearing operation is complete, the sample bottles are sent to the refrigerator by a carry cart to await the results of the bacterial test. In step S121, the recycled bottles are sent to the refrigerator for storage. Finally, in step S122, the bacterial test dishes with lines drawn on them are placed in an incubator and cultured for two days (approximately 48 hours) at 37°C ± 20% to support the interpretation and analysis of follow-up colony growth.

[0043] In step S12, during the process of interpreting the results of the bacterial test, the operator reads the identification card with the bacterial testing device 100, confirms the interpreter's information, and records it.

[0044] In step S13, the operator places the bacterial sample dish into the bacterial testing device 100, captures an image, and records the interpretation time.

[0045] In step S14, after capturing an image of the bacterial inspection dish, the colony discrimination model performs AI inference on the captured colony video material. The self-attention mechanism acquires global context-related features of the image to distinguish between true microbial colonies and background noise. It also integrates multiscale features from different sensory fields of view to infer colony localization and counting, and constructs a feedback mechanism to update and correct the AI-inferred bacterial inspection results. As shown in Figure 9, the system corrects and updates identification errors and incomplete labeling data by providing feedback with AI labels. This retrains the colony discrimination model and optimizes the model, and the results of the AI-inferred bacterial inspection are displayed in Figure 10.

[0046] In step S15, during the process of verifying the bacterial test record, the operator uses the bacterial test record established by the PMS reagent to perform a data search and obtain and record information such as the reading time, the interpreter, and the bacteriologist.

[0047] In step S16, relevant data generated during bacterial testing and interpretation is stored in a database for centralized management, supporting subsequent inquiries, verification, and tracking.

[0048] This invention is a colony discrimination model obtained by introducing the deep learning object detection models described above, and achieves the following technical effects.

[0049] Its key features are high-precision detection and automatic overlap separation. The colony discrimination model can learn the entire visual characteristics of the colony, including irregularities in contour lines and subtle differences in brightness distribution, through training with a large number of colony images. Compared to conventional algorithms that rely on fixed thresholds, this device can effectively utilize image contour cues to precisely identify overlapping colonies in crowded environments, even when the lighting is not uniform or the contrast of the colony boundaries is relatively low, thereby reducing underestimation of counts.

[0050] Its strengths lie in its robustness and interference resistance. The colony discrimination model can autonomously learn the general visual characteristics of colonies and automatically ignore scratches on the bacterial sample dish, scratches on the culture medium, bubbles, and uneven light interference, significantly reducing the need for human parameter adjustment and realizing a highly automated detection process.

[0051] As described above, the bacterial testing device according to the present invention effectively overcomes the drawbacks of conventional methods. Since it can perform multiple existing artificial bacterial testing processes using a bacterial testing device that incorporates RFID labels, it solves the problems of complex coding and operation in conventional bacterial testing processes, thereby achieving the objective of reducing human labor consumption. Furthermore, the automated functions of the system, such as imaging, archiving, and identification, effectively shorten working time and improve the overall efficiency of bacterial testing. Therefore, this invention is more progressive and practical, and we will file a claim for utility model registration in accordance with the law.

[0052] The above are merely better embodiments of the present invention, and the present invention is not limited thereto. All equivalent changes and modifications made based on the claims and specifications relating to the present invention are included within the scope of the claims of the present invention. [Brief explanation of the drawing]

[0053] [Figure 1] This is a three-dimensional conceptual diagram of the bacterial testing apparatus according to the present invention. [Figure 2] This is a three-dimensional conceptual diagram of the bacterial testing apparatus according to the present invention from another angle. [Figure 3] This is a magnified conceptual diagram of some elements of the bacterial testing device according to the present invention. [Figure 4] This is a conceptual diagram of the block architecture of the first embodiment of the bacterial testing device according to the present invention. [Figure 5] This is a conceptual diagram of the block architecture of a second embodiment of the bacterial testing device according to the present invention. [Figure 6] This is a conceptual diagram of the bacterial testing process of the bacterial testing apparatus according to the present invention. [Figure 7] This is a conceptual diagram of the bacterial testing recording interface according to the present invention. [Figure 8] This is a conceptual diagram of the interface according to the present invention, showing the selection of the type of bacterial test and the manufacturing / expiration date. [Figure 9] This is a photograph showing the process of optimizing a model using AI label feedback according to the present invention. [Figure 10] This is a photograph of the results of a bacterial test performed using AI inference according to the present invention. [Explanation of symbols]

[0054] 1. Identity Recognition Unit 10 cases 100 Bacteriological testing equipment 11 Identity and Encoding Module 2. Adjustable light source unit 21 Spectrum Light Source Board 22 Uniform Light Panel 3 Image Capture Unit 4. Bacterial Inspection Dish Information Recognition Unit 41. Bacteriological testing recording module 5 System Units 51 AI Recognition Unit 511 AI Model Training Module 512 AI Inference Modules 6 Database Units 7. Bacterial testing dish 71 Identity Verification Unit S11~S16 Step

Claims

1. The case includes an identity recognition unit that identifies and records the identity of each operator, allowing for the investigation of the operator who entered each record. The case is equipped with an adjustable light source unit that adjusts different light sources based on the size of different bacterial testing dishes, The case is equipped with an image capture unit that automatically captures video footage of colonies in each bacterial test dish, In this case, a bacterial testing dish information recognition unit is provided near the bottom or sides of the bacterial testing dish, each bacterial testing dish having its own identification unit, each identification unit identifies the contents of each bacterial testing dish, and records this in the corresponding bacterial testing record, and In the case, an AI recognition unit is provided, which is the core for information transmission and control of each unit, and which includes an AI model training module and an AI model training module, wherein the AI ​​model training module collects video material of multiple colonies captured by the image capture unit and labels the colonies based on their contours and brightness distribution to obtain labeled video material of multiple colonies, and a deep learning object detection model is trained based on the labeled video material of the colonies to detect the colonies The system unit includes: generating a discrimination model; performing AI inference based on the discrimination model of the colony in the AI ​​inference module; capturing video footage of the colony containing the required bacterial test content in the image capture unit; transmitting the video footage of the colony requiring bacterial test to the discrimination model of the colony in the system unit; effectively distinguishing between true microbial colonies and background noise based on global context-related features of the image acquired by the self-attention mechanism; inferring colony localization and counting by integrating multiscale features of different sensory fields of view; and establishing a feedback mechanism to update and correct the results of the bacterial test identified by the AI. A bacterial testing device characterized by the following features.

2. The bacterial testing apparatus according to claim 1, characterized in that the adjustable light source unit is composed of a spectral light source board and a uniform light panel installed between the spectral light source board and the bacterial testing dish.

3. The bacteriological testing apparatus according to claim 1, characterized in that the identity recognition unit has an identity and encoding module that provides information on the bacteriological testing operator and the bacteriological testing sample collection operator and records the identity of the operator at each stage, and has a single coding that binds to the content of the bacteriological testing, thereby ensuring that the content of each bacteriological testing does not overlap or is not mistyped, so that the content of the bacteriological testing matches the record of the corresponding bacteriological testing and thus has traceability corresponding to the content of the bacteriological testing.

4. The bacterial testing apparatus according to claim 3, further comprising a system unit connected to which a database unit is connected for storing the results of the bacterial testing, managing the operator information recorded by the identity and encoding module and the contents of the bacterial testing, and supporting subsequent questioning, verification, and tracking.

5. The bacterial testing apparatus according to claim 1, characterized in that the identity recognition unit is a card reader.

6. The bacterial testing apparatus according to claim 1, characterized in that the identity authentication unit is one or more combinations of a QR code (registered trademark), a radio frequency identification (RFID) label, a near-field communication (NFC) label, and a beacon.

7. The bacterial inspection apparatus according to claim 1, characterized in that the image capture unit is an autofocus camera.

8. The bacterial inspection apparatus according to claim 1, characterized in that the AI ​​recognition unit is either built into the system unit or connected externally.

9. The bacterial inspection apparatus according to claim 1, further characterized in that the AI ​​model training module is configured to re-annotate and update images with identification errors based on user feedback, and then feed these updates to the deep learning object detection model for retraining and model optimization.

10. The bacterial testing apparatus according to claim 1, characterized in that the AI ​​inference module outputs a prediction set that directly includes the colony location and confidence level using an ensemble prediction mechanism.

Images