Excrement detection target object identification AI training method, calculation processing device and storage device
By diluting and staining the feces, exposing and distinguishing the targets, combining dilution and enrichment technology, the problems of difficult targets and impurities interference in the feces are solved, and efficient AI recognition and counting effects are achieved.
Patent Information
- Application Number
- CN202410136751.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2023-12-09
- Filing Date
- 2024-01-31
- Publication Date
- 2025-06-10
AI Technical Summary
The prior art is difficult to effectively identify and count the content of target substances in feces, especially because the target substances in feces are low in relative solids and there are a large number of interfering substances, which makes AI training difficult and experimental workloads high.
By diluting the feces, it is dispersed, the target object is exposed, and the target object is clearly distinguished from the background by staining. The stained pictures are taken for AI training to obtain an effective feature data set. At the same time, the concentration of target substances is increased through dilution and enrichment techniques and impurity interference is reduced.
The natural suspension and significant distinction of target substances for feces detection are achieved, the AI recognition rate is improved, the experimental workload is reduced, and the problems of low target substance content and impurities are solved.
Smart Images
Figure CN120125867A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the technical field of analysis of formed components based on microscopic magnified images, and particularly relates to an AI training method for identifying fecal test targets, as well as a computing and processing and storage device. Background Art
[0002] It is one of the routine clinical laboratory examination items. Through this examination, some pathological phenomena of the gastrointestinal tract can be more intuitively understood, and the functional status of the digestive tract, pancreas, and hepatobiliary can be indirectly judged. It is divided into macroscopic general character observation, microscopic examination, and chemical examination.
[0003] Cells can be found in microscopic examination. Occasional white blood cells are seen in normal feces, and no red blood cells. In enteritis, the number of white blood cells is less than 15 per high-power field; in acute bacillary dysentery, it is more than 15 per high-power field, and even the field of view is full. Red blood cells can be seen in inflammation of the lower intestine (such as colitis, bacillary dysentery) and bleeding (such as polyps, tumors, hemorrhoids, etc.). Eosinophils can be seen in allergic enteritis and intestinal parasite infections, accompanied by Charcot-Leyden crystals; macrophages can be seen in bacillary dysentery and rectal inflammation; cancer cells can sometimes be found in the feces of colorectal cancer patients.
[0004] Food residues can be found in microscopic examination. A small amount of starch granules, muscle fibers, and fat droplets are normally visible. If their amount increases, it indicates poor digestion and absorption, which is more common in chronic pancreatitis and pancreatic insufficiency (such as pancreatic head cancer, etc.).
[0005] Intestinal yeasts can be found in microscopic examination. Human yeasts and common yeasts are normally visible, and Candida albicans can be seen when there is intestinal flora imbalance.
[0006] Parasites can be found in microscopic examination. When the human body is infected with different parasites, the corresponding eggs can appear in the feces, and common ones include Ascaris eggs, hookworm eggs, pinworm eggs, Clonorchis sinensis eggs, Fasciolopsis buski eggs, and Entamoeba trophozoites, etc.
[0007] When routinely examining feces under a microscope, it is difficult to detect by manually identifying the number of test targets in the high-power field of view. The key is that there is no quantitative standard, and many substances in fecal test targets cannot be identified and counted, and the technology is backward.
[0008] The applicant has proposed a series of Chinese patents, such as
[0009] 1. CN2020112669290, "Cell Analysis Method and System and Quantitative Method and System";
[0010] 2. CN2020112669182, "Cell Suspension Sample Imaging Method and System and Kit";
[0011] 3. CN2022104799126, “Fast focusing method of microscopic image acquisition device and microscopic image acquisition method”;
[0012] 4. CN2023110496905, “A device, chip and method for detecting formed components of blood, urine and feces (multi-parameters)”
[0013] Use a new technical solution to measure the content of target substances in feces.
[0014] Artificial intelligence image recognition technology is an image processing technology based on artificial intelligence algorithms. It recognizes and understands image content by analyzing and learning images.
[0015] Data assets refer to data resources that are owned or controlled by individuals or enterprises and can bring future economic benefits to enterprises and are recorded in physical or electronic form. Data assets are data sets in cyberspace that have data ownership (exploration rights, use rights, ownership rights), are valuable, measurable, and readable.
[0016] The target object to be detected in the feces detection target is low relative to the solid feces, such as Figure 13 In the process, a large amount of fecal residue interferes with recognition, and the content of many components is zero. However, AI recognition requires samples to train the AI model. How to collect enough samples is the key to the application of fecal AI recognition technology.
[0017] How to make AI training samples is a blank in the industry. Although the applicant has carried out AI recognition research on blood cells earlier, the model and feature data of blood AI recognition have low recognition rate in the feces recognition scenario and have no application value at all. To carry out AI detection of fecal suspension, how to train AI model is a technical problem that needs to be solved urgently. Figure 14 , the content of many targets to be detected is extremely low, and it is very difficult to collect the targets to be detected. Conventional technical means are almost impossible to achieve. The AI training technology in blood recognition technology, especially sample processing, is completely different and requires re-development and a lot of experimental verification, such as Figure 14, in feces, parasite eggs, intestinal protozoa, and various microorganisms need to collect samples for AI training. When conducting AI training, special treatment of the samples is required, rather than simply taking pictures of the samples. For example, there are a large number of living organisms in feces. In microscopic images, the movement of living organisms causes the deformation of the graphics, and the recognition rate of the captured pictures is very low and cannot be used. Even if the ratio of the fecal suspension is incorrect and the target sample cannot be in a natural suspension state, it will affect the recognition rate. To obtain various pathogenic samples, a large number of samples need to be collected, and the training workload is very large, resulting in the abnormal development of the industry. The applicant has collected a large number of animal fecal samples, conducted a large number of experiments, accumulated a vast amount of data, and gradually initiated the application of AI technology in fecal detection. How to reduce the difficulty of AI training for fecal detection and reduce the experimental workload is the technical problem to be solved in this application. Summary of the Invention
[0018] In this application, the inventor proposes to dilute the fecal detection target in a suspended state, so that the feces are dispersed, which can effectively expose the fecal detection target and enable the fecal detection target to be naturally suspended. Through staining, the fecal detection target can be distinguished from the background. Using the stained pictures for AI training, the obtained AI feature dataset A can effectively support the AI detection of the fecal detection target in the suspended state feces. By first diluting the feces and then enriching them, the detection target is aggregated, the concentration of the fecal detection target is increased, but the concentration of impurities is not increased, making it possible to find the fecal detection target.
[0019] An AI training method for identifying fecal detection targets includes: preprocessing feces to obtain fecal samples; the fecal preprocessing includes diluting the feces with a diluent; spreading the fecal samples flat to allow the fecal detection targets to be suspended in the liquid and present a natural state; taking pictures of the fecal samples to obtain pictures; using the obtained pictures for AI training to obtain the fecal detection target feature dataset A.
[0020] The above-mentioned AI training method for identifying fecal detection targets includes: the fecal preprocessing includes staining the fecal samples, and a staining agent is added to the feces for the staining treatment of the fecal samples.
[0021] The above-mentioned AI training method for identifying fecal detection targets includes:
[0022] The staining agent is dissolved in the diluent.
[0023] The above-mentioned AI training method for identifying fecal detection targets includes:
[0024] The dilution treatment is carried out with reference to a turbidimetry card to dilute the feces within a set concentration range.
[0025] The above-mentioned AI training method for identifying fecal detection targets includes: adding a staining diluent in a set proportion to the fecal sample obtained by turbidimetric dilution, and the staining diluent includes a stain and / or an insecticide.
[0026] The above-mentioned AI training method for identifying fecal detection targets enriches the obtained fecal sample, and the enrichment is precipitation enrichment and / or centrifugal enrichment; after enrichment, the suspension in the layer where the fecal detection target is located is taken as the fecal sample.
[0027] According to the AI training method for identifying fecal detection targets described in claim 6, the fecal detection target is a lipid droplet, and the lipid droplet floats on the upper layer of the fecal sample. The upper layer sample is taken to obtain the enriched sample; the fecal sample is photographed by photographing the upper layer of the sample to obtain a picture.
[0028] The above-mentioned AI training method for identifying fecal detection targets includes image preprocessing and AI training processing. The image preprocessing outputs the labeled fecal detection targets, and the AI training processing outputs the fecal detection target feature dataset according to the labeled fecal detection targets.
[0029] The above-mentioned AI training method for identifying fecal detection targets includes live insect shaping before fecal pretreatment.
[0030] The above-mentioned AI training method for identifying fecal detection targets, the live insect shaping process includes any one of the following technical features:
[0031] TA10: Shaping the live insect by refrigeration, and restoring to room temperature after refrigeration. The refrigeration temperature is higher than 2 degrees and lower than 8 degrees;
[0032] TA20: Shaping the live insect by heating, and restoring to room temperature after heating. The heating temperature is higher than 39 degrees and lower than 60 degrees.
[0033] According to the AI training method for identifying fecal detection targets described in claim 9, the live insect shaping is to shape the live insect or eggs by a medicament.
[0034] The above-mentioned AI training method for identifying fecal detection targets, the medicament includes any one or more of formic acid, acetic acid, propionic acid, malonic acid, butyric acid, methanol, ethanol, formaldehyde, acetaldehyde, propionaldehyde, butyraldehyde, n-valeraldehyde, glutaraldehyde.
[0035] Any of the above fecal test target recognition AI training methods further includes using the fecal test target feature dataset A as the feature dataset, using AI software to recognize one or more pictures of the said pictures to obtain marked pictures, where the marked pictures include the marked fecal test targets, performing manual review on the marked pictures to form manually reviewed marked pictures, and using the manually reviewed marked pictures for AI training to obtain the fecal test target feature dataset B.
[0036] Any of the above fecal test target recognition AI training methods, where the fecal test targets include any one of intestinal protozoa, Ascaris eggs, hookworm eggs, pinworm tapeworms, tapeworm eggs, intestinal protozoa, pathogenic microorganisms, cells, starch granules, lipid droplets, plant fibers, muscle fibers, and unknown tangible substances.
[0037] A computing device that includes the above fecal test target recognition AI training method.
[0038] A computing device, where the memory of the computing device includes the above fecal test target AI feature dataset A or fecal test target AI feature dataset B.
[0039] A data storage device, where the data storage includes the above fecal test target AI feature dataset A or fecal test target AI feature dataset B.
[0040] The technical effects of the above technical solutions include: making the fecal test targets suspended in the liquid, presenting a natural state; taking pictures of fecal samples to obtain pictures, and having a high recognition rate for the same suspended detection application scenarios.
[0041] The technical effects of the above technical solutions include: by allowing the target to be detected to naturally suspend and unfold, its morphological features can be fully unfolded to form an effective sample, and there can be a sufficient recognition rate for training. Without a naturally unfolded sample, manual annotation cannot be standardized, and the AI training work cannot be carried out.
[0042] The technical effects of the above technical solutions include: through staining, the target detection objects can be distinguished from background impurities, and the targets can be found during manual annotation.
[0043] The technical effects of the above technical solutions include: diluting and staining at the same time, pre-configuring the stain and diluent into a staining diluent with a certain concentration, which can effectively control the concentration and uniformity of the stain.
[0044] The technical effects of the above technical solutions include: diluting with reference to the turbidimetry card, diluting the feces within the set concentration range, so that the impurity distribution of different pictures for AI training is the same or similar.
[0045] The technical effects of the above technical solutions include: it can be diluted first, and then burned proportionally after dilution, and the dye concentration of dyeing can be controlled.
[0046] The technical effects of the above technical solutions include: the obtained fecal samples are enriched, and the enrichment methods include precipitation enrichment and / or centrifugation enrichment; after enrichment, the suspension in the layer where the target substance is located in the fecal test is taken as the fecal sample, thus reducing the difficulty of finding the target.
[0047] The technical effects of the above technical solutions include: before the pretreatment of feces, it includes the shaping treatment of live worms to prevent the interference of live worms on the quality of the captured pictures.
[0048] The technical effects of the above technical solutions include: the live worms are shaped by refrigeration or heating, with low cost and convenient treatment.
[0049] The technical effects of the above technical solutions include: the live worms are shaped by refrigeration or heating, within a specific temperature range, without affecting the properties of the proteins therein and changing the external morphology.
[0050] The technical effects of the above technical solutions include: the live worms or eggs are shaped by medicaments, and the operation is simple and effective.
[0051] The technical effects of the above technical solutions include: diluting first and then enriching can ensure the removal of impurities, enabling the target substance to aggregate, and at the same time ensuring that the impurities do not interfere with imaging.
[0052] The technical effects of the above technical solutions include: there are many unknown components in feces. For suspected substances, the unknown tangible substances are marked, which can aggregate the unknown substances and facilitate medical research or identification by others. Description of the Drawings
[0053] Figure 1 It is a schematic diagram of the steps of an AI training method for identifying fecal test target substances Figure 1 ;
[0054] Figure 2 It is a schematic diagram of the steps of an AI training method for identifying fecal test target substances Figure 2 ;
[0055] Figure 3 It is a schematic diagram of the steps of an AI training method for identifying fecal test target substances Figure 3 ;
[0056] Figure 4 It is a schematic diagram of the steps of staining and dilution in fecal pretreatment;
[0057] Figure 5 It is a schematic diagram of the steps of enrichment after staining and dilution in fecal pretreatment;
[0058] Figure 6It is a schematic diagram of the steps for enrichment after quantitative staining of feces diluted with a reference turbidimetry card;
[0059] Figure 7 It is a schematic diagram of the steps for enrichment after diluting and killing insects in feces diluted with a reference turbidimetry card;
[0060] Figure 8 It is a schematic diagram of the steps after dilution after in vivo typing;
[0061] Figure 9 It is a schematic diagram of the steps for diluting and killing insects after dilution with a reference turbidimetry card;
[0062] Figure 10 It is a schematic diagram of the steps for dilution, layering, and enrichment after in vivo typing;
[0063] Figure 11 It is a schematic diagram of the steps for dilution, layering, and enrichment after refrigerating and in vivo typing;
[0064] Figure 12 It is a schematic diagram of a turbidimetry card;
[0065] Figure 13 It is an image of diluted feces, with color converted to grayscale;
[0066] Figure 14 It is an output report for detecting feces using suspension imaging;
[0067] Figure 15 They are worms and eggs identified by secondary training;
[0068] Figure 16 They are bacteria identified during the secondary training process;
[0069] Figure 17 They are cells identified during the secondary training process;
[0070] Figure 18 They are various food residues identified during the secondary training process. Detailed implementation manner
[0071] The following further details the content of this application in conjunction with each attached drawing. It should be noted that the following is a description of the preferred embodiments of the present invention and does not constitute any limitation to the present invention. The description of the preferred embodiments of the present invention is only for the description of the general principles of the present invention. The numbers "first", "second", and the combinations of letters and numbers such as "A", "B", "TA", "TB", "H" involved in the present invention are only for the convenience of description, and do not represent the order relationship in time or space. The specific meanings of the combinations of letters and numbers are determined by the specific words they represent.
[0072] As Figure 1, 2, 3, Fecal detection target recognition AI training method, including: preprocessing feces to obtain fecal samples; fecal preprocessing includes diluting feces with a diluent; spreading the fecal samples flat so that the fecal detection targets are suspended in the liquid, presenting a natural state; photographing the fecal samples to obtain pictures; using the obtained pictures for AI training to obtain the fecal detection target feature dataset A.
[0073] By allowing the target to be detected to naturally suspend and unfold, its morphological characteristics can be fully unfolded, so as to form an effective sample, and there is sufficient recognition rate for training. For samples that are not naturally unfolded, even manual annotation cannot be standardized, and the AI training work cannot be carried out.
[0074] Such as Figure 4 , fecal preprocessing includes fecal sample staining treatment, and a staining agent is added to the feces during fecal sample staining treatment.
[0075] Through staining, the target detection objects can be distinguished from background impurities. When manually annotating, the targets can be found. For example, target objects such as white blood cells and bacteria are scarce in feces, and annotators have to spend a very large amount of work to find the target objects.
[0076] Figure 4 , Figure 5 , the staining agent is dissolved in the diluent.
[0077] For fecal samples, they need to be diluted to a certain extent so that the targets can be separated from fecal impurities. Diluting and staining at the same time, pre-configuring the staining agent and the diluent into a staining diluent with a certain concentration can effectively control the concentration and uniformity of the staining agent.
[0078] Figure 6 , Figure 12 , the dilution treatment refers to a turbidimetric card for dilution, and the feces are diluted within a set concentration range. It can be diluted first and then diluted proportionally by burning, which can control the dye concentration of the staining.
[0079] For fecal samples, they need to be diluted to a certain extent so that the targets can be separated from fecal impurities. The dilution degree is difficult to control. Through the turbidimetric card, the fecal samples can be more accurately diluted to the set concentration range.
[0080] Figure 7 , add a set proportion of staining diluent to the fecal samples obtained by turbidimetric dilution. The staining diluent includes a staining agent and / or an insecticide.
[0081] Figure 10 , Figure 11 , enrich the obtained fecal samples. The enrichment is precipitation enrichment and / or centrifugal enrichment; after enrichment, take the suspension in the layer where the fecal detection target is located as the fecal sample.
[0082] Figure 11 If the fecal detection target is a lipid droplet, and the lipid droplet floats on the upper layer of the fecal sample, take the upper layer sample to obtain an enriched sample; the photographing of the fecal sample is to photograph the upper layer of the sample to obtain a picture.
[0083] Figure 1 , Figure 2 , AI training includes image preprocessing and AI training processing. The image preprocessing outputs the labeled fecal detection target, and based on the labeled fecal detection target, the AI training processing outputs the fecal detection target feature dataset.
[0084] Figure 10 , before fecal pretreatment, it includes the fixed-type treatment of live worms.
[0085] Figure 11 , fix the live worms by refrigeration, and restore to normal temperature after refrigeration. The refrigeration temperature is higher than 2 degrees and lower than 8 degrees;
[0086] Fix the live worms by heating, and restore to normal temperature after heating. The heating temperature is higher than 39 degrees and lower than 60 degrees.
[0087] Figure 9 , the fixed-type treatment of live worms is to fix live worms or eggs by medicaments.
[0088] For the above-mentioned fecal detection target recognition AI training method, the medicaments include any one or more of formic acid, acetic acid, propionic acid, malonic acid, butyric acid, methanol, ethanol, formaldehyde, acetaldehyde, propionaldehyde, butyraldehyde, n-valeraldehyde, and glutaraldehyde.
[0089] Figure 3 , use two-stage AI training. Use the fecal detection target feature dataset A as the feature dataset, use the AI software to identify one or more pictures of the said picture to obtain a marked picture. The marked picture includes the labeled fecal detection target, conduct manual review on the marked picture to form a manually reviewed marked picture, and use the manually reviewed marked picture for AI training to obtain the fecal detection target feature dataset B.
[0090] Such as Figures 14 to 18 , through a large number of experiments, various target samples are successfully found. Use the original target samples to successfully obtain the object feature dataset A. Use the object feature dataset A for AI recognition to detect more target samples. Interpret the objects identified at the first level, conduct multiple rounds of secondary AI training, gradually enhance the recognition accuracy of the AI model, and establish various targets, including intestinal protozoa, Ascaris eggs, hookworm eggs, pinworm tape, tapeworm eggs, intestinal protozoa, pathogenic microorganisms, cells, starch granules, lipid droplets, plant fibers, muscle fibers, and unknown tangible substances.
[0091] Figure 15, the eggs, some parasites identified by the first-level AI, and the found targets, after precise annotation, can be used for the training of the second-level AI. In the figure, many samples were discovered in feces for the first time, and the AI recognition data was successfully established.
[0092] Figure 16 , the fungi identified by the first-level AI, and the found targets, after precise annotation, can be used for the training of the second-level AI. By solidifying the living body and obtaining clear images under a high-power microscope, although the movement frequency of some living bodies is small and taking pictures does not affect the clarity, the movement frequency of many living bodies is high, and the pictures taken have serious blurring and cannot be manually interpreted and annotated, nor can they be used for AI recognition training. By fixing the living body, the AI recognition model for this type of living body was established for the first time. This model and the feature dataset have reached the accuracy rate for medical applications.
[0093] Figure 17 , various cells identified by the first-level AI, after precise annotation, are used for the training of the second-level AI. After multiple trainings of the second-level AI, cells of various morphologies are found. After expert interpretation, classification models for epithelial cells, red blood cells, white blood cells, etc. are successfully established.
[0094] Figure 18 , various food residues identified by the first-level AI, after precise annotation, are used for the training of the second-level AI. After multiple trainings of the second-level AI, secondary classifications such as starch granules, lipid droplets, plant fibers, and muscle fibers of various classification morphologies are found, and these parameters can evaluate the digestive ability. The above-mentioned lipid droplets need to obtain the target image by microphotographing the upper liquid surface.
[0095] The unknown tangible objects are identified by the identification personnel. For the detected objects that cannot be accurately classified and located, by establishing an unknown classification project, the target objects of this type are continuously detected and enriched, and then other experienced identification personnel continue to identify them. This can speed up the overall work of AI training and also discover unknown parasites, bacteria, or other unknown tangible objects in feces for medical research use.
[0096] A computing device, the computing device includes the above-mentioned AI training method for the identification of fecal detection targets.
[0097] A computing device, the memory of the computing device includes the above-mentioned AI feature dataset A for fecal detection targets or the AI feature dataset B for fecal detection targets.
[0098] A data storage device, the data storage includes the above-mentioned AI feature dataset A for fecal detection targets or the AI feature dataset B for fecal detection targets.
[0099] Although the present invention has been described and illustrated with reference to preferred embodiments and several alternatives, the invention is not limited to the specific descriptions in this specification. Other additional alternatives or equivalent components may also be used to practice the present invention.
Claims
1. A feces detection target recognition AI training method, characterized in that: include: Pre-treating the stool to obtain a stool sample; the stool pre-treating includes diluting the stool with a diluent; Spread the stool sample flat so that the stool test target is suspended in the liquid and presents a natural state; Take photos of stool samples to obtain images; The obtained images are used for AI training to obtain the feces detection target feature dataset A.
2. The feces detection target recognition AI training method according to claim 1, characterized in that: include: The stool pretreatment includes a stool sample staining process, in which a stain is added to the stool.
3. The feces detection target recognition AI training method according to claim 3 is characterized in that: include: The dye is dissolved in the diluent.
4. The feces detection target recognition AI training method according to claim 1, characterized in that: include: The dilution process is performed by referring to a turbidimetric card to dilute the feces to within a set concentration range.
5. The feces detection target recognition AI training method according to claim 4, characterized in that: include: A set proportion of a dye diluent is added to the stool sample obtained by turbidimetric dilution, and the dye diluent includes a dye and / or a pesticide.
6. According to the feces detection target recognition AI training method according to any one of claims 1 to 5, the obtained feces sample is enriched, and the enrichment is precipitation enrichment and / or centrifugal enrichment; after enrichment, the suspension of the layer where the feces detection target is located is taken as the feces sample.
7. The feces detection target recognition AI training method according to claim 6, characterized in that: The feces detection target is fat droplets, which float on the upper layer of the feces sample. The upper layer sample is taken to obtain an enriched sample; the feces sample is photographed by photographing the upper layer of the sample to obtain a picture.
8. The feces detection target recognition AI training method according to claim 1, characterized in that: The AI training includes image preprocessing and AI training processing. The image preprocessing outputs the marked feces detection target object. The AI training processing outputs the feces detection target object feature data set according to the marked feces detection target object.
9. The feces detection target recognition AI training method according to claim 1, characterized in that: Pretreatment of feces includes live worm sizing.
10. The feces detection target recognition AI training method according to claim 9, characterized in that: The live insect shaping process includes any one of the following technical features: TA10: The live worms are shaped by refrigeration and then returned to normal temperature after refrigeration, wherein the refrigeration temperature is higher than 2 degrees and lower than 8 degrees; TA20: The living worms are shaped by heating, and then returned to normal temperature, the heating temperature is higher than 39 degrees and lower than 60 degrees.
11. The feces detection target recognition AI training method according to claim 9, characterized in that: The live worm shaping is to shape the live worm or worm eggs by using a drug.
12. The feces detection target recognition AI training method according to claim 11, characterized in that: The insecticide includes any one or more of formic acid, acetic acid, propionic acid, malonic acid, butyric acid, methanol, ethanol, formaldehyde, acetaldehyde, propionaldehyde, butyraldehyde, n-valeraldehyde, and glutaraldehyde.
13. The method for AI training of feces detection target recognition according to any one of claims 1 to 12 is characterized in that: A feces detection target feature data set A is used as a feature data set, and one or more images of the image are identified using AI software to obtain a marked image, wherein the marked image includes the marked feces detection target, and the marked image is manually reviewed to form a manually reviewed marked image. The manually reviewed marked image is used for AI training to obtain a feces detection target feature data set B.
14. The method for AI training of feces detection target object recognition according to any one of claims 1 to 12, characterized in that: The feces detection targets include any one of intestinal protozoa, ascarid eggs, hookworm eggs, pinworm tapes, tapeworm eggs, intestinal protozoa, pathogenic microorganisms, cells, starch granules, lipid droplets, plant fibers, muscle fibers, and unknown tangible objects.
15. A computing device, characterized in that: The computing and processing device includes the AI training method for feces detection target recognition as described in any one of claims 1 to 10.
16. A computing device, characterized in that: The memory of the computing and processing device includes the feces detection target AI feature data set A or the feces detection target AI feature data set B described in any one of claims 1 to 9.
17. A data storage device, characterized in that: The data storage includes the feces detection target AI feature data set A or the feces detection target AI feature data set B described in any one of claims 1 to 9.