CN-121983283-A - AI data management method for medical institutions

CN121983283ACN 121983283 ACN121983283 ACN 121983283ACN-121983283-A

Abstract

The invention discloses an AI data treatment method for medical institutions, which relates to the field of intelligent medical treatment and solves the problem that the existing AI data treatment method has poor treatment effect on medical data, and comprises the following steps of S1: the method comprises the steps of obtaining medical text data and medical image data, dividing the medical text data into independent medical analysis texts, performing diagnosis rationality analysis on the medical analysis texts, and obtaining medical institution acquisition data according to analysis results, wherein step S2 is to perform diagnosis matching analysis on medical analysis images according to the medical institution acquisition data, obtain image diagnosis rational indexes corresponding to the medical analysis images according to the analysis results, screen the medical analysis images according to the image diagnosis rational indexes, obtain medical image primary screening data, and step S3 is to perform medical data treatment according to the medical image primary screening data and the medical institution acquisition data, obtain treatment qualified data, and improve AI data treatment effects.

Inventors

WU DI
WU XIU
XIANG JINBIAO
ZHAO YAQING
LIU XUEBIN

Assignees

福鑫数科(杭州)人工智能有限公司

Dates

Publication Date: 20260505
Application Date: 20260123

Claims (10)

1. The AI data management method facing the medical institution is characterized by comprising the following steps: Step S1, acquiring medical text data and medical image data, dividing the medical text data into independent medical analysis texts, performing diagnosis rationality analysis on the medical analysis texts, screening the medical analysis texts according to analysis results to obtain medical text primary screening data, and defining the medical text primary screening data and the medical image data as medical institution acquisition data; Step S2, performing diagnosis matching analysis on the medical analysis image according to the acquired data of the medical institution, acquiring an image diagnosis reasonable index corresponding to the medical analysis image according to an analysis result, and screening the medical analysis image according to the image diagnosis reasonable index to obtain medical image primary screening data; and step S3, medical data treatment is carried out according to the medical image preliminary screening data and the medical institution acquisition data, and treatment qualified data are obtained.
2. The AI data management method for medical institutions of claim 1, wherein in step S1, further comprising the steps of: Step S11, acquiring medical institutions for medical data management, and selecting a target medical institution from the acquired medical institutions; Step S12, acquiring medical texts generated by a target medical institution at the current moment to obtain medical text data, and acquiring medical images generated by the target institution at the current moment to obtain medical image data; s13, performing treatment demand analysis on the medical text data, and obtaining medical text primary screening data according to analysis results; And S14, defining the medical text preliminary screening data and the medical image data as medical institution acquisition data.
3. The AI data management method for medical institutions of claim 2, wherein in step S13, further comprising the steps of: S131, dividing medical text data into a plurality of medical analysis texts, and randomly selecting one sample analysis text from the plurality of medical analysis texts; Step S132, performing symptom diagnosis matching analysis on the sample analysis text to obtain symptom diagnosis matching degree corresponding to the sample analysis text; S133, performing index diagnosis matching analysis on the sample analysis text to obtain index diagnosis matching degree corresponding to the sample analysis text; Step S134, obtaining a text diagnosis reasonable index corresponding to the sample analysis text through calculation on the symptom diagnosis matching degree and the index diagnosis matching degree; Step S135, obtaining a text diagnosis reasonable index corresponding to each medical analysis text; Step S136, acquiring a text diagnosis reasonable index preset interval, screening the corresponding medical analysis text into qualified medical texts if the text diagnosis reasonable index is in the text diagnosis reasonable index preset interval, and screening the corresponding medical analysis text into medical texts to be treated if the text diagnosis reasonable index is not in the text diagnosis reasonable index preset interval, so as to obtain medical text primary screening data.
4. The AI data governance method for a medical institution of claim 3, further comprising the step of: Obtaining clinical diagnosis in a sample analysis text to obtain target clinical diagnosis; Acquiring clinical symptoms of patients related to the sample analysis text to obtain clinical symptoms of a plurality of patients; acquiring a plurality of diagnosis typical symptoms, and acquiring clinical symptoms of a patient and the same symptoms in the diagnosis typical symptoms to obtain Z1 common clinical symptoms to Za common clinical symptoms; Counting the number of historical patients with target clinical diagnosis received by a target medical institution to obtain a historical target diagnosis accumulated number; counting the number of historical target patients with Z1 common clinical symptoms to obtain the number of Z1 symptom cases, calculating the ratio of the number of Z1 symptom cases to the cumulative number of historical target diagnosis to obtain the historical occurrence ratio of Z1 symptoms, and similarly, calculating the historical occurrence ratio of Za symptoms; Obtaining the number of symptom cases corresponding to each typical symptom, calculating the ratio of the number of symptom cases to the accumulated number of historical target diagnosis, obtaining the historical occurrence ratio corresponding to each typical symptom, and summing the obtained historical occurrence ratios to obtain the accumulated occurrence ratio of the typical symptom; And calculating the Z1 symptom history occurrence ratio to the Za symptom history occurrence ratio and the typical symptom cumulative occurrence ratio to obtain the symptom diagnosis matching degree corresponding to the sample analysis text.
5. The AI data governance method for a medical institution of claim 3, further comprising the step of, in step S133: Obtaining clinical diagnosis in a sample analysis text to obtain target clinical diagnosis; obtaining abnormal medical indexes contained in the sample analysis text to obtain a plurality of patient abnormal indexes; Obtaining typical medical indexes related to target clinical diagnosis to obtain a plurality of diagnosis typical indexes, acquiring the same medical treatment in the abnormal index and the diagnosis typical index of the patient to obtain a Z1 common medical treatment index to a Zb common medical treatment index; Acquiring a plurality of historical target patients, carrying out a number statistics to obtain a historical target diagnosis cumulative number, carrying out a number statistics to the historical target patients with Z1 common medical indexes to obtain a Z1 index case number, calculating the ratio of the Z1 index case number to the historical target diagnosis cumulative number to obtain a Z1 index historical occurrence ratio, and the like, counting a Zb index case number and calculating a Zb index historical occurrence ratio; Obtaining the index case number corresponding to each diagnosis typical index, calculating the ratio of the obtained index case number to the historical target diagnosis cumulative number to obtain the historical occurrence ratio corresponding to each diagnosis typical index, and summing the obtained historical occurrence ratios to obtain the typical index cumulative occurrence ratio; And calculating the historical occurrence ratio of the Z1 index to the historical occurrence ratio of the Zb index and the cumulative occurrence ratio of the typical index to obtain the index diagnosis matching degree corresponding to the sample analysis text.
6. The AI data management method for medical institutions of claim 1, wherein in step S2, further comprising the steps of: S21, acquiring medical institution acquisition data, acquiring medical image data according to the medical institution acquisition data, splitting the medical image data into a plurality of medical analysis images, and arbitrarily selecting one sample analysis image from the acquired plurality of medical analysis images; s22, performing diagnosis matching degree analysis on the sample analysis image to obtain an image diagnosis reasonable index corresponding to the sample analysis image; Step S23, obtaining an image diagnosis reasonable index and an image diagnosis reasonable index preset interval corresponding to each medical analysis image, screening the corresponding medical analysis image into a qualified medical image if the image diagnosis reasonable index is in the image diagnosis reasonable index preset interval, and screening the corresponding medical analysis image into a medical image to be treated if the image diagnosis reasonable index is not in the image diagnosis reasonable index preset interval, so as to obtain medical image primary screening data.
7. The AI data management method for a medical institution of claim 6, wherein in step S22, further comprising the steps of: step S221, obtaining a medical diagnosis result corresponding to the sample analysis image to obtain a sample image diagnosis; Step S222, acquiring a plurality of historical analysis images obtained by acquiring a historical medical image with a medical diagnosis result being sample image diagnosis, and randomly selecting one sample historical image from the acquired plurality of historical analysis images; step S223, if the sample image is diagnosed as lumbar disc herniation, performing similarity analysis on the sample analysis image and the sample history image, and acquiring medical image similarity of the sample analysis image and the sample history image according to an analysis result; In the step S223, the method further includes the following steps: Adjusting the sample analysis image and the sample history image to the same image scaling ratio, and creating a plane rectangular coordinate system by taking a geometric center corresponding to the sample analysis image as a coordinate origin to obtain an image plane rectangular coordinate system; And marking analysis cone connection lines in the sample analysis images, and marking historical cone connection lines in the sample historical images.
8. The AI data governance method for a medical institution of claim 7, further comprising the step of: In an image plane rectangular coordinate system, covering a sample historical image with a sample analysis image, if the length value of an analysis cone connecting line is more than or equal to that of the historical cone connecting line, completely covering the analysis cone connecting line with the analysis cone connecting line, and if the length value of the sample cone connecting line is less than that of the historical cone connecting line, completely covering the analysis cone connecting line with the historical cone connecting line; And respectively carrying out pixel point number ratio analysis on the lumbar disc region in the sample analysis image and the sample historical image, and acquiring the first disc roundness and the second disc roundness according to analysis results.
9. The AI data governance method for a medical institution of claim 8, further comprising the step of: Calculating a difference value between the roundness of the first intervertebral disc and the roundness of the second intervertebral disc, and calculating a ratio of the obtained difference value to the roundness of the second intervertebral disc to obtain the roundness deviation degree of the image lumbar intervertebral disc; acquiring an overlapping region of the first lumbar intervertebral disc region and the second lumbar intervertebral disc region to obtain a third lumbar intervertebral disc region, and respectively counting the number of pixel points in the first lumbar intervertebral disc region, the second lumbar intervertebral disc region and the third lumbar intervertebral disc region to obtain a first region pixel point number value, a second region pixel point number value and a third region pixel point number value; calculating the first region pixel point number value, the second region pixel point number value and the third region pixel point number value to obtain the image intervertebral disc region coincidence rate; calculating the superposition rate of the lumbar intervertebral disc region and the deviation degree of the circle degree of the lumbar intervertebral disc to obtain the medical image similarity between the sample analysis image and the sample history image; and obtaining medical image similarity corresponding to the sample analysis image and each historical analysis image, obtaining a plurality of medical image similarity, and setting the medical image similarity with the largest numerical value as an image diagnosis reasonable index corresponding to the sample analysis image.
10. The AI data management method for medical institutions of claim 1, wherein in step S3, further comprising the steps of: acquiring medical institution acquisition data, and acquiring medical text preliminary screening data according to the medical institution acquisition data; Acquiring a plurality of medical texts to be treated according to the preliminary screening data of the medical texts, feeding the medical texts to be treated back to a text uploading terminal, correcting the medical texts to be treated to obtain medical texts to be judged, creating a text treatment judgment model, finishing treatment of the medical texts to be treated if the text treatment judgment model screens the medical texts to be judged as qualified medical texts, and repeatedly treating the medical texts to be treated if the text treatment judgment model screens the medical texts to be treated until the text treatment judgment model screens the medical texts to be judged as qualified medical texts; Acquiring a plurality of medical images to be treated according to the primary screening data of the medical images, feeding the medical images to be treated back to an image uploading terminal, correcting the medical images to be treated to obtain medical images to be judged, creating an image treatment judgment model, finishing treatment of the medical images to be treated if the image treatment judgment model screens the medical images to be judged as qualified medical images, and repeatedly treating the medical images to be treated if the image treatment judgment model screens the medical images to be treated as the medical images to be treated until the image treatment judgment model screens the medical images to be judged as qualified medical images; and defining the qualified medical image and the qualified medical text as treatment qualified data.

Description

AI data management method for medical institutions Technical Field The invention belongs to the field of intelligent medical treatment, relates to an artificial intelligence technology, and in particular relates to an AI data treatment method for medical institutions. Background The existing AI data management method has the following defects when managing medical text data and medical image data of a medical institution: 1. When the existing AI data treatment method is used for carrying out data treatment on medical text data, the medical text data can be subjected to formatting comparison and missing value supplementation generally, clinical diagnosis in the medical text data is difficult to carry out rationality evaluation by combining with historical diagnosis big data of a medical institution, AI model treatment cannot be carried out on medical text with unreasonable diagnosis according to evaluation, training of a medical diagnosis model is not carried out on the reasonable diagnosis medical text which is obtained by the AI model, and therefore, the text training quality of the medical diagnosis model is difficult to provide guarantee, and the situation of insufficient interaction accuracy occurs when the medical diagnosis model processes the medical text data; 2. When the conventional AI data management method is used for data management of medical text data, the medical image data can only be subjected to formatting comparison and definition restoration, the image diagnosis in the medical image data is difficult to reasonably evaluate by combining with the historical image diagnosis big data of a medical institution, the AI model management can not be performed on the medical image with unreasonable diagnosis according to the evaluation result, the training of the medical diagnosis model can not be performed on the medical image with unreasonable diagnosis outputted by the AI model, the guarantee for the graphic training quality of the medical diagnosis model is difficult to be provided, and the interaction quality of the medical image data is further influenced. Therefore, we propose an AI data governance method for medical institutions. Disclosure of Invention Aiming at the defects existing in the prior art, the invention aims to provide an AI data treatment method oriented to a medical institution, and aims to improve the AI treatment quality of medical data. In order to achieve the purpose, the invention adopts the following technical scheme that the AI data management method facing the medical institution comprises the following steps: Step S1, acquiring medical text data and medical image data, dividing the medical text data into independent medical analysis texts, performing diagnosis rationality analysis on the medical analysis texts, screening the medical analysis texts according to analysis results to obtain medical text primary screening data, and defining the medical text primary screening data and the medical image data as medical institution acquisition data; Step S2, performing diagnosis matching analysis on the medical analysis image according to the acquired data of the medical institution, acquiring an image diagnosis reasonable index corresponding to the medical analysis image according to an analysis result, and screening the medical analysis image according to the image diagnosis reasonable index to obtain medical image primary screening data; and step S3, medical data treatment is carried out according to the medical image preliminary screening data and the medical institution acquisition data, and treatment qualified data are obtained. Further, in the step S1, the method further includes the following steps: Step S11, acquiring medical institutions needing to carry out treatment on medical data AI, and arbitrarily selecting a target medical institution from the acquired medical institutions; Step S12, acquiring medical texts generated by a target medical institution at the current moment to obtain medical text data, and acquiring medical images generated by the target institution at the current moment to obtain medical image data; s13, performing treatment demand analysis on the medical text data, and obtaining medical text primary screening data according to analysis results; And S14, defining the medical text preliminary screening data and the medical image data as medical institution acquisition data. Further, in the step S13, the method further includes the following steps: Step S131, dividing the medical text data into a plurality of medical analysis texts, and randomly selecting one analysis text from the acquired plurality of medical analysis texts; step S132, performing symptom diagnosis matching analysis on the sample analysis text, and acquiring symptom diagnosis matching degree corresponding to the sample analysis text according to an analysis result; S133, performing index diagnosis matching analysis on the sample analysis text, and acquiring index diagnosis matchi