CN-122025101-A - Method and system for evaluating skin lesions of herpes zoster in stages based on AI image recognition

CN122025101ACN 122025101 ACN122025101 ACN 122025101ACN-122025101-A

Abstract

The invention discloses a method and a system for evaluating the skin damage of herpes zoster in stages based on AI image recognition, which relate to the technical field of the evaluation of the skin damage of the herpes zoster in stages and comprise the following steps of collecting the skin damage image data of the herpes zoster continuously recorded in a multi-period mode, synchronously embedding time marks in the collecting process, uniformly recording the shooting time, the illumination state and the shooting angle information of each frame of image to form an original image sequence; based on the original image sequence, the adaptive ordering adjustment is performed on the time intervals between adjacent frames according to the progressive order of the time identifiers. The invention constructs a stable skin damage time evolution track through the time mark and the continuous processing flow, avoids the stage reverse judgment caused by the disorder of the frame sequence, improves the continuity and the reliability of the stage evaluation of the skin damage of the herpes zoster, simultaneously unifies the color, the area and the texture changes into the natural time rhythm constraint, ensures that the stage result more accords with the clinical course rule, and improves the safety of remote follow-up and intelligent evaluation.

Inventors

Ji Zhoujing
Zhao Caichou
YANG SHENGJU

Assignees

南通大学附属医院

Dates

Publication Date: 20260512
Application Date: 20260130

Claims (10)

1. The method for evaluating the skin lesions of the herpes zoster stage based on AI image recognition is characterized by comprising the following steps of: Step one, collecting the herpes zoster skin damage image data continuously recorded in multiple time periods, synchronously embedding a time mark in the collecting process, uniformly recording the shooting time, the illumination state and the shooting angle information of each frame of image, and forming an original image sequence; step two, based on the original image sequence, performing self-adaptive sequencing adjustment on the time intervals between adjacent frames according to the progressive sequence of the time marks, restoring the original time sequence of the frames according to the time marks, resetting the dislocation frames to the corresponding time points, and reconstructing the natural time rhythm of the skin loss change; step three, in combination with the natural time rhythm after sequencing and adjustment, performing smooth transition processing on the color gradient change of the adjacent frame images, introducing time continuity into color distribution, eliminating abnormal color difference mutation caused by frame dislocation, and reconstructing a natural evolution track of the skin damage region in the time dimension; Step four, depending on the skin damage evolution track subjected to color smooth transition treatment, performing dynamic matching on the shape boundary of the skin damage area, fusing the color change trend into the area proportion adjustment process, and keeping the stable expansion relation of the skin damage shape in the time dimension; and fifthly, based on a stable extension relation obtained through dynamic matching, performing rhythm balance processing on the texture change trend of the skin damage area between continuous frames, and taking the area proportion change into texture transition analysis to weaken the reverse mutation influence caused by time sequence drift.
2. The method for evaluating the skin lesions of shingles based on AI image recognition according to claim 1, wherein the step of collecting the continuously recorded image data of the skin lesions of shingles for a plurality of time periods comprises: Performing image acquisition operation according to a preset shooting interval and time period, keeping the framing range consistent through a fixed shooting distance and a positioning reference object by taking a skin damage area as a center, automatically generating a time stamp by a time synchronization device, and embedding shooting date and time information of each frame of image into metadata fields of an image file; Synchronously recording illumination state information in the image acquisition process, wherein the illumination state comprises light intensity, illumination direction, color temperature value, reflection coefficient and ambient brightness distribution, and the acquisition device adjusts exposure time according to real-time illumination parameters to maintain stable brightness and stores the illumination state and a time mark; Recording shooting angle information of each frame of image under the condition that the time mark and the illumination state are synchronously recorded, wherein the shooting angle comprises a horizontal rotation angle, a vertical pitch angle and distance data of the camera and the skin damage area, and storing an angle change value in association with the time mark and the illumination state; And sorting the multi-period image frames based on the time mark, establishing a time sequence index, calculating the time interval of the adjacent frames, correspondingly associating the illumination state and shooting angle information with the time mark, and generating an original image sequence with three-dimensional time, illumination and space attributes.
3. The AI-image-recognition-based shingles skin lesion stage evaluation method according to claim 2, wherein the step of performing adaptive ordering adjustment of the time intervals between adjacent frames according to the progressive order of the time markers comprises: extracting time identification information of each frame of image from an original image sequence with time continuity, establishing a time index table taking time identification as a main sequence, and recording time intervals and abnormal interval marks of adjacent frames in the index table; performing self-adaptive sequencing adjustment on the image frames according to the progressive sequence of the time marks, and resetting the time difference values of the frames before and after frame reference with abnormal time mark intervals to a reasonable time point to restore the time sequence; Under the condition of completing time sequence adjustment, performing self-adaptive balance processing on the time intervals of the image sequence, and reallocating the time differences of adjacent frames by referring to the average shooting period of the whole group of images to form uniform and continuous intervals; The image frames subjected to sorting and balancing processing are recombined according to the new time sequence to generate a continuous original image sequence with strictly progressive time sequence and balanced time interval.
4. The method for evaluating the skin lesion stage of shingles based on AI image recognition according to claim 3, wherein the step of performing a smooth transition process on the color gradient change of the adjacent frame images comprises: Reading color information of each frame of skin-loss image in the continuous time sequence under the natural time rhythm after sequencing and adjustment, extracting main color components and spatial distribution characteristics of skin-loss areas, and establishing a color distribution data set by combining time marks; Performing smooth transition processing on the color gradient change between adjacent frames according to the progressive sequence of the time marks, analyzing the color difference of pixels at the same position in the adjacent frames according to the time gradient of the skin damage color change, and performing transition adjustment on the color intensity of an abnormal color difference area; performing global consistency correction on skin-loss color change in a time sequence under the condition of finishing color gradient smoothing treatment, and adjusting brightness balance of a skin-loss central region and an edge region according to continuous frame color distribution so as to form uniform gradual change; Arranging the smoothed and corrected image frames in time sequence, superposing the color gradient change curves, generating a complete color evolution sequence, and maintaining the color continuity of the skin damage region in the time dimension and the space layout.
5. The AI-image-recognition-based method for stage assessment of skin lesions of shingles according to claim 4, wherein in global consistency correction, the luminance difference between the center region and the edge region of the lesion is adjusted synchronously in the time dimension by referring to the color gradient change curves of the adjacent frames, and the color distribution between the adjacent frames is maintained continuously during the generation of the color evolution sequence.
6. The method for stage-by-stage evaluation of herpes zoster based on AI image recognition as recited in claim 4, wherein the step of performing dynamic matching of morphological boundaries of the lesion region comprises: Extracting basic form information of a skin damage region in each frame of image from a skin damage evolution track subjected to color smooth transition treatment, identifying and extracting skin damage edge out-of-band contour lines in a layered manner according to a color distribution change trend, and recording space coordinates and color change rates of contour points to form a boundary track; Performing dynamic matching on boundary morphology between adjacent frames according to natural time rhythm of color evolution, and keeping boundary variation continuity by performing point-by-point comparison on the spatial position of boundary contour and morphology variation in continuous frames and adjusting boundary point position by referring to color variation trend; Performing time dimension adjustment on the area proportion of the skin lesion area under the condition of finishing dynamic matching, taking the color change rate as a control factor of area change, and synchronously adjusting the area proportion according to the skin lesion process to form an extension relation of color and area coordination; and connecting the adjusted boundary profiles according to a time sequence to form a continuous boundary curve on a time axis, reconstructing a boundary track through time interpolation processing, and maintaining a stable extension relation of the skin damage morphology in a time dimension.
7. The method for evaluating the skin lesion of the herpes zoster based on the AI image recognition according to claim 6, wherein when the spatial position of the boundary outline of the continuous frame is adjusted in the dynamic matching, the boundary point is subjected to position correction along the color gradient direction according to the color change trend, and the movement rate of the boundary is controlled by combining the time mark, so that the continuous expansion relation of the skin lesion boundary on the time sequence is maintained in the process of adjusting the color change and the area proportion.
8. The AI-image-recognition-based shingles skin lesion stage evaluation method according to claim 6, wherein the step of performing a rhythm balance process on a skin lesion area texture variation trend between successive frames comprises: Extracting texture feature information of skin damage areas in continuous frames under a stable extension relation obtained by dynamic matching, carrying out layered extraction on the texture direction, density and contrast of the skin damage surface of each frame according to color distribution and morphological boundaries, and establishing texture feature data by combining time marks; Performing rhythm balance processing on texture variation trend between adjacent frames according to stable extension relation obtained by dynamic matching, comparing texture direction, density and contrast variation frame by frame with reference to time mark, and adjusting abnormal region; Under the condition of finishing the rhythm balance processing, the area proportion change is incorporated into texture transition analysis, and the skin area proportion and the texture density are subjected to time correlation calculation to form a time map of the coordinated change of the area and the texture; And recombining all frame images according to the time identification sequence based on the rhythm balance and area fusion processing results, and forming a complete texture change curve on a time axis.
9. The method for stage evaluation of herpes zoster skin lesions based on AI image recognition according to claim 8, wherein the time correspondence between the area ratio of the lesion area and the texture density is synchronously associated by a time mark by incorporating the area ratio change into the step of texture transition analysis, the texture density is correspondingly increased when the lesion area is in a shrinkage trend, and the texture density is correspondingly reduced when the lesion area is in an expansion trend.
10. The herpes zoster skin damage stage assessment system based on AI image recognition is used for realizing the herpes zoster skin damage stage assessment method based on AI image recognition as set forth in any one of claims 1 to 9, and is characterized by comprising an image acquisition recording module, a time sequence ordering adjustment module, a color smooth transition module, a form dynamic matching module and a texture rhythm balancing module; the image acquisition and recording module acquires the herpes zoster skin damage image data continuously recorded in multiple time periods, synchronously embeds time marks in the acquisition process, uniformly records the shooting time, the illumination state and the shooting angle information of each frame of image, and forms an original image sequence; The time sequence ordering adjustment module is used for executing self-adaptive ordering adjustment on the time intervals between adjacent frames according to the progressive sequence of the time marks based on the original image sequence, restoring the original time sequence of the frames according to the time marks, resetting the dislocation frames to the corresponding time points, and reconstructing the natural time rhythm of the skin loss change; The color smooth transition module is used for executing smooth transition processing on the color gradient change of the adjacent frame images by combining the natural time rhythm after sequencing and adjustment, introducing time continuity into color distribution, eliminating abnormal color difference mutation caused by frame dislocation, and reconstructing a natural evolution track of the skin region in the time dimension; The morphology dynamic matching module performs dynamic matching on morphology boundaries of the skin damage areas by means of skin damage evolution tracks subjected to color smooth transition treatment, fuses color change trends into an area proportion adjustment process, and keeps a stable expansion relation of the skin damage morphology in a time dimension; And the texture rhythm balance module is used for executing rhythm balance processing on the texture change trend of the skin damage area between continuous frames based on the stable extension relation obtained by dynamic matching, and taking the area proportion change into texture transition analysis to weaken the reverse mutation influence caused by time sequence drift.

Description

Method and system for evaluating skin lesions of herpes zoster in stages based on AI image recognition Technical Field The invention relates to the technical field of herpes zoster skin lesion stage evaluation, in particular to an AI image recognition-based herpes zoster skin lesion stage evaluation method and system. Background The stage evaluation of the skin lesions of the herpes zoster based on AI image recognition refers to the process of automatically recognizing and analyzing the skin lesions of the patients with the herpes zoster in a stage by utilizing an artificial intelligence image recognition technology. The method is characterized in that a visible light or multispectral image of a skin damage area of a patient is acquired through high-resolution image acquisition equipment, and a deep learning model (such as a convolutional neural network CNN) is utilized to extract and classify the characteristics of skin rash in the image, so that different stage states such as a erythema stage, a blister stage, a crusting stage, a healing stage and the like are automatically distinguished. The model learns the dynamic change rule of skin lesions in terms of color, form, luster, boundary structure and the like by learning a large number of marked clinical sample images, so that a skin lesion stage result can be rapidly, objectively and accurately given when a new image is input, an auxiliary diagnosis basis is provided for doctors, and standardization and intellectualization of herpes zoster disease process assessment are realized. The prior art has the following defects: In the dynamic evaluation process of the skin lesion stage of the herpes zoster, the image acquisition device is easy to generate the condition that partial frame data are disordered in sequence when continuously recording in multiple time periods, so that the time sequence characteristics drift at the input end of the model. Because the algorithm mainly carries out stage judgment according to time sequence information such as skin damage color change, area change, texture update and the like, when a frame sequence is disturbed, the model can erroneously identify the skin damage originally in a healing stage as a reappearance active focus, so that a 'recurrence period' conclusion is output. The misjudgment not only can cause reverse jump of the course evaluation result, but also can interfere with the judgment of the curative effect trend by doctors, so that the subsequent intervention scheme is improperly adjusted. Especially in the automatic follow-up or remote diagnosis and treatment scene, the problem is more easily ignored, and serious consequences such as delay of treatment period or misuse of medicines can be finally caused. The above information disclosed in the background section is only for enhancement of understanding of the background of the disclosure and therefore it may include information that does not form the prior art that is already known to a person of ordinary skill in the art. Disclosure of Invention The invention aims to provide a herpes zoster skin lesion stage assessment method and system based on AI image recognition so as to solve the problems in the background art. In order to achieve the purpose, the invention provides the following technical scheme that the herpes zoster skin lesion stage evaluation method based on AI image recognition comprises the following steps: step one, collecting the herpes zoster skin damage image data continuously recorded in multiple time periods, synchronously embedding a time mark in the collecting process, uniformly recording the shooting time, the illumination state and the shooting angle information of each frame of image to form an original image sequence with time continuity, and accurately judging the follow-up time sequence adjustment; Step two, based on the original image sequence with time continuity, executing self-adaptive sequencing adjustment on the time interval between adjacent frames according to the progressive sequence of the time mark, restoring the original time sequence of the frames according to the time mark, resetting the dislocation frames to the corresponding time points, reconstructing the natural time rhythm of skin loss change, and providing a continuous input sequence for color processing; Step three, in combination with the natural time rhythm after sequencing and adjustment, performing smooth transition processing on the color gradient change of the adjacent frame images, introducing time continuity into color distribution, eliminating abnormal color difference mutation caused by frame dislocation, reconstructing a natural evolution track of a skin damage area in a time dimension, and providing color continuity support for morphological boundary processing; Step four, depending on the skin damage evolution track subjected to color smooth transition treatment, performing dynamic matching on the shape boundary of the skin damage area, fusing the color