AI-assisted diabetic foot patient nutrition auxiliary system

Abstract

The invention provides an AI-assisted diabetic foot patient nutrition auxiliary system, which comprises a camera system arranged at an entrance of an indoor living room and a house, a home computer connected with the camera system and used for shooting walking images of a diabetic foot patient at multiple angles, and a server used for analyzing foot and body orientations and medical institutions according to a preset first AI algorithm, and the control module is used for accessing the home computer, transmitting the analysis result of the first AI algorithm, further analyzing the gait by adopting a second AI algorithm, and making a nutrition decision. According to the nutrition auxiliary system, monitoring data is not needed, the tissue damage type is judged by means of the walking posture, and the nutrition recommendation decision needed by the diabetic foot patient is efficiently achieved.

Description

Technical Field

[0001] This invention relates to a nutritional support system for diabetic foot patients, and more particularly to an AI-assisted nutritional support system for diabetic foot patients. Background Technology

[0002] Diabetic foot ulcer (DFU) is one of the most serious complications of diabetes. Statistics show that the amputation rate for DFU patients is as high as 23%, and the 5-year mortality rate is as high as 30.5%, placing a heavy burden on patients and the healthcare system. Studies have shown that more than half of DFU patients suffer from moderate to severe malnutrition, which not only increases the risk of infection and amputation but also affects patient prognosis and quality of life. The inventors' team's earlier paper, "Meta-analysis of factors influencing malnutrition in patients with diabetic foot ulcers" (published in the October 2025 issue of the *Chinese Journal of Nursing*, Volume 60), specifically investigated the quantitative analysis of the incidence and influencing factors of malnutrition in DFU patients through meta-analysis of existing evidence. The aim was to provide a scientific basis for early identification of high-risk groups for malnutrition and the development of personalized intervention measures. This study has been registered in the International Systematic Reviews Database (CRD4202506-23406).

[0003] In another research paper by the inventors' team, a scheme entitled "Construction and Validation of a Predictive Model for Malnutrition Risk in Patients with Diabetic Foot" was proposed (published in the November 2024 issue of *Journal of Clinical Internal Medicine*, Volume 41). This study selected 223 patients with diabetic foot (DF) as the modeling group, dividing them into a nutrition group (148 cases) and a malnutrition group (75 cases) based on their malnutrition status. General and clinical data were collected for univariate analysis. Binary logistic regression analysis was used to explore the influencing factors of malnutrition in DF patients, and a nomogram prediction model was constructed. The predictive value of the model was evaluated using receiver operating characteristic (ROC) curves; the accuracy of the model was assessed using the Hosmer-Lemeshow (HL) test. Clinical data from 91 DF patients during the same period were selected for external validation of the model. Clinical decision-making (DCA) curves were plotted to test the practical application efficacy of the model. Univariate analysis showed that BMI, dietary regularity, foot ulcer infection, Wagner classification, DM complications, CRP, abnormal albumin levels, HbA1c, abnormal LDL, and abnormal TG were all influencing factors for malnutrition in DF patients (P < 0.05). Multivariate logistic regression analysis showed that BMI, dietary regularity, and albumin (ALB) were protective factors against malnutrition in DF patients, while Wagner classification, foot ulcer infection, and glycated hemoglobin (HbA1c) were risk factors (P < 0.05). The area under the ROC curve (AUC) of the model was 0.895 (95% CI 0.850–0.941), and the HL test showed P = 0.248, with the slope of the calibration curve approaching 1. Validation showed that the model's AUC was 0.773 (95% CI 0.674–0.872), and the HL test showed P = 0.882, indicating good calibration and discriminative ability and stable results. DCA curve analysis results show that the model has a high net benefit. This model can effectively predict the occurrence of malnutrition in patients with DF (dystrophic malnutrition), has high clinical application value, and can provide a reference for medical staff to identify high-risk groups for malnutrition in the early stages.

[0004] In addition, studies such as "Nutritional Intervention for Diabetic Foot Ulcers—An Interpretation of the 2022 American Limb Preservation Society Expert Consensus and Working Guidelines: Nutritional Intervention for Adult Diabetic Foot Ulcers" (June 2022, Vol. 23, No. 2, Infection, Inflammation and Repair) and the Master's Thesis from Anhui Medical University, "The Predictive Value of Nutritional Risk Screening and Assessment for Amputation Risk in Patients with Diabetic Foot Ulcers," have all discussed the crucial role of nutritional risk screening and assessment in timely intervention for diabetic foot ulcers.

[0005] However, current technologies often rely on monitoring nutritional and physiological indicators to diagnose symptoms in diabetic foot patients and propose interventions or treatments accordingly. This requires extensive, real-time data monitoring, demands high-quality and continuous data analysis, and necessitates various additional instruments, hindering normal daily activities. While portable monitoring devices are a reliable method, carrying or wearing them is still less convenient than wearing nothing at all. Furthermore, diabetic foot patients often overlook subtle changes in their condition, making remote data collection and monitoring by family members a significant challenge requiring long-term care and patience. The proverb "Prolonged illness tests filial piety" may be a common saying, but it reflects a harsh reality in life. Therefore, finding an optimal solution that reconciles the contradictions between human and machine monitoring is a pressing issue.

[0006] The invention team believes that gait changes in the prediabetic foot stage are an important physiological and behavioral characteristic in the early stages of diagnosis. As the disease progresses, statistically significant changes in gait become important reference standards. For example, shuffling gait, shortened stride, heel-first strike, and discontinuous directional adjustments indicate deterioration of nerve and vascular systems. Therefore, appropriate nutritional intervention is needed to reverse and improve these changes as early as possible.

[0007] However, there is an urgent need for a technological solution that keeps pace with the times and uses AI to assist in nutritional decision-making for patients with diabetic foot, in order to intervene as early as possible and improve the condition in a timely manner. Summary of the Invention

[0008] 1. The core technology of this invention 1. Install video recording devices in areas where patients frequently move around to collect data on their walking posture, body type, and eating habits, and analyze their gait, whether their body type is appropriate, and whether their eating habits are reasonable. 2. Analyze the etiology based on the gait analysis results and decide on the nutritional plan.

[0009] 2. Technical Content of the Invention To embody the core technologies mentioned above and address the problems existing in current technologies, this invention provides an AI-assisted nutritional support system for diabetic foot patients. The system includes a camera system installed in the living room and at the entrance of the house to capture walking images of diabetic foot patients from multiple angles. A home computer connected to the camera system analyzes the foot and body orientation based on a preset first AI algorithm. A server at a medical institution is connected to the home computer to transmit the analysis results of the first AI algorithm. A second AI algorithm is then used to further analyze the gait and make nutritional decisions.

[0010] Optionally, the multiple angles include the front, side, and back, and the camera system includes one camera for each angle.

[0011] Optionally, the methods by which the first AI algorithm analyzes foot and body orientation include: S1: Collect multiple images from each angle, train the first convolutional neural network to complete the recognition of feet in multiple images from each angle, and train the second convolutional neural network to complete the recognition of body orientation in multiple images from each angle. S2: Select images from each angle identified in S1 for each fixed time period to form multiple time-based image sequences for each angle.

[0012] Optionally, the method for selecting the fixed time period is to select any one day in each month as a fixed time node, and then use 0.2s-0.5s intervals in the patient's image at the corresponding angle within the fixed time node as the fixed time period.

[0013] Preferred, the camera system is turned on only when the selected day arrives. If no image of the patient appears within 8-15 hours after it is turned on, it is automatically turned off and turned on again the next day until an image of the patient appears. If no image of the patient appears for two consecutive days, the system is reported to the medical institution's server via the home computer. The medical institution's server then sends a notification to the patient's selected contact person that the patient has not been home for two consecutive days.

[0014] This can save electricity and notify family members when the patient is not at home, in order to prevent accidents.

[0015] Specifically, the second convolutional neural network is trained to perform a method for recognizing body orientation in multiple images from each angle, including: S1-1 acquires an image and identifies an angle of the patient based on a pre-trained second convolutional neural network; S1-2 continues to identify another angle of the patient and stops acquiring images from that angle.

[0016] In other words, once any camera detects an angle, it can continue to detect until another camera detects a different angle. To achieve this, specific camera angles need to be adjusted so that during image acquisition, only one camera detects an angle, while the other cameras fail to detect an angle of the patient.

[0017] The training method for the second convolutional neural network is: The first step involves obtaining multiple images from three angles—front, side, and back—from a manually selected source via the medical institution's server. The second step involves constructing a second convolutional neural network, which is trained using multiple images from the front, side, and back.

[0018] Optionally, methods for using a second AI algorithm to analyze gait and make nutritional decisions include: The first step is to collect the image sequence and sort the image sequence on the timeline; It is important to note that since each camera angle can only be used to identify images at that specific angle, images from angles where the effect is not identified are not listed in the timeline.

[0019] The second step is to calculate the average foot height and the rate of height reduction, and compare them with the normal threshold. The third step is to determine whether the average foot height is less than 20% of the corresponding threshold, i.e., the first percentage is less than 20%, the rate of height reduction is less than 20% of the corresponding threshold, and the second percentage is less than 20%. If the first and second percentages are between 20% and 40%, it is considered moderate; greater than 40% and 60% is considered severe; and greater than 60% is considered dangerous. It is important to emphasize that insufficient foot lift indicates the severity of shuffling gait; as the condition progresses, the foot lift becomes lower and lower.

[0020] The fourth step is to select two angles, find two consecutive images from different angles to determine the turning process, and calculate the turning time based on the time interval between the two images. Compare this time with the normal time threshold to determine whether the balance dysfunction is mild, moderate, severe, or dangerous. The fifth step involves setting up the first to fourth nutrition plans and the first to fourth dietary meal recommendations based on the four categories of problems: nerve damage, vascular damage, and balance dysfunction.

[0021] Optionally, fixed-position cameras can be installed for different customer levels to periodically take photos of diabetic foot wounds, body shape, dietary habits, and gait, which are then uploaded together to assist in providing nutritional decisions.

[0022] 3. Beneficial effects By relying on image recognition of the patient's (or those with early signs) posture, body type, diabetic foot wound healing status, and the severity of possible nerve damage, vascular damage, and balance dysfunction, nutritional recommendation decisions for diabetic foot patients can be made conveniently and quickly without any laboratory data. Attached Figure Description

[0023] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1This is a schematic diagram of the camera system layout and patient indoor activities of the AI-assisted nutritional support system for diabetic foot patients according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the algorithm for judging the severity of three types of problems, which is the basis of the second AI algorithm in this embodiment of the invention for analyzing gait and making nutrition decisions. Detailed Implementation

[0024] The embodiments of the present invention are described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the present invention and are not restrictive.

[0025] This invention provides an AI-assisted nutritional support system for diabetic foot patients, including a camera system installed in the living room and at the entrance of the house to capture walking images of diabetic foot patients from multiple angles, a home computer connected to the camera system to analyze foot and body orientation according to a preset first AI algorithm, and a server of a medical institution to enable the home computer to access and transmit the analysis results of the first AI algorithm, further using a second AI algorithm to analyze gait and make nutritional decisions.

[0026] Optionally, the methods by which the first AI algorithm analyzes foot and body orientation include: S1: Collect multiple images from each angle, train the first convolutional neural network to complete the recognition of feet in multiple images from each angle, and train the second convolutional neural network to complete the recognition of body orientation in multiple images from each angle. S2: Select images from each angle identified in S1 for each fixed time period to form multiple time-based image sequences for each angle.

[0027] like Figure 1 As shown, the AI-assisted nutritional support system for diabetic foot patients includes three cameras installed on three walls of the living room (one of which is the front door), which are responsible for recognizing images from the front, back, and side angles, respectively, and obtaining multiple images.

[0028] All three cameras are connected to a home computer to analyze foot and body orientation based on a preset first AI algorithm. The medical institution's server is used to enable the home computer to access the system, transmit the analysis results of the first AI algorithm, and further use a second AI algorithm to analyze gait and make nutritional decisions.

[0029] Among them, the methods used by the first AI algorithm to analyze foot and body orientation include: S1: Collection Figure 1Multiple images from each angle are used to train a first convolutional neural network to identify feet in multiple images from each angle, and a second convolutional neural network is used to train body orientation in multiple images from each angle. S2: Determine the tenth day of the first quarter, using 0.3s as a fixed time period. Select images from each angle identified in S1 to form multiple time-series images for each angle. Optionally, the fixed time period can be selected by choosing any day of the month as a fixed time node. Within this fixed time node, images of the patient at the corresponding angle are used, with each interval of 0.2s-0.5s serving as the fixed time period. In this embodiment, 0.3s is used. The camera system is activated only on the chosen day. If no patient image appears within 8-15 hours after activation, it automatically shuts off and resumes operation the following day until a patient image appears. If no patient image appears for two consecutive days, the system is reported to the medical institution's server via the home computer. The medical institution's server then sends a notification to the patient's selected contact person that the patient has been away from home for two consecutive days.

[0030] In other words, the three cameras will only turn on on the chosen day. If no image of the patient is captured within 10 hours of turning on, they will automatically turn off and turn on again the following day until an image of the patient is captured. If no image of the patient is captured for two consecutive days, the system will report this to the medical institution's server via the home computer. The medical institution's server will then send a notification to the patient's designated contact person that the patient has not been home for two consecutive days. This system can save electricity and promptly notify family members that the patient is not at home, preventing accidents.

[0031] In this embodiment, the first convolutional neural network can be an object detection model based on YOLO or Faster R-CNN architecture, and the second convolutional neural network can be a network based on ResNet or Vision Transformer architecture.

[0032] In an optional embodiment of the present invention, the method for training the second convolutional neural network to identify body orientation in multiple images at each angle includes: S1-1: Image acquisition identifies an angle of the patient based on a pre-trained second convolutional neural network; S1-2: Continue to identify another angle of the patient and stop acquiring images from that angle.

[0033] In other words, if any one camera detects an angle, it can continue to detect until another camera detects a different angle. To achieve this, specific camera angles need to be adjusted so that during image acquisition, only one camera detects an angle, while the other cameras fail to detect an angle of the patient.

[0034] like Figure 1 As shown, the patient starts moving forward from point A in the living room. Although the rear and front cameras are on, they cannot detect the patient's back and front images respectively, so image acquisition has not actually begun. As the patient continues moving forward, they are captured by the side camera (using a wide-angle lens) on the front door. Figure 2 The image sequence shown consists of multiple images of the patient walking at fixed 0.3-second intervals. At a certain time point, starting from position B, the patient turns around and begins walking back. At this point, the rear camera begins capturing images of the patient's back, while the side cameras stop capturing images. The front camera, however, has not yet detected an angle image of the patient's front and therefore does not capture any images.

[0035] In an optional embodiment of the present invention, the training method of the second convolutional neural network in step S1-1 is: The first step involves obtaining multiple images from three angles—front, side, and back—from the medical institution's server, which were manually selected. The second step involves constructing a second convolutional neural network, which is trained using multiple images from the front, side, and back.

[0036] Combination Figure 2 As shown, the method for analyzing gait and making nutrition decisions using a second AI algorithm includes: The first step is to acquire image sequences and store the image sequences in... Figure 2 Sort in the timeline; The second step is to calculate the average foot height (foot height identified by the boxes in the illustration) and the rate of height reduction, and compare them with a normal threshold. The normal threshold in this embodiment can be set according to different subjects; it can be established based on gait databases of large-scale healthy populations and diabetic foot patients at different disease stages, and verified and calibrated through clinical studies. This threshold can be updated according to research progress, and this embodiment does not limit its specific value.

[0037] The third step is to determine whether the average foot height is less than 20% of the corresponding threshold, i.e., the first percentage is less than 20%, the rate of height reduction is less than 20% of the corresponding threshold, and the second percentage is less than 20%. If the first and second percentages are between 20% and 40%, it is considered moderate; greater than 40% and 60% is considered severe; and greater than 60% is considered dangerous. The fourth step involves selecting two angles and finding two consecutive images from different angles to represent the turning process. Based on the time interval between the two images, the turning time is calculated and compared to a normal time threshold to determine whether the balance impairment is mild, moderate, severe, or dangerous. Figure 2 Select two angles, the back and the side, and find two consecutive images from different angles to determine the turning process. Calculate the turning time based on the time interval between the two images and compare it with the normal time threshold to determine whether the balance dysfunction is mild, moderate, severe, or dangerous. Understandably, if the patient travels back and forth between A and B multiple times, the fourth step can be performed from either the front or side angles.

[0038] Figure 2 From the corresponding Figure 1 Point A in the image is first identified by the side camera at the patient's side angle. Multiple images, represented by ellipses, are then used to form an image sequence.

[0039] The time intervals are actually much longer than 0.3 seconds. It's not a fixed time interval (0.3 seconds as shown in the image) between multiple images from a single-angle camera, but rather a sequence of images from cameras at different angles. This represents the start of image acquisition by the side camera and the start of image acquisition by the rear camera, and also characterizes... Figure 1 The turning motion ended.

[0040] In other words, a turning motion is completed within one time interval. The degree of balance impairment is judged based on whether the turning motion takes too long. The rear camera stops acquiring images until the side camera recognizes the patient's side angle (hence the use of three dots to represent multiple images from the back angle).

[0041] The fifth step involves developing the first to fourth nutritional plans and recommended meal plans (one course of treatment) based on the four classifications of the three categories of problems: nerve damage, vascular damage, and balance dysfunction. Table 1 below shows the first to fourth nutritional plans and recommended meal plans: Table 1

[0042] The study results showed differences in the incidence of malnutrition among DFU patients using different assessment tools. The incidence of malnutrition in DFU patients using the MNA-SF, MNA, and GLIM assessment tools were 52.2%, 70.2%, and 41.4%, respectively. Among them, MNA had the highest detection rate, possibly related to its comprehensive assessment content and high sensitivity and specificity. MNA-SF, as a simplified version, is suitable for rapid screening, while GLIM, based on international consensus, provides standardized diagnostic criteria for malnutrition. The incidence of malnutrition in DFU patients was higher than that in general diabetic patients (32.0%), suggesting that DFU patients are more prone to malnutrition problems. Therefore, nursing staff should conduct regular nutritional screenings based on patients' clinical characteristics and develop personalized nutritional intervention plans according to the screening results.

[0043] The study found that elderly DFU patients have a higher risk of malnutrition, possibly related to factors such as decreased physiological function, reduced appetite, and weakened digestive and absorptive capacity in the elderly. High BMI reduces the risk of malnutrition in DFU patients, possibly due to insulin resistance and abnormal energy metabolism. Furthermore, DFU patients with strong self-care abilities have a lower risk of malnutrition because they can independently ensure nutritional intake, rationally adjust their diet, and maintain metabolic activity. This suggests that caregivers should improve patients' nutritional status by enhancing their self-management abilities. Therefore, caregivers should pay attention to the nutritional status of elderly DFU patients, improving their dietary structure by providing easily digestible and absorbable nutritional supplements; for patients with low BMI, nutritional support should be strengthened to ensure adequate calorie and protein intake; simultaneously, health education should be used to improve patients' self-care abilities and encourage active participation in self-management, thereby improving their nutritional status.

[0044] The study results indicate that glycated hemoglobin (HbA1c) >7%, low albumin levels, low hemoglobin levels, high C-reactive protein (CRP) levels, and the presence of infection increase the risk of malnutrition in DFU patients. High HbA1c levels reflect long-term poor glycemic control, which may lead to metabolic disorders and increased nutrient consumption. Low albumin and hemoglobin levels suggest insufficient protein reserves and hypoxia in ischemic tissues of the foot, affecting wound healing and immune function. Elevated CRP and the presence of infection indicate an exacerbated inflammatory response, which may further worsen malnutrition. For patients with high HbA1c levels, nursing staff should assist physicians in optimizing glycemic control protocols to reduce the impact of metabolic disorders on nutritional status; for patients with low albumin levels, protein intake should be increased, and enteral or parenteral nutritional support should be used when necessary; for patients with elevated CRP and infection, infection and inflammation should be actively controlled to reduce nutrient consumption. Therefore, comprehensive assessment of patients' nutritional status and provision of personalized nutritional support are key to improving the prognosis of DFU patients.

[0045] The study found that patients with diabetic foot ulcers (DFU) classified as Wagner grade 3-5 and those with a long duration of diabetes had a higher risk of malnutrition. As ulcer severity and disease duration increased, patients' nutritional status deteriorated, possibly related to worsening infection, metabolic disorders, and increased complications. For patients with higher Wagner grade DFU, nursing staff should strengthen wound care, control infection, and reduce nutritional consumption. For patients with a longer disease duration, their nutritional status should be assessed regularly, and nutritional support plans should be adjusted promptly to prevent further deterioration of malnutrition.

[0046] Therefore, developing more nutritional assessment tools specifically for DFU patients, constructing risk prediction models, and conducting more high-quality prospective studies will be of paramount importance in this field. In addition, the long-term effects of nutritional intervention on the prognosis of DFU patients should be explored to provide more comprehensive guidance for clinical practice.

[0047] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps related to the aforementioned AI-assisted nutritional support system for diabetic foot patients.

[0048] This invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps related to the aforementioned AI-assisted nutritional support system for diabetic foot patients.

[0049] This invention also provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement steps related to an AI-assisted nutritional support system for diabetic foot patients.

[0050] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. An AI-assisted nutritional support system for patients with diabetic foot, characterized in that, The system includes a camera system installed in the living room and at the entrance of the house to capture images of diabetic foot patients walking from multiple angles. A home computer connected to the camera system is used to analyze the foot and body orientation according to a preset first AI algorithm. A server at the medical institution is used to enable the home computer to access the system, transmit the analysis results of the first AI algorithm, and further use a second AI algorithm to analyze the gait and make nutritional decisions.

2. The system according to claim 1, characterized in that, The multiple angles include the front, side, and back, and the camera system includes one camera for each angle.

3. The system according to claim 2, characterized in that, The first AI algorithm to analyze foot and body orientation includes: S1: Collect multiple images from each angle, train the first convolutional neural network to complete the recognition of feet in multiple images from each angle, and train the second convolutional neural network to complete the recognition of body orientation in multiple images from each angle. S2: Select images from each angle identified in S1 for each fixed time period to form multiple time-based image sequences for each angle.

4. The system according to claim 3, characterized in that, The method for selecting the fixed time period is to randomly select one day in each month as a fixed time node, and then use 0.2s-0.5s intervals in the patient's image at the corresponding angle within that fixed time node as the fixed time period.

5. The system according to claim 3, characterized in that, The camera system will be turned on on the selected day. If no image of the patient appears within 8-15 hours after it is turned on, it will be turned off automatically and will be turned on again the next day until an image of the patient appears. If no image of the patient appears for two consecutive days, the system will be reported to the medical institution's server via the home computer. The medical institution's server will then send a notification to the patient's selected contact person that the patient has not been home for two consecutive days.

6. The system according to any one of claims 2-5, characterized in that, The second convolutional neural network is trained to perform a method for recognizing body orientation in multiple images from each angle, including: S1-1: Image acquisition: An angle of the patient is identified based on a pre-trained second convolutional neural network. S1-2: Continue to identify another angle of the patient and stop acquiring images from that angle.

7. The system according to claim 6, characterized in that, The training method for the second convolutional neural network is: The first step involves obtaining multiple images from three angles—front, side, and back—from a manually selected source via the medical institution's server. The second step is to construct a second convolutional neural network by training it with multiple images from the front, side, and back. Methods for using a second AI algorithm to analyze gait and make nutritional decisions include: The first step is to collect the image sequence and sort the image sequence on the timeline; The second step is to calculate the average foot height and the rate of height reduction, and compare them with the normal threshold. The third step is to determine whether the average foot height is less than 20% of the corresponding threshold, i.e., the first percentage is less than 20%, the rate of height reduction is less than 20% of the corresponding threshold, and the second percentage is less than 20%. If the first and second percentages are between 20% and 40%, it is considered moderate; greater than 40% and 60% is considered severe; and greater than 60% is considered dangerous. The fourth step is to select two angles, find two consecutive images from different angles to determine the turning process, and calculate the turning time based on the time interval between the two images. Compare this time with the normal time threshold to determine whether the balance dysfunction is mild, moderate, severe, or dangerous. The fifth step involves setting up the first to fourth nutrition plans and the first to fourth dietary meal recommendations based on the four categories of problems: nerve damage, vascular damage, and balance dysfunction.

Images