A machine learning based gait assessment method for patients following mastectomy
By using machine learning gait assessment methods and analyzing trunk acceleration data after mastectomy using BP and GA-BP neural networks, the problem of early diagnosis and intervention of scoliosis after mastectomy was solved, and the quality of life of patients was improved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- UNIV OF SHANGHAI FOR SCI & TECH
- Filing Date
- 2023-04-03
- Publication Date
- 2026-06-02
AI Technical Summary
After mastectomy, women may experience changes in their body that affect their self-esteem and posture, leading to scoliosis. Current technology makes early diagnosis and effective intervention difficult, impacting their quality of life.
A machine learning-based gait assessment method was adopted. Trunk movement signals were acquired through sample collection sensors, and BP neural networks and GA-BP neural networks were constructed. Genetic algorithms were used for optimization, trunk acceleration data were analyzed, gait balance was predicted, and scoliosis was determined.
Accurate prediction of gait balance in research subjects, early detection of scoliosis, early intervention to prevent the occurrence of scoliosis, and improvement of quality of life.
Smart Images
Figure CN116350185B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of computer-aided medical technology, specifically to a machine learning-based method for gait assessment after mastectomy. Background Technology
[0002] Breast cancer is the most common malignant tumor among women worldwide, accounting for 29% of new cancer cases globally. Based on the current number of breast cancer patients, it is estimated that 23 million people may be diagnosed with breast cancer by 2030. Moreover, despite the availability of sufficient treatments in current technologies, the treatment and cure of breast cancer remains an obstacle in the clinical management of this disease.
[0003] Currently, surgical treatment is the primary method in clinical practice, with mastectomy being the most common. However, research shows that after mastectomy, women may experience a decline in self-esteem due to changes in their bodies, potentially leading to depression. Furthermore, mastectomy can disrupt a patient's balance and affect posture. In particular, total mastectomy, which involves the complete removal of the affected breast, causes a shift in the weight distribution across the chest walls, affecting posture, gait stability, and ultimately spinal balance.
[0004] Furthermore, scoliosis is a common outcome after unilateral mastectomy. Among patients without scoliosis before surgery, more than 70.0% develop scoliosis after either a total mastectomy of the left or right breast. Once scoliosis develops, it can cause symptoms such as shoulder pain, lower back pain, or nerve damage, severely impacting the patient's quality of life.
[0005] However, because each breast in a woman's body accounts for an average of 4%-5% of body fat and 1% of total body weight, it does not significantly affect body shape and gait. Therefore, the subtle differences are difficult to detect with the naked eye, making scoliosis difficult to identify in daily life. Thus, early diagnosis, intervention, and treatment of scoliosis in patients after mastectomy are of great significance. Summary of the Invention
[0006] This invention was made to solve the above-mentioned problems, and its purpose is to provide a gait assessment method for patients after mastectomy based on machine learning.
[0007] This invention provides a machine learning-based gait assessment method for patients after mastectomy, characterized by the following steps:
[0008] Step S1: Select multiple research subjects according to the data acquisition standards, and obtain multiple sample datasets by means of sample acquisition sensors to obtain the movement signals of the torso of multiple research subjects before and after the operation, and select one of the sample datasets as the target dataset.
[0009] Step S2: Perform data preprocessing on the sample dataset and the target dataset to obtain the preprocessed sample dataset and the preprocessed target dataset, respectively;
[0010] Step S3: Divide the multiple preprocessed sample datasets into training sample datasets and validation sample datasets according to the proportions;
[0011] Step S4: Use the Sigmoid function as the activation function and construct a BP neural network using the backpropagation algorithm. The BP neural network includes an input layer, a hidden layer, and an output layer.
[0012] Step S5: Use a genetic algorithm to optimize the BP neural network to obtain a GA-BP neural network;
[0013] Step S6: Train the BP neural network and the GA-BP neural network using the training sample dataset, and compare the training results of the BP neural network and the GA-BP neural network.
[0014] Step S7: Input the validation sample dataset into the BP neural network and the GA-BP neural network to verify the accuracy of the BP neural network and the GA-BP neural network after training;
[0015] Step S8: Input the preprocessed target dataset into the trained BP neural network or GA-BP neural network to obtain the evaluation results.
[0016] The machine learning-based gait assessment method for patients after mastectomy provided by this invention may also have the following features: In step S1, the data collection criteria are patients aged 18 to 90 who are scheduled for total mastectomy and breast-conserving surgery and who voluntarily participate and sign informed consent forms. Patients must not have received spinal surgery or other spinal physical therapy, have not been diagnosed with vestibular or nervous system diseases, have no unresectable local cancer lesions, have no distant metastases, have no other malignant tumors, and have not experienced back or neck and shoulder pain caused by trauma or inflammation after total mastectomy and breast-conserving surgery.
[0017] The gait assessment method for patients after mastectomy based on machine learning provided by the present invention may also have the following feature: in step S1, the sample acquisition sensor is an electromyography sensor with acceleration acquisition function.
[0018] The gait assessment method for patients after mastectomy based on machine learning provided by the present invention may also have the following features: in step S1, the movement signals are the acceleration, angle and angular velocity in the X, Y and Z directions during trunk movement.
[0019] The gait assessment method for patients after mastectomy based on machine learning provided by the present invention may also have the following features: In step S2, data preprocessing includes data filtering, handling of null and outlier values and data labeling, and data feature selection is performed on each sample dataset so that a set of 1*48-dimensional data is selected from each sample dataset.
[0020] The machine learning-based gait assessment method for patients after mastectomy provided by this invention may also have the following feature: in step S3, multiple preprocessed sample datasets are divided into training sample datasets and validation sample datasets in a 4:1 ratio.
[0021] The machine learning-based gait assessment method for patients after mastectomy provided by this invention may also have the following feature: in step S4, the BP neural network includes 48 input layers, 97 hidden layers and 1 output layer.
[0022] The gait assessment method for patients after mastectomy based on machine learning provided in this invention may also have the following features: in step S5, the population size of the genetic algorithm is 15, the crossover probability is 0.4, the mutation probability is 0.01, and the maximum number of genetic iterations is 30.
[0023] The machine learning-based gait assessment method for patients after mastectomy provided by this invention may also have the following features: in step S6, the training rounds of the BP neural network and the GA-BP neural network are 200, and the initial learning rate is 0.01.
[0024] The role and effect of invention
[0025] The gait assessment method for post-mastectomy patients based on machine learning according to the present invention includes the following steps: Step S1: Selecting multiple research subjects according to data acquisition standards, and acquiring the trunk movement signals of the multiple research subjects before and after surgery using sample acquisition sensors to obtain multiple sample datasets, and selecting one of the sample datasets as the target dataset; Step S2: Preprocessing the sample dataset and the target dataset to obtain a preprocessed sample dataset and a preprocessed target dataset, respectively; Step S3: Dividing the multiple preprocessed sample datasets into training sample datasets and validation sample datasets according to a certain ratio; Step S4: Using the Sigmoid function as the activation function, and adopting... A backpropagation (BP) neural network is constructed using the backpropagation algorithm. The BP neural network includes an input layer, a hidden layer, and an output layer. Step S5: The BP neural network is optimized using a genetic algorithm to obtain a GA-BP neural network. Step S6: The BP neural network and the GA-BP neural network are trained using a training sample dataset, and the training results of the BP neural network and the GA-BP neural network are compared. Step S7: The validation sample dataset is input into the BP neural network and the GA-BP neural network to verify the accuracy of the BP neural network and the GA-BP neural network after training. Step S8: The preprocessed target dataset is input into the trained BP neural network or the GA-BP neural network to obtain the evaluation results.
[0026] Therefore, the gait assessment method for patients after mastectomy based on machine learning in this invention can accurately predict the gait balance of the research subjects by analyzing the acceleration data of the trunk during movement with the help of network learning algorithms. This allows for the determination of whether the body balance of the research subjects has been disrupted, thereby determining whether the human body has developed scoliosis. In turn, it enables intervention and treatment of scoliosis, so as to achieve the purpose of early intervention and prevention of scoliosis. Attached Figure Description
[0027] Figure 1 This is a flowchart of the machine learning-based gait assessment method for patients after mastectomy according to the present invention.
[0028] Figure 2 This is a schematic diagram of some research data from the machine learning-based gait assessment method for patients after mastectomy in this invention.
[0029] Figure 3 This is a schematic diagram of the data acquisition method in the machine learning-based gait assessment method for patients after mastectomy in this invention;
[0030] Figure 4 This is a schematic diagram of the BP neural network in the machine learning-based gait assessment method for patients after mastectomy in this invention.
[0031] Figure 5This is a flowchart of the GA-BP neural network in the machine learning-based gait assessment method for patients after mastectomy in this invention.
[0032] Figure 6 This is a schematic diagram comparing the performance of a BP neural network and a GA-BP neural network in the machine learning-based gait assessment method for patients after mastectomy according to the present invention. Detailed Implementation
[0033] To make the technical means, creative features, objectives and effects of the present invention easy to understand, the following embodiments, in conjunction with the accompanying drawings, specifically illustrate a machine learning-based gait assessment method for patients after mastectomy.
[0034] In this embodiment, a machine learning-based method for gait assessment after mastectomy is provided.
[0035] Figure 1 This is a flowchart of the machine learning-based gait assessment method for patients after mastectomy according to the present invention.
[0036] like Figure 1 As shown, the machine learning-based gait assessment method for post-mastectomy patients involved in this embodiment includes the following steps:
[0037] Step S1: Select multiple research subjects according to the data acquisition standards, and obtain multiple sample datasets by means of sample acquisition sensors to obtain the movement signals of the torso of multiple research subjects before and after the operation, and select one of the sample datasets as the target dataset.
[0038] In this embodiment, the data collection criteria are patients aged 18 to 90 years who are scheduled for total mastectomy and breast-conserving surgery, and who voluntarily participate and sign informed consent forms. Patients must not have received spinal surgery or other spinal physical therapy, have not been diagnosed with vestibular or nervous system diseases, have no unresectable local cancer lesions, have no distant metastases, have no coexisting malignancies, and have not experienced back or neck / shoulder pain due to trauma or inflammation after total mastectomy and breast-conserving surgery. Furthermore, patients must not withdraw their informed consent throughout the data collection process. Ideally, patients must also not have a history of back or neck / shoulder pain, immune system diseases, or a history of spinal diseases before surgery.
[0039] Figure 2 This is a schematic diagram of some research data from the machine learning-based gait assessment method for patients after mastectomy, as described in this invention.
[0040] like Figure 2As shown, the movement signals are the acceleration, angle, and angular velocity in the X, Y, and Z directions during trunk movement. In this embodiment, the sample acquisition sensor mainly acquires the movement signal of the spinous process of the third lumbar vertebra during the subject's walking process.
[0041] Figure 3 This is a schematic diagram of the data acquisition method in the machine learning-based gait assessment method for patients after mastectomy according to the present invention.
[0042] like Figure 3 As shown, the sample acquisition sensor is an electromyography (EMG) sensor with acceleration acquisition function. In this embodiment, the sample acquisition sensor is preferably an EMG sensor manufactured by Shenzhen Weite Intelligent Technology Co., Ltd., and by setting the sample acquisition sensor on the waist belt, it can be fixed to the waist of the research subject, thereby facilitating the acquisition of the movement signal of the torso during the movement of the research subject. The sample acquisition sensor then transmits the acquired movement signal through wired or wireless means.
[0043] S2: Perform data preprocessing on the sample dataset and the target dataset to obtain the preprocessed sample dataset and the preprocessed target dataset, respectively.
[0044] In this embodiment, data preprocessing includes data filtering, handling of null and outlier values, and data labeling. Data feature selection is performed on each sample dataset to select a set of 1*48 dimensional data from each dataset. Specifically, Lagrange interpolation is used to interpolate null and outlier values in the sample datasets.
[0045] S3: Divide multiple preprocessed sample datasets into training sample datasets and validation sample datasets according to the proportions.
[0046] The preprocessed sample datasets are divided into training and validation datasets in a 4:1 ratio. Specifically, each preprocessed sample dataset includes a set of 1*48 dimensional data. In this embodiment, there are 32 preprocessed sample datasets, so the 32 sets of 1*48 dimensional data are divided into training and validation datasets in a 4:1 ratio.
[0047] Figure 4 This is a schematic diagram of the BP neural network in the machine learning-based gait assessment method for patients after mastectomy in this invention.
[0048] like Figure 4 As shown, step S4: Using the Sigmoid function as the activation function, a backpropagation algorithm is used to construct a BP neural network, which includes an input layer, a hidden layer, and an output layer.
[0049] In this embodiment, the BP neural network includes 48 input layers, 97 hidden layers, and 1 output layer.
[0050] Figure 5 This is a flowchart of the GA-BP neural network in the machine learning-based gait assessment method for patients after mastectomy according to the present invention.
[0051] like Figure 5 As shown, step S5: Use a genetic algorithm to optimize the BP neural network to obtain a GA-BP neural network. In this embodiment, the optimized GA-BP neural network can quickly find the globally optimal weights and thresholds.
[0052] In this embodiment, the population size of the genetic algorithm is 15, the crossover probability is 0.4, the mutation probability is 0.01, and the maximum number of genetic iterations is 30.
[0053] Figure 6 This is a schematic diagram comparing the performance of a BP neural network and a GA-BP neural network in the machine learning-based gait assessment method for patients after mastectomy according to the present invention.
[0054] like Figure 6 As shown, step S6: Train the BP neural network and the GA-BP neural network using the training sample dataset, and compare the training results of the BP neural network and the GA-BP neural network.
[0055] In this embodiment, the BP neural network uses the backpropagation algorithm to adjust the coefficients of the training sample data during training, and then updates and iterates the training sample data so that the value Y obtained by the output layer of the BP neural network from each training sample data is consistent with the actual value Y. out By comparing the results, the mean squared error (Y) can be obtained. out -Y) 2 The loss function is used to record the ERR. During training, the GA-BP neural network employs a genetic algorithm to adjust the coefficients of the training sample data until the coefficients are optimized to an acceptable error range. Furthermore, the number of training iterations and the learning rate are kept the same by comparing the BP neural network with the GA-BP neural network.
[0056] In this embodiment, the BP neural network and the GA-BP neural network are trained for 200 epochs, and the initial learning rate is 0.01.
[0057] Step S7: Input the validation sample dataset into the BP neural network and the GA-BP neural network to verify the accuracy of the BP neural network and the GA-BP neural network after training. In this embodiment, the accuracy of the BP neural network and the GA-BP neural network after training is 99.51% and 100%, respectively.
[0058] Step S8: Input the preprocessed target dataset into the trained BP neural network or GA-BP neural network to obtain the evaluation results. In this embodiment, the evaluation results can be used to determine whether the research subject has developed scoliosis after mastectomy.
[0059] The role and effect of the embodiments
[0060] The gait assessment method for post-mastectomy patients based on machine learning according to the present invention includes the following steps: Step S1: Selecting multiple research subjects according to data acquisition standards, and acquiring the trunk movement signals of the multiple research subjects before and after surgery using sample acquisition sensors to obtain multiple sample datasets, and selecting one of the sample datasets as the target dataset; Step S2: Preprocessing the sample dataset and the target dataset to obtain a preprocessed sample dataset and a preprocessed target dataset, respectively; Step S3: Dividing the multiple preprocessed sample datasets into training sample datasets and validation sample datasets according to a certain ratio; Step S4: Using the Sigmoid function as the activation function, and adopting... A backpropagation (BP) neural network is constructed using the backpropagation algorithm. The BP neural network includes an input layer, a hidden layer, and an output layer. Step S5: The BP neural network is optimized using a genetic algorithm to obtain a GA-BP neural network. Step S6: The BP neural network and the GA-BP neural network are trained using a training sample dataset, and the training results of the BP neural network and the GA-BP neural network are compared. Step S7: The validation sample dataset is input into the BP neural network and the GA-BP neural network to verify the accuracy of the BP neural network and the GA-BP neural network after training. Step S8: The preprocessed target dataset is input into the trained BP neural network or the GA-BP neural network to obtain the evaluation results.
[0061] Therefore, the gait assessment method for patients after mastectomy based on machine learning in this invention can accurately predict the gait balance of the research subjects by analyzing the acceleration data of the trunk during movement with the help of network learning algorithms. This allows for the determination of whether the body balance of the research subjects has been disrupted, thereby determining whether the human body has developed scoliosis. In turn, it enables intervention and treatment of scoliosis, so as to achieve the purpose of early intervention and prevention of scoliosis.
[0062] The above embodiments are preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention.
Claims
1. A machine learning-based method for gait assessment in patients after mastectomy, characterized in that, Includes the following steps: Step S1: Select multiple research subjects according to the data acquisition standards, and obtain multiple sample datasets by means of sample acquisition sensors to obtain the movement signals of the torso of the multiple research subjects before and after the operation, and select one of the sample datasets as the target dataset. Step S2: Perform data preprocessing on the sample dataset and the target dataset to obtain a preprocessed sample dataset and a preprocessed target dataset, respectively; Step S3: Divide the multiple preprocessed sample datasets into training sample datasets and validation sample datasets according to the proportions; Step S4: Using the Sigmoid function as the activation function, a backpropagation algorithm is used to construct a BP neural network, which includes an input layer, a hidden layer, and an output layer. Step S5: Optimize the BP neural network using a genetic algorithm to obtain a GA-BP neural network; Step S6: Train the BP neural network and the GA-BP neural network using the training sample dataset, and compare the training results of the BP neural network and the GA-BP neural network; Step S7: Input the verification sample dataset into the BP neural network and the GA-BP neural network to verify the accuracy of the BP neural network and the GA-BP neural network after training; Step S8: Input the preprocessed target dataset into the trained BP neural network or GA-BP neural network to obtain the evaluation result. In step S1, the data collection criteria are patients aged 18 to 90 who are scheduled for total mastectomy and breast-conserving surgery, and who voluntarily participate and sign an informed consent form. The patient must not have undergone spinal surgery or other spinal physiotherapy, not have been diagnosed with vestibular or nervous system diseases, not have unresectable local cancer lesions, and have no distant metastases, no coexisting malignant tumors, and must not have experienced back or neck and shoulder pain due to trauma or inflammation after total mastectomy or breast-conserving surgery. In step S1, the movement signal is the acceleration, angle, and angular velocity in the X, Y, and Z directions during the movement of the torso. In step S2, the data preprocessing includes data filtering, handling of null and outlier values, and data labeling. Furthermore, data feature selection is performed on each sample dataset to select a set of 1*48 dimensional data from each sample dataset. In step S4, the BP neural network includes 48 input layers, 97 hidden layers, and 1 output layer. In step S5, the population size of the genetic algorithm is 15, the crossover probability is 0.4, the mutation probability is 0.01, and the maximum number of genetic iterations is 30. In step S6, the BP neural network and the GA-BP neural network are trained for 200 rounds, and the initial learning rate is 0.
01.
2. The machine learning-based gait assessment method for post-mastectomy patients according to claim 1, characterized in that: In step S3, the multiple preprocessed sample datasets are divided into the training sample dataset and the validation sample dataset in a 4:1 ratio.
3. The machine learning-based gait assessment method for post-mastectomy patients according to claim 1, characterized in that: In step S1, the sample acquisition sensor is an electromyography sensor with acceleration acquisition function.