Intelligent Wearable Systems for Diagnosis of Peripheral Neuropathy

Version	Summary	Created by	Modification	Content Size	Created at	Operation
1		Maria Kyrarini	--	4990	2023-11-29 08:02:50	\|
2	format correct	Catherine Yang	Meta information modification	4990	2023-11-30 01:51:53	\|

This entry is adapted from the peer-reviewed paper 10.3390/technologies11060163

The automated diagnosis of peripheral neuropathy (PN) has been performed in many studies where only a single sensor or technique is used for the data acquisition purpose. These techniques are based on analyzing foot plantar pressure or gait analysis, vibration- and sensitivity-based confocal microscopic image processing, thermal image processing, ultrasound scanner, and nerve conduction studies (NCS) or electromyography (EMG). However, in order to diagnose peripheral neuropathy, a continuous source of data is necessary to diagnose and track the progression of PN. Image-based approaches can provide a feasible way of detecting PN. In a number of research studies, a subject’s gait is analyzed using cameras, but, due to the camera’s limited field of view, it cannot provide continuous data which can result in missed data points. Similarly, other non-wearable techniques such as vibration- or perception-based methods and thermal images provide suitable ways for the detection of PN, but these types of methods are not suitable for long-term data collection which is vital for the diagnosis and progression of PN. For this purpose, wearable sensors are the most suitable way to collect continuous data, as wearable devices are designed in such a way that they can be easily worn by patients and attached to the human body directly. They also provide a way to collect and store patient activity data for later use.

wearable systems peripheral neuropathy intelligent systems sensors

1. Wearable Inertial Sensor-Based Intelligent Systems for the Diagnosis of PN

Wearable inertial sensors are mostly used for activity recognition tasks such as human gait or postural positions with potential applications in healthcare and well-being, as they offer reliable and accurate methods for studying human motion ^[1]. There are different configurations of inertial sensors that can be used for recording and analyzing human gait for the purpose of diagnosis. In most cases, inertial sensors are attached to the leg, foot, or waist of the human body ^[2]. The most commonly used sensor in most studies is an accelerometer in order to study human motion ^[3]. The combination of an accelerometer and gyroscope or inertial measurement unit has also recently been used by many researchers ^[4]^[5]^[6]^[7] for studying human motion for different applications.

Chen and Shanshan ^[8] use a wearable inertial sensor, i.e., accelerometer for the early screening of peripheral neuropathy in diabetic patients and evaluate how much valuable information can be added by the wearable inertial sensor in the screening process of peripheral neuropathy. In this study, the authors focus on developing a wearable system based on inertial sensors that distinguish the gait of diabetic patients with and without peripheral neuropathy. The diagnosis of PN is based on the degradation in gait which results in a slower gait, limited knee and ankle mobility, and shorter steps. The experimental study was performed on 106 participants (aged 38 to 83 years old—54 female and 52 male); among those, 30 were diagnosed as confirmed diabetic peripheral neuropathic patients, and 76 healthy individuals were included. The study was approved by the Ethics Committee of Shanghai Ninth People’s Hospital, China. The ground truth in this study was taken from the results of NCS studies and physicians’ diagnoses. For data collection, an ear-worn 3three-axis accelerometer was used to capture movements in lateral, forward, and vertical directions at a sampling rate of 100 Hz. During walking trials, each participant had to walk 10 m for three rounds while capturing motion data.

The data were transmitted wirelessly to a tablet in real time. To obtain the gait feature, vertical acceleration was used to segment gait cycles. To identify heel-strike events, a customized peak detection algorithm was used by delineating each gait cycle from a heel-strike event to the instant before the next heel-strike event. The gait features that were used in this research include heel-strike events, toe-off events, variance, skewness, and kurtosis of the gait signal. Other features like the maximum amplitude of the gait signal, cadence per minute, and gait speed during the 10-m walk test were also included for analysis.

After data processing, five categories of logistic regression models (Model A to Model E) were used to predict the presence of peripheral neuropathy using different databases having different feature points. The first model, i.e., Model A, only used the total score from the Michigan neuropathy screening instrument (MNSI) history; the second model was based on physical examination of motor and sensory functions using nerve conduction studies (NCS). Model C used both the data from MNSI history and NCS studies. The fourth model, Model D, only used data from an ear-worn accelerometer, and the last model, Model E, was based on the combination of MNSI history and gait features from the inertial sensor. Age and gender were also used as feature points in all five proposed models. The main reason for creating different models was to quantitatively measure the contribution of wearable sensors to the early screening of PN. In order to compare the performance of the proposed models, the authors applied the likelihood ratio (LR) test to find the best possible gait features for the diagnosis of PN. Significant increased LR test stats show that the new feature or biomarker enhances the performance of the models and improves model fit statistically. To examine overfitting, the Brier score was calculated by observing the difference between the true probability and observed probability. The results of this study show that three gait features can play a great role in distinguishing abnormal walking in PN patients. These features include the skewness of lateral acceleration, the maximum amplitude of lateral acceleration, and the range normalized maximum amplitude of lateral acceleration. The overall results show that these three gait features can be used to diagnose PN, as they reveal that gait patterns in DPN patients are less extreme and exhibit less sway in the lateral direction. Additionally, step length and gait speed are relatively higher in PN patients compared to healthy individuals. Among all five proposed models in this research, Model E, which consists of inertial data as well as MNSI history, outperformed all other proposed models and suggests that valuable information can be added by wearable sensors in the screening of PN in order to make the diagnosis process more accurate and diagnose the disease in time.

Cohen et al. ^[9] proposed a framework for the diagnosis of peripheral neuropathy based on an inertial measurement unit (IMU) and tandem walking test. The main focus of this study was to determine whether the tandem walking test can be performed successfully for the screening of peripheral neuropathy. In the experimental study, both healthy subjects and subjects with peripheral neuropathy were recruited. All subjects could walk without any gait aid. The experiment included 21 subjects with PN (13 males, 8 females—age 60 ± 12.4 years) (mean ± SD) including small fiber, large fiber, or mixed large and small fiber neuropathy, and 61 healthy subjects (31 males, 30 females—age 49.6 ± 16.0 years). This study was approved by the Institutional Review Board for Human Subjects Research of Baylor College of Medicine, Houston, TX, USA. The ground truth was provided by a medical expert using electromyography (EMG). At the time of the walking trials, subjects wore only socks for hygiene purposes. The walking task consists of a 10-step walk on the industrial carpeting in two different phases. In the first phase, the subjects were asked to walk with their eyes open; in the second phase, the subjects were asked to perform the walking task with their eyes closed. The IMU sensor was mounted on the torso of each subject in order to collect gait data. From the raw sensor data, mean square values of resultant acceleration, angular velocity about the roll axis, angular velocity about the pitch axis, and angular velocity about the yaw axis were measured. To calculate the differences in the dependent measures, multilevel statistical techniques were used including separate models fitted to each dependent variable. Receiver operating characteristic (ROC) measures were taken into consideration for the statistical analysis of the sensor signal. Chi-square distribution was used to determine changes in eye open/closed conditions between pathological and non-pathological subjects. The experimental results in this study show that it is more likely for PN subjects to take significantly more consecutive steps with eyes open than eyes closed, while healthy individuals took more consecutive steps than PN subjects in both eyes open and closed conditions. The motion data analysis results indicated that PN subjects have higher angular velocity about the roll axis, angular velocity about the pitch axis, and angular velocity about the yaw axis. The results also showed that PN subjects show greater instability compared to healthy persons while performing the tandem task with their eyes closed.

Esser et al. ^[10] used a single IMU to analyze gait in order to diagnose peripheral neuropathy in diabetic patients. The main aim of the study was to analyze human gait using an IMU sensor during a 10-m walk test. The IMU was mounted on the lower back, and data from the accelerometer and gyroscope were recorded at the sampling rate of 100 Hz for further processing. By using the collected data, spatiotemporal gait parameters were extracted from the sensor data. The extracted parameters included step time, cadence, stride length, and walking speed. The data analysis for each group in order to find group differences was performed using the chi-square statistical method. This method compares the distribution of categorical variables in a sample with the distribution of categorical variables in another sample. For statistical analysis, this study uses ROC curves by means of the area under the curve (AUC). The experimental set-up included 17 participants (14 males, 3 females) with DPN, and 42 healthy participants (30 males, 12 females) aged around 63.2 ± 9.2 years. There was no difference in age, gender ratio, height, or BMI between groups. The participants were recruited from the Oxford Centre for Diabetes, Endocrinology and Metabolism at the Oxford University Hospitals NHS Foundation Trust (Oxford, UK), and the study had approval from the National Research Ethics Committee (NRES: 11/SC/0218). The PN patients were confirmed using a monofilament test carried out by a specialist diabetes podiatrist. The obtained results in this study indicate that significant differences were found for all spatiotemporal parameters between PN patients and healthy subjects except for stride length. However, walking speed differed significantly in unhealthy and control subjects compared to any other gait-related parameters, while producing the largest discriminatory power (AUC = 0.975).

Wang et al. ^[11] proposed a framework for the diagnosis of peripheral neuropathy and other neurological disorders using two IMU sensors ((InvenSense MPU-6050) in order to analyze human gait. The two IMUs were used to measure five key kinematic and three spatiotemporal gait parameters that can help in distinguishing the type of neurological disorder. These parameters capture the kinematics of the ankle’s linear motion and shank rotation, as dysfunction of lower-limb segments and joints would impact the motion of the ankle and shank. The IMUs were mounted on the ankle of each shank in the sagittal plane on the lateral side. The data were collected at a sampling frequency of 100 Hz. The study included 8 patients with PN (3 males, 5 females—age 49 + 8), 13 patients with post-stroke (PS) (9 males, 4 females—age 61 ± 15 years), 15 patients with Parkinson’s disease (PD) (9 males, 6 females—age 76 ± 7 years), and 13 healthy subjects (HC) (7 males, 6 females—age 49 ± 20 years). Additionally, information about subjects’ heights and weights was provided in ^[11]. The study was approved by the Medical Ethics Committee of the School of Medicine at Zhejiang University, China. As can be observed, the data between each group are not balanced based on age or female/male ratio. In walking trials, the participants were asked to walk at a convenient speed for more than 12 m on a flat surface. From the IMU data, eight features were extracted based on detected gait phases and calculated motion trajectories i.e., stride length (SL), gait cycle duration (GD), percentage swing phase (PSP), max ankle velocity (MV), max ankle height (MH), ankle horizontal displacement (MHD), range of shank motion (RS), and kinematic asymmetry (KS). Using extracted features, a support vector machine (SVM) classifier was used to distinguish among four classes. The algorithm for training and validation of the proposed system was carried out in MATLAB. The classification accuracy achieved was 93.9% in this case. In this study, the authors used separate SVM classifiers for each of the four classes. In this way, four SVM classifiers were trained to distinguish data points of one class of subject from another class. During the training process, linear kernel function and sequential minimal optimization (SMO) method were used.

2. Pressure Sensor-Based Intelligent Wearable System for Diagnosis of PN

Foot plantar pressure is the distribution of the pressure field that acts between the foot and the surface ^[12]. It plays a very important role in the diagnosis of peripheral neuropathy. As in most cases, due to the damage of foot nerves which is common in diabetic patients, the patient cannot feel the right amount of plantar pressure required to walk smoothly. Due to nerve damage, the sensitivity of the foot decreases. Hence, patients having peripheral neuropathy in the lower limb will always exert more pressure than the healthy person while walking ^[13]. This makes it important for diabetic patients to have regular check-ups, as symptoms of PN will appear in later stages of the disease where it is not possible to recover the organ, and it may result in amputation of the foot ^[14]. By using a wearable pressure sensor, the early diagnosis of PN can be made possible, as wearable systems can be used in clinical settings as well as remotely ^[15]. Further, it is difficult for doctors to examine patients by simply observing the gait of the patients. Wearable sensors can provide an objective measurement and detailed knowledge about the physiology of the patient which is necessary to keep people’s lives healthy ^[16]. There are a variety of pressure measurement systems available; however, they are broadly classified into two types: (1) platform systems and (2) in-shoe systems.

Cao et al. ^[17] proposed a method for the diagnosis of PN based on foot plantar pressure distribution. The main aim of the study was to analyze foot plantar pressure changes while measuring foot plantar pressure distribution in order to prevent ulcers in elderly diabetic people. The study further investigates the role of plantar pressure in elderly diabetic patients with and without PN and compares pressure distribution between healthy subjects and diabetic patients with and without peripheral neuropathy. This study includes foot plantar pressure data from 19 diabetic patients with peripheral neuropathy (DPN) (10 males, 9 females—age 65.7 ± 2.4 years), 17 diabetic patients without peripheral neuropathy (D) (9 males, 8 females—age 65.2 ± 6.8 years), and 20 healthy subjects (H) (11 males, 9 females—age 65.2 ± 5.4 years). Patients were recruited from Tianjin Medical University Chu Hsien-I Memorial Hospital in China. All subjects agreed to participate and signed informed consent after being fully informed of the study’s procedure. The data were collected by using an insole wireless plantar pressure monitoring system designed by Medilogic, USA, at a sampling frequency of 300 Hz. The sensor can provide continuous data up to 10 m. The sensor was placed between the sole and the socks. The experimental protocol consisted of 10 m of walking trials at a speed suitable and comfortable for the subject. From plantar pressure data, the authors divided the plantar area into seven regions based on anatomical structure.

After collecting the plantar pressure data, the changes in peak pressures in segmented areas were analyzed while the subject was performing the walking task. In this study, the peak pressure was taken as the highest pressure in each segmented area during one gait cycle. A value of pressure higher than 200 KPa was considered high pressure. In this study, two insole pressure sensors were used to record pressure distribution in both feet. From sensor data, average pressure values of the right and left foot were calculated for each segment area in order to find statistical differences in each subject. The value of peak pressures in normal and DPN patients were investigated, clearly showing that DPN patients tend to have higher pressure in every segmented area of the foot compared to healthy or diabetic subjects without DPN. However, no differences were found in the peak pressure of healthy subjects and diabetic patients without PN. Results showed that the most sensitive areas related to the change in foot plantar pressure include the inner forefoot and medial forefoot region. Additionally, the peak pressure of the forefoot region in DPN patients is much higher compared to the peak pressure in the rear foot region. The overall study suggests that a significant increase in plantar pressure at the forefoot region was observed in DPN patients in the standing position, while healthy individuals in the standing position put more pressure on the rear foot.

Corpin et al. ^[18] proposed a model for the prediction of DPN by analyzing foot plantar pressure data. The Tekscan Medical Sensor 3000E hardware and Tekscan F-Scan 7.50 Research Software were used to collect the plantar pressure distribution of healthy subjects and DPN patients and to train different machine learning classifiers to distinguish between pathological and non-pathological subjects. The data collection step included 36 normal and diabetic volunteers and both female and male volunteers; however, the ratio between female and male volunteers was not provided. The volunteers’ ages were between 49 and 56 years. The study was approved by the Ethics Committee under the supervision of the University of Santo Tomas Hospital (USTH), Manila, Philippines. The ground truth about the diagnosed case was obtained by using the Michigan Screening Instrument-questionnaire (MNSI-q). Nerve conduction velocity studies (NCV) were also conducted on each subject in order to find the ground truth. This research aims to classify among three classes: (1) healthy individuals (N); (2) individuals with diabetes but without peripheral neuropathy (DM); and (3) diabetic patients with peripheral neuropathy (DPN). Both male and female volunteers were included in these experimental trials. During walking trials, each subject had to walk 7 m in a straight line. Each subject had to perform the same walking trial eight times in order to obtain a sizable dataset. The subjects were asked to walk in their normal style during the trial. The Tekscan Medical Sensor 3000E hardware consists of 960 individual pressure sensing points to collect dynamic plantar pressure data. The overall system was based on an in-shoe pressure measurement system. The output of each sensor point was divided into 256 increments to make the visualization better. The software divided the foot into 13 different regions by mapping the pressure data to accurately form the outline of the foot.

The parameters that were calculated by the software included peak pressure (PP), the instant of maximum force (IMxF), the instant of peak pressure (IPP), pressure–time integral (PTI), force–time integral (FTI), length of contact (LC), and contact area (CA). The data were then analyzed statistically using one-way ANOVA. Two different datasets were created for each leg separately, as some features were significantly different between the left and right foot. Principle component analysis (PCA) was also used to determine potential features and to remove those features that contributed less to the learning process by reducing the dimension of the dataset. From 208 features, PCA reduces the dimension of the dataset to only 29 new features that can successfully represent the overall data, i.e., 95%. Different machine learning classifiers such as SVM, random forest, multilayer perceptron (MLP), K-nearest neighbor (KNN), and Gaussian process (GP) were trained and tested in order to distinguish between three given classes. The k-fold cross-validation algorithm was also used to validate the system’s performance. The results of this research showed that the parametric difference between the right and left foot is an indication of asymmetric plantar pressure distribution. Hence, separate datasets were used to compare parameters from both feet. The results also stated that instant maximum force time (IMxFT) and contact area (CA) on the right foot exhibit multiple significance in different regions. However, if there is a significant difference in the contact area between the left and right foot, it indicates the presence of DPN. For classification purposes, among the five classifiers, the SVM classifier outperforms other classifiers, with the highest accuracy of 91.91%.

Wang et al. ^[19] proposed a wireless footwear system to monitor diabetic foot ulcers due to peripheral neuropathy in diabetic patients. The system consists of an insole pressure sensor array that captures pressure changes during walking and transfers data via Bluetooth to a mobile phone in real time. So, the proposed methodology offers the continuous monitoring of plantar pressure using an insole flexible pressure system. A composite piezoresistive flexible sensor was developed to fulfill long-term monitoring requirements. The construction of the sensor was composed of carbon black and silicon rubber. The functioning of the sensor was validated by a custom pressure testing platform, which consists of a keyboard testing machine, a pressure testing machine, and a desktop computer. The data of 5 healthy subjects (HC) (4 males, 1 female—age 48.5 ± 3.5 years), 5 diabetic patients without PN (D) (2 males, 3 females—age 55.8 ± 5.6 years), and 5 DPN patients (2 males, 3 females—age 59.00 ± 10.71 years) were recorded by the proposed pressure measurement system. All subjects were able to walk without any gait aids. The study was approved by the Research Ethics Committee of the Body Data Science Engineering Center of Guangdong Province and the First Affiliated Hospital of Jinan University of Guangdong Province in China. In the walking trial, subjects were asked to walk 20 m along the corridor and stairs for one minute at their regular speed. The data were captured at a sampling frequency of 20 Hz with a 12-bit sampling resolution. The application interface for smartphones was also designed to allow users to visualize the pressure distribution in real time. The final database consists of 2403 samples, including 779 samples from the HC group, 736 from the D group, and 888 samples recorded for DPN patients. From the raw sensor data, peak plantar pressure (PPP), the pressure–time integral (PTI), the maximum pressure gradient (MaxPG), the minimum pressure gradient (MinPG), the full width at half maximum (FWHM), the forefoot-to-rearfoot plantar pressure ratio (F/R), and the symmetry index (SI) were extracted for the database feature points. The feature points used in this study were calculated using the method given by Botros et al. ^[20]. Five different machine learning classifiers, support vector machine (SVM), K-nearest neighbors (KNN), RF (random forest), GBDT (gradient-boosted decision trees), and AdaBoost classifiers were trained, and they tested the performance of the proposed system. A 10-fold cross-validation was also used to validate the ML models. The overall average accuracy of all five classifiers used was 85%, with the highest accuracy of 94.7% in the case of the random forest (RF) classifier. It should be noted that the sample size of the participants (five per class) is a small number, and further evaluation is needed with a higher number of participants.

3. ECG-Based Intelligent Wearable Systems for Diagnosis of PN

According to ^[21], almost one-third of acute myocardial infarction patients have diabetes, which is one of the leading causes of peripheral neuropathy. The damage to these peripheral nerves that regulate the heart mechanism is called cardiac autonomic neuropathy (CAN) ^[22]. Electrocardiography (ECG) is a quick and non-invasive procedure for the early detection of CAN, which is caused by the damage of those peripheral nerves that are responsible for the proper functioning of the heart.

In Ref. ^[23], an ECG-based platform for the diagnosis of peripheral neuropathy was proposed. The diagnosis was based on the fact that uncontrolled high glucose levels in diabetic patients cause cardiovascular diseases because they affect heart rate variability (HRV). This study is focused on analyzing HRV parameters in DPN patients so that normal and abnormal ECG can be distinguished. The HRV analysis has the capability to recognize variations in the autonomic nervous system (ANS), which is responsible for keeping the heart functioning properly ^[24]. The data collection method involved the recording of an ECG for all participants for 24 h using a four-channel Holter machine. Twenty subjects were enrolled for the data collection task. All 20 subjects (10 males and 10 females with a mean age of 55.7 years old) were type-II diabetes mellitus patients and were all 40 years old or above. Of the 20 subjects, 10 subjects were confirmed diagnosed with DPN using nerve conduction studies (NCS). The study was approved by the ethical committee of Bangladesh University of Health Sciences. The ECG data were collected at a 200 Hz sampling rate and then the signal was processed in order to remove any possible noise. The ECG signal was then processed in order to extract nine feature points related to HRV to distinguish between DPN-positive and DPN-negative groups. Both time-domain and frequency-domain parameters of the ECG signal were extracted for classification purposes. The Wilcoxon rank-sum test ^[25] was used to find statistical significance between these two classes. The obtained results showed the feasibility of ECG data for the purpose of diagnosing peripheral neuropathy related to the heart.

In Ref. ^[26], Jelinek et al. also used HRV attributes of the ECG signal in order to distinguish between the DPN group and the healthy control group. This study proposed a new classification technique for diagnosis purposes and compared the performance of other machine learning (ML) classifiers with the proposed one. The authors utilized their prior collected dataset called Diab Health ^[27] and selected a subset of 21 patients with severe diabetic neuropathy. However, no information was provided on how the subset was selected. The main aim of the study was to investigate the contribution of HRV parameters in an automated disease classification task. A multi-level clustering technique was used to improve diagnostic accuracy. The data collection procedure included 20 min of ECG recording in a supine position for all participants. The ECG signals were recorded at a sampling frequency of 400 Hz with a lead II configuration ^[28]. The ground truth for CAN patients was acquired from Ewing battery criteria ^[29]. The results of this study indicate the significance of HRV parameters for the diagnosis of cardiovascular diseases and the proposed graph-based machine learning classification algorithm (GBML) performed better compared to other conventional clustering techniques. The performances of ML algorithms were based on sensitivity and specificity, which are common metrics for ML algorithms ^[30]. The best sensitivity of 0.98 and the best specificity of 0.89 were achieved in the case of the GBML clustering technique.

Sharanya and Sridhar ^[31] proposed a system for the diagnosis of CAN that is based on a convolutional neural network (CNN) for prediction purposes. The aim of this study was to classify CAN-positive and CAN-negative subjects by analyzing ECG signals. In the experimental trials, 13 male and 6 female subjects participated. Among them, 9 subjects were confirmed positive for CAN, and 10 were healthy subjects, labeled as CAN-negative class. No information on the age of the participants or approval of the study from an institutional review board or an ethical committee was provided. The ECG signals were acquired at a sampling frequency of 400 Hz in lead II configuration for 20 min. After acquiring the ECG signal, the signal was processed to remove possible noise, and then feature points such as RR-intervals, etc., were extracted from the clean ECG signal. Using the extracted features, a CNN model was trained to distinguish between normal ECG and abnormal ECG, representing the presence of CAN. The overall results show that CNN can take care of analyzing the complexity of the ECG attributes for the early diagnosis of CAN. The overall accuracy for the diagnosis of CAN achieved in this study is 95.42%. Research work discussed here for the diagnosis purpose using a single wearable sensor is summarized in Table 1.

Table 1. Summary of the wearable systems for detecting peripheral neuropathy (PN) using a single-type sensor. H: healthy; DPN: diabetic PN; CAN: cardiac autonomic neuropathy; PD: Parkinson’s disease; ACC: accelerometer; IMU: inertial measurement unit; N/A: not available.

References	No. of Participants	Distribution of Participants	No. of Sensors and Placement	Data Collection Procedure	Methodology	Results
Chen and Shanshan ^[8]	106	PN: 30 H: 76	3-axis ACC	Single ear-worn ACC	10-m walking test	Gait analysis using logistic regression models for training and testing
Cohen et al. ^[9]	72	PN: 21 H: 61	IMU	Single torso-mounted IMU	Tandem walking test	ROC and Chi-square methods were used to evaluate the gait data.
Esser et al. ^[10]	56	PN:14 H: 42	IMU	Single IMU attached to lower back	Standard 10-m walking test	Chi-square distribution using IMU Sensor data for classification. Statistical analysis was conducted using ROC.
Wang et al. ^[11]	49	PN: 9 Stroke: 13 PD: 14 H: 13	IMUs	Two IMUs were attached to the ankle of each shank	12-m walking trail	The gait parameters were extracted using method ^[32] based on wavelet analysis.
Cao et al. ^[17]	56	DPN: 19 Diabetic: 17 H: 20	Insole wireless plantar pressure monitoring system designed by Medilogic	The sensor was placed on one foot between the soles and socks of the participants	10-m walking test recorded at 300 Hz sampling	Peak pressure was recorded in each case by dividing the foot into seven segments and then comparing the pressure distribution of each region in each of the two classes
Corpin and Ryan Rey A. ^[18]	36	N/A	Tekscan Medical Sensor 3000E	Single Tekscan Medical sensor placed on the right foot only.	7 m walking in a straight line and repeat the procedure eight times	In-shoe pressure monitoring system was used. The Tekscan software provides a number of gait and pressure parameters that can be used as features for ML algorithms
Wang et al. ^[19]	20	DPN: 5 H: 5	Insole piezoresistive pressure sensor array	Two insole pressure sensors that each contained eight pressure measuring points were placed on both feet.	20-m walking test	Using the proposed insole system, the pressure data were collected from each sensing point, and peak pressures were recorded to create a database of healthy and unhealthy subjects. Five different classification algorithms were then trained for the diagnosis, and the model was validated by using k-fold validation.
Morshed et al. ^[23]	20	DPN: 10 H: 10	Holter device	Four-channel (RA-LA, LA-LL, LL-RA, and Vx-RL) Holter device	24-h ECG recording at 200 Hz	HRV parameters were extracted from ECG data using a method in ^[33]. Both time-domain and frequency-domain features of the ECG signal were used in the diagnosis of PN.
Jelinek et al. ^[26]	21	DPN: 21	ECG	ECG sensor with lead II configuration	20-min ECG recording in spine position	Using HRV attributes of the ECG signal, a new multi-level clustering technique was proposed and implemented to distinguish between two classes.
Sharanya, S. and P.A. Sridhar ^[31]	19	CAN: 9 H: 10	ECG	ECG sensor with lead II configuration	20-min ECG recording	A CNN network was used to distinguish between PN and healthy subjects. A 20-min-long ECG was recorded for each subject.

References

Cuesta-Vargas, A.I.; Galan-Mercant, A.; Williams, J.M. The use of inertial sensors system for human motion analysis. Phys. Ther. Rev. 2010, 15, 462–473.
Tao, W.; Liu, T.; Zheng, R.; Feng, H. Gait analysis using wearable sensors. Sensors 2012, 12, 2255–2283.
Pan, Q.; Chen, Y.; Ma, X.; Wang, C.; Chen, W. Application of Wearable Technologies in Fall Risk Assessment and Improvement in Patients with Peripheral Neuropathy: A Systematic Review. J. Sens. 2023, 2023, 1746536.
Anwary, A.R.; Yu, H.; Vassallo, M. Gait Evaluation Using Procrustes and Euclidean Distance Matrix Analysis. IEEE J. Biomed. Health Inform. 2019, 23, 2021–2029.
Wang, C.; Li, Z.; Sarpong, B. Multimodal adaptive identity-recognition algorithm fused with gait perception. Big Data Min. Anal. 2021, 4, 223–232.
Talha, M.; Soomro, H.A.; Naeem, N.; Ali, E.; Kyrarini, M. Human Identification Using a Smartphone Motion Sensor and Gait Analysis. In Proceedings of the PETRA ‘22: The15th International Conference on PErvasive Technologies Related to Assistive Environments, Corfu, Greece, 29 June–1 July 2022; pp. 197–202.
Wang, L.; Sun, Y.; Li, Q.; Liu, T.; Yi, J. IMU-Based Gait Normalcy Index Calculation for Clinical Evaluation of Impaired Gait. IEEE J. Biomed. Health Inform. 2021, 25, 3–12.
Chen, S.; Kang, L.; Lu, Y.; Wang, N.; Lu, Y.; Lo, B.; Yang, G.-Z. Discriminative Information Added by Wearable Sensors for Early Screening—A Case Study on Diabetic Peripheral Neuropathy. In Proceedings of the 2019 IEEE 16th International Conference on Wearable and Implantable Body Sensor Networks (BSN), Chicago, IL, USA, 19–22 May 2019; pp. 1–4.
Cohen, H.S.; Mulavara, A.P.; Peters, B.T.; Sangi-Haghpeykar, H.; Kung, D.H.; Mosier, D.R.; Bloomberg, J.J. Sharpening the tandem walking test for screening peripheral neuropathy. South Med. J. 2013, 106, 565–569.
Esser, P.; Collett, J.; Maynard, K.; Steins, D.; Hillier, A.; Buckingham, J.; Tan, G.D.; King, L.; Dawes, H. Single Sensor Gait Analysis to Detect Diabetic Peripheral Neuropathy: A Proof of Principle Study. Diabetes Metab. J. 2018, 42, 82–86.
Wang, L.; Sun, Y.; Li, Q.; Liu, T.; Yi, J. Two Shank-Mounted IMUs-Based Gait Analysis and Classification for Neurological Disease Patients. IEEE Robot. Autom. Lett. 2020, 5, 1970–1976.
Abdul Razak, A.H.; Zayegh, A.; Begg, R.K.; Wahab, Y. Foot plantar pressure measurement system: A review. Sensors 2012, 12, 9884–9912.
Volmer-Thole, M.; Lobmann, R. Neuropathy and Diabetic Foot Syndrome. Int. J. Mol. Sci. 2016, 17, 917.
Cade, W.T. Diabetes-related microvascular and macrovascular diseases in the physical therapy setting. Phys. Ther. 2008, 88, 1322–1335.
Majumder, S.; Mondal, T.; Deen, M.J. Wearable Sensors for Remote Health Monitoring. Sensors 2017, 17, 130.
Schlachetzki, J.C.M.; Barth, J.; Marxreiter, F.; Gossler, J.; Kohl, Z.; Reinfelder, S.; Gassner, H.; Aminian, K.; Eskofier, B.M.; Winkler, J.; et al. Wearable sensors objectively measure gait parameters in Parkinson’s disease. PLoS ONE 2017, 12, e0183989.
Cao, Z.; Wang, F.; He, Y.; Zhang, Y.; Zhang, J. Analysis of plantar pressure in elderly diabetic patients with peripheral neuropathy. In Proceedings of the 2021 International Conference on Public Health and Data Science (ICPHDS), Chengdu, China, 9–11 July 2021; pp. 184–187.
Corpin, R.R.A.; Guingab, H.A.R.; Manalo, A.N.P.; Sampana, M.L.B.; Abello, A.N.A.; Cruz, A.R.D.; Roxas, E.A.; Suarez, C.G.; Serrano, K.K.D. Prediction of Diabetic Peripheral Neuropathy (DPN) using Plantar Pressure Analysis and Learning Models. In Proceedings of the 2019 IEEE 11th International Conference on Humanoid, Nanotechnology, Information Technology, Communication and Control, Environment, and Management (HNICEM), Laoag, Philippines, 29 November–1 December 2019; pp. 1–6.
Wang, D.; Ouyang, J.; Zhou, P.; Yan, J.; Shu, L.; Xu, X. A Novel Low-Cost Wireless Footwear System for Monitoring Diabetic Foot Patients. IEEE Trans. Biomed. Circuits Syst. 2021, 15, 43–54.
Botros, F.S.; Taher, M.F.; ElSayed, N.M.; Fahmy, A.S. Prediction of diabetic foot ulceration using spatial and temporal dynamic plantar pressure. In Proceedings of the 2016 8th Cairo International Biomedical Engineering Conference (CIBEC), Cairo, Egypt, 15–17 December 2016; pp. 43–47.
Voulgari, C.; Tentolouris, N.; Stefanadis, C. The ECG vertigo in diabetes and cardiac autonomic neuropathy. Exp. Diabetes Res. 2011, 2011, 687624.
Serhiyenko, V.A.; Serhiyenko, A.A. Cardiac autonomic neuropathy: Risk factors, diagnosis and treatment. World J. Diabetes 2018, 9, 1–24.
Morshed, M.G.; Mukit, M.A.; Ahmed, K.I.U.; Mostafa, R.; Parveen, S.; Khandoker, A.H. Heart rate variability analysis for diagnosis of diabetic peripheral neuropathy. In Proceedings of the 2020 IEEE Region 10 Symposium (TENSYMP), Dhaka, Bangladesh, 5–7 June 2020; pp. 1253–1256.
Vinik, A.I.; Erbas, T.; Casellini, C.M. Diabetic cardiac autonomic neuropathy, inflammation and cardiovascular disease. J. Diabetes Investig. 2013, 4, 4–18.
Bridge, P.D.; Sawilowsky, S.S. Increasing physicians’ awareness of the impact of statistics on research outcomes: Comparative power of the t-test and and Wilcoxon Rank-Sum test in small samples applied research. J. Clin. Epidemiol. 1999, 52, 229–235.
Jelinek, H.F.; Cornforth, D.J.; Kelarev, A.V. Machine Learning Methods for Automated Detection of Severe Diabetic Neuropathy. J. Diabet. Complicat. Med. 2016, 1, 1000108.
Herbert Jelinek, C.W.; Tinley, P. An innovative multi-disciplinary diabetes complications screening program in a rural community: A description and preliminary results of the screening. Aust. J. Prim. Health 2006, 12, 14–20.
Tung, R.T. Electrocardiographic Limb Leads Placement and Its Clinical Implication: Two Cases of Electrocardiographic Illustrations. Kans. J. Med. 2021, 14, 229–230.
Pafili, K.; Trypsianis, G.; Papazoglou, D.; Maltezos, E.; Papanas, N. Simplified Diagnosis of Cardiovascular Autonomic Neuropathy in Type 2 Diabetes Using Ewing’s Battery. Rev. Diabet Stud. 2015, 12, 213–219.
Hossin, M.; Sulaiman, M.N. A Review on Evaluation Metrics for Data Classification Evaluations. Int. J. Data Min. Knowl. Manag. Process 2015, 5, 1–11.
Sharanya, S.; Sridhar, P.A. A model for early diagnosis of Cardiac Autonomic Neuropathy (CAN). J. Phys. Conf. Ser. 2021, 2089, 012053.
Aminian, K.; Najafi, B.; Büla, C.; Leyvraz, P.-F.; Robert, P. Spatio-temporal parameters of gait measured by an ambulatory system using miniature gyroscopes. J. Biomech. 2002, 35, 689–699.
Pan, J.; Tompkins, W.J. A real-time QRS detection algorithm. IEEE Trans. Biomed Eng. 1985, 32, 230–236.

© Text is available under the terms and conditions of the Creative Commons Attribution (CC BY) license; additional terms may apply. By using this site, you agree to the Terms and Conditions and Privacy Policy.

Upload a video for this entry

Information

Subjects: Engineering, Electrical & Electronic

Contributors MDPI registered users' name will be linked to their SciProfiles pages. To register with us, please refer to https://encyclopedia.pub/register :

Muhammad Talha

Maria Kyrarini

Ehsan Ali Buriro

View Times: 185

Update Date: 30 Nov 2023

Table of Contents

Video Upload Options

Confirm

1. Wearable Inertial Sensor-Based Intelligent Systems for the Diagnosis of PN

2. Pressure Sensor-Based Intelligent Wearable System for Diagnosis of PN

3. ECG-Based Intelligent Wearable Systems for Diagnosis of PN

References