>

Severity assessment of spastic cerebral palsy patients using gait cycle and fuzzy logic

   Massoud S. (1)* Ismaiel E. (2) Massoud R ⁽¹⁾  AL-Mawaldi M.  ⁽¹⁾ 

¹ Biomedical Engineering Department, Faculty of Mechanical and Electrical Engineering, Damascus University, Syria.

² Faculty of Biomedical Engineering, Al-Andalus University, Syria

* Corresponding author: saleh2.massoud@damascusuniversity.edu.sy

 

 

مؤلفون / Authors

 

 Introduction

Methods

 Results and discussion

Conclusion

References

Abstract

Gait analysis has many applications in quantitatively assessing normal and impairment in human gait. In gait monitoring, many parameters, including spatiotemporal and kinetic parameters, are calculated to quantify or score the human's gait. In this paper, we use fuzzy logic to assess the gait impairment of SCP and healthy subjects using the dataset of gait analysis. This proposed methodology starts with extracting distinct features from the force data during the gait cycle. Initializing FL system based on extracted features. Finally, we evaluate the available samples from the dataset using the proposed FL system to assess the available cases using a percentage scale of severity. The FL results show that the severity of SCP children ranged between 82 and 88%, while the evaluations of healthy children ranged between 5 and 13%. The proposed methodology shows how important it is to work on extracting features from children's gait data and how efficient input provides for fuzzy logic. The paper also presents a first model for a medical decision support system that can assess pathological cases, as well as evaluate the efficiency of treatment and intervention based on FL.

 

Keywords:

 Gait Analysis, spastic cerebral palsy, fuzzy logic.

Introduction

Gait analysis applications include quantitatively assessing normal and pathological gait function and monitoring various orthopedic and neurological disorders, planning and evaluating surgery outcomes, documenting functional changes during patient follow-up, and evaluating the effectiveness of rehabilitation protocols. Many parameters, including spatiotemporal and kinetic parameters, are calculated to assess the patient's gait numerically or on a scale [1]. The existence of numerous parameters, as well as the uncertainty associated with each parameter, makes accurately assessing a patient's gait difficult. Furthermore, comparing results for a particular patient at various times or clinical points makes determining treatment efficacy difficult.  Accordingly, data complexity hampered the clinical use of gait analysis in many clinical trials, raising awareness of the need for a summary index, a single measure of the quality of human gait pattern that addresses the maximum number of gait parameters that have the greatest impact on medical diagnosis [2].

Accordingly, considerable effort has been expended in developing indices that summarize and condense the information derived from many gait parameters as a result of gait analysis into a singular score.

Fuzzy logic (FL) is one of the most reliable artificial intelligence tool in evaluating and classifying the human gait due to his ability to mimic human reasoning. FL produces accurate outputs by using reasoning methods based on a set of inputs and knowledge bases that representing the human expert. Moreover, it enables accurate gait evaluation [3,4,5,6] and assisting clinicians in understanding the mechanism of gait [7].

The application of fuzzy logic in the clinical studies of human gait is still rare. On the other hand, studies have shown that it has the ability to deal with the large number of parameters studied in gait analysis and the possibility of assembling and expressing them according to a semantic expression that describes the degree of weakness or deviation in human gait.

 

Cerebral palsy (CP) is the most prevalent motor impairment in childhood. Cerebral palsy is a collection of diseases that impair a person's ability to move, balance, and keep equilibrium. Spastic CP (SCP) refers to the damage or issue in is in the motor cortex of the brain. Where the motor cortex is in charge of action planning and regulation, a SCP kid suffers from gait impairment and instability [8]. Recently, Al-Mawaldi and Khadour (2022) presented a clinical study involving gait analysis of SPC and healthy children [9]. The study included 10 individuals with spastic CP who endured (SEMLS) at the age of (9.2±3.2) years. And a set of 12 healthy children aged (8.2±2.2) years. They compared kinetic parameters such as walking Profile Score (GPS) before and after Single Event Multilevel Surgery (SEMLS) on walking for Spastic Cerebral Palsy children.

In this paper, we employ fuzzy logic to assess the gait impairment of SCP and healthy subjects using the dataset of [8]. This proposed methodology starts with extracting distinct features from the force data during the gait cycle. Initializing FL system based on extracted features. Finally, we evaluate the available samples from the [9] dataset using the proposed FL system.

 

 

 

 

Methods

 

 Data collection

 

The dataset of this study involves gait analysis of 10 children, 5 subjects are diagnosed with SCP (9.2±3.2 years) and the 5 healthy subjects (8.2±2.2 years). The experimental procedures of data acquisition involved a computerized three-dimensional gait analysis system to determine the distance and time parameters, and the kinetic and kinematic activities in a group of children with spastic paralysis and a group of healthy children. The gait analysis was conducted in the Biomechanics Laboratory at Medical Engineering, Department of Mechanical and Electrical Engineering, University of Damascus, using SMART-D Motion Capture System [9]. 

The laboratory has two Swiss Kistler force plates working synchronously with the camera, with an sampling frequency of 200 Hz. When a force is applied to the plate, an electric charge is generated and converted into an electrical ouput corresponding to the force applied to the plates. The numerical output contains the reaction forces (Fx, Fy, Fz) and the momentum components of these forces (Mx, My, Mz) about the center of the force plate.

 

 Features extraction

By visual inspection among force data of SCP and healthy samples, we found a sufficient difference in Fx and Fz components during the gait cycle. The first important feature is the negative amplitude of Fx which equals the vertical distance between the maximum negative peak of force and the average of the same Fx components as shown in Fig. 1 which labelled as “NP”.

The second important feature represents the peak-to-peak distance in the Fz component. This distance where found to have a very distinguished change among healthy and SCP subjects. The “PP” second feature is computed from the maximum and minimum amplitude of 5-25% of the gait cycle.

Figure 1. Extracted features from Fx and Fz components during the 60% of gait cycle.

 

 

 

3Fuzzy logic

 

In this study, a Mamdani-type FL system was used to examine the Fx and Fz components of the gait cycle by using two extracted features as input and estimating the score of SP severity on a scale of 100 (Figure 2). 

Figure 2. Proposed FL system to estimate the severity of SP.

 

To process the extracted features and obtain the prospective score. FL goes through several stages to reach a complete solution as follows [10]:

(1) Fuzzification: includes defining the membership functions (MF) for the input variables to determine the degree of truth in each rule. The FL has two inputs. The first input is the NP feature where we express its changes among healthy and SP cases using three triangle MFs (Figure 3). Where the first MF of NP feature is labelled as “ModP” referring to moderate SP, the second “MilP” referring to mild SP and the third “H” MF referring to healthy values. The second input represents the PP feature which is divided into four MFs (Figure 3) and those functions are labelled as “H”, “MilP”, “ModP” and “SevP” according to the severity of the SP case. The MF of FL output has similar MFs of PP input as labels and MF type but the horizontal axis represents the severity from 0 to 100%.

 

Figure 3. The membership functions of FL inputs

 

.

(2) Inference: contains the fuzzy if-then rules (Figure 4). In this study, the fuzzy rules were built based on the experimental observations of SP data and prior knowledge about the case diagnosis. Where the “max” function is used as the OR method, the implication used the “min” function and the “max” function is used for aggregation

 

Figure 4. The fuzzy rules of proposed FL system. 

(3) Defuzzification: the Mean of Maximum (MoM) defuzzification method was used to calculate the severity of inputs.

 

 

Results and discussion

This study presents a system for evaluating the gait cycle and detecting cases of SCP in addition to their level of severity by extracting some important features and using them as input for a fuzzy system. Those features differ in their values between the healthy and SCP cases.

In this study, two important features were extracted from Fx and Fz components that related to the patient's weight.

After extracting features from force components of 5 healthy and 5 SCP subjects, then implementing the fuzzy logic as shown in Figure 1. The FL output is illustrated in Table 1 which shows the evaluation output as a percentage.

 Table 1. The severity of examined subjects using FL.

This percentage represents the severity of examined subject and how much is close to being healthy (0%) and severe SCP (100%).  The table shows that the evaluation of SCP children ranged between 82 and 88%, while the evaluations of healthy children ranged between 5 and 13%.

As preliminary results, these results show the efficiency of using a fuzzy logic system based on the extracted features in assessing normal and SCP cases effectively and broadly. The proposed methodology shows how important it is to work on extracting features from children's gait data and how efficient input provides for fuzzy logic. The paper also presents a first model for a medical decision support system that can assess pathological cases, as well as evaluate the efficiency of treatment and intervention [9] based on FL.

 

 

 

Conclusion

 

In this paper, we found that the two components Fx and Fz have an important role in differentiating between healthy and SCP children. The study proposed an FL system to assess the severity of SCP compared with normal values of healthy subjects. This proposed methodology started with extracting distinct features from the force data during the gait cycle. Those extracted features were used to build an appropriate FL system. Our findings illustrated the efficiency of using the fuzzy logic system based on the extracted features in assessing normal and SCP cases effectively and broadly. The proposed methodology shows how important it is to work on extracting features from children's gait data and how efficient input provides for fuzzy logic.

 

 

 References