International Journal of Engineering and Advanced Technology (IJEAT)

ISSN: 2249 – 8958, Volume-6 Issue-4, April 2017

Foot Classification and Influence of Pattern

Recognition
Saleh S. AlTayyar, Ahmed E. Negm
 – Kinematics;
Abstract: The article presents the application of neural network – Kinetics;
and decision tree techniques to investigating barometric data got – Anthropometrics;
with instruments measuring the weight of the human plantar onto
– Electromyography’s;
contact surface while strolling. The examination was completed
on a gathering of plantar foot photo taken while the subject – Others.
remained on the reflected photograph box. We gather 35 The estimation of some of those parameters is important to
understanding, 30 of them are male and 5 female with various play out the quantitative human step appraisal. The appraisal
ages. Numerical qualities for foot examination for every patient of human stride is an imperative errand to for example, assess
foot part get measuring 12 property. Some foot plant pathologies, the level of ailment and measuring the impacts of recovery
similar to buckle and level foot, are ordinarily identified by a
human master by method for impression pictures. All things process or surgical mediation. The achievement of adopting
considered, the absence of prepared individual to finish such huge the traditional strategy to human step by clinician is
first screening discovery endeavors blocks the routinely analytic of emphatically restricted by his capacity. Great clinician ought
the previously mentioned pathologies. In this work an imaginative to be acquainted with the specialized and therapeutic side of
programmed framework for foot plant pathologies in view of the examination. He ought to likewise have the capacity to
neural systems (NN) and Decision Tree (DT) are introduced. The
handle substantial arrangements of information and to make
outcomes accomplished with this framework confirm the
attainability of setting up programmed conclusion frameworks in legitimate aptitude in light of his insight and experience.
light of the impression and example acknowledgment. The order These days, the techniques for programmed human step
settled on by the resultant choice tree was right for all the more examination are exceptionally prevalent, in light of the fact
than 94% steps. This permits to point the parameters which are that they break the impediments of manual assessment of the
the best discriminators between the explored sorts of human walk. information concerning walk [3, 4, 5, 6, 7, 8, 9]. The
Index Terms: Foot Deformities, Photography, Pattern manmade brainpower techniques for programmed step
Recognition, Neural Network, Decision Trees. examination are as per the following: neural systems [9],
fluffy rationales [10] and others [5, 11].
I. INTRODUCTION
Numerous potential and proven benefits of pattern
recognition to clinical biomechanics data analysis and
Data mining is the arrangement of techniques which
prosthesis control mislead most often to their application by
empowers to deal with an immense and multidimensional neglecting some key issues. A key issue is related to the pre-
arrangement of measured Data [1]. Data digging and post-processing of data. A combination of experience and
accommodates quick and proficient dissecting of the trial and error will be necessary for appropriate pre- and
information and finding new, infrequently unforeseen, post-processing data. Pre-processing covers the selection of
associations between different parameters [2]. Biomedical input data, techniques like scaling, normalization, Fourier or
building is an exceptionally specific and essential field of wavelet transformation, rectification and averaging [12]
information mining application. Above all else, biomedical which all will significantly affect the performance. Especially
designing aides in enhancing the nature of human life. under consideration of an adequate relationship of the number
Second, biomedical information have uncommon elements, of input variables and number of adjustable weights with an
for example, high between and intrasubject changeability, ANN these issues are of specific interest. Although formal
high nonlinear reliance between a few parameters, feature selection methods from statistical pattern recognition
multidimensionality and so on, which cause troubles in are offered [13], they have not been widely applied in clinical
dissecting instances of specific subjects with routine biomechanical data analysis. A further and closely related
strategies [3]. One of the imperative issues in biomedical issue to the first one is the issue in developing ANNs
building is a programmed instrumented human step concerning generalization.
investigation. Walk is a fundamental human movement. It This is identified with the nature of the forecasts for cases
empowers us to move our body. Stride is additionally an that are not in the preparation set. Like in other adaptable
exceptionally complex human movement. It is depicted by non-straight estimation techniques, over-fitting and under
tremendous number of parameters which include: fitting is a basic issue to manage. Under-fitting can come
about because of a not adequately complex ANN that neglects
Revised Version Manuscript Received on April 26, 2017. to recognize the flag in a muddled information set.
Saleh S. AlTayyar, Biomedical Technology Department, College of Over-fitting may come about because of an ANN that is
applied medical science/ King Saud University, Riyadh, KSA,
[email protected]
excessively mind boggling and drives,
Ahmed E. Negm, Consultant, Barq Consulting Engineer, Healthcare
Technology Administration, King Fahad Medical City. Riyadh, KSA,
[email protected]

Published By:
Blue Eyes Intelligence Engineering
218 & Sciences Publication Pvt. Ltd.
 Foot Classification and Influence of Pattern Recognition

Making it impossible to fit the commotion, not only the flag. decides the position for the two separations, sizes them,
[14] Consider over-fitting particularly hazardous with a hefty computes the proportion and groups the foot. Indeed, even
portion of the normal sorts of neural systems since it intense the criteria for arranging impressions appears to be
effectively can prompt forecasts that are past the scope of the exceptionally straightforward, the utilization of a classifier in
preparation information. [15]Recommends the utilization of light of pattern recognition offers the accompanying points of
expansive measure of preparing information with a specific interest contrasted and more conventional methodologies: (1)
end goal to stay away from over-fitting. Over-fitting is by all it is not easy to build up a calculation to decide with exactness
accounts improbable, if there are no less than 30 times the the right position to quantify the separations, and (2) it can be
same number of preparing cases as there are weights in the prepared to perceive different pathologies or to enhance their
system. execution as more cases are accessible.
A standout amongst the most encouraging procedures of We will probably think about between example
information mining is choice tree. Choice trees empower to acknowledgment methods as Neural Network and Decision
separate the information covered up in the information and Tree ordering a gigantic information sent out from screening
showing it in an exceptionally clear manner. They give for the recognition of pathologies as level foot and give in
extremely straightforward conditions in the tree hubs and foot. These information were arranged by a specialist.
prompt conclusion (class) on the least level of the tree. It is This paper portrays contrast of a programmed
critical that the outcomes are simple for translation and could characterizing information with analyze foot plant
be utilized by the staff with neither scientific nor building pathologies. We define analyze of foot plant pathologies as an
foundation. This settles on choice trees helpful device for example acknowledgment issue. This paper is sorted out as
clinical applications. A standout amongst the most intriguing takes after. Area 2 present the foot plant pathologies. Segment
properties of choice trees is no from the earlier presumptions. 3 depicts material and methods used. Segment 4 introduces
Also, to work legitimately, the choice trees don't require as result and discussion of use classifiers. Area 5 demonstrates
much information as neural systems. It is particularly vital in conclusion of using classifier.
biomedical applications, where the quantity of subjects is
exceptionally restricted. Obviously, a greater arrangement of II. MATERIAL AND METHOD
information give more exact and more dependable results.
Choice trees have as of now been effectively utilized as a part A. Material
of the human walk examination in clinical applications [16, Electronic systems for recording and evaluating pressure
17]. Be that as it may, in late papers the fundamental fate of distribution under the foot in static and dynamic conditions.
choice trees is an order assignment. The platforms provide accurate, reliable information for the
At the point when the foot is planted, not all the sole is in analysis of foot function and diagnosis of foot pathologies.
contact with the ground, the impression is the surface of the Foot deformities and malfunction can be detected during
foot plant in contact with the ground. The give in foot and the analysis of the barefoot pressure data. The Technical
level foot are pathologies exhibited. On the off chance that specification of emed®-xl platform, dimension (mm) is 1,529
these foot deformities are not distinguished and treated on x 504 x 21, Sensor area (mm) is 1,440 x 440, number of
time, they turn out to be most noticeably awful amid sensors 25,344, resolution (sensor/cm2) is 4 and with 100 Hz
adulthood creating a few aggravations, agony and stance frequency. Foot type was determined by obtaining arch index
related disarranges [20]. values. We collect 35 patient, 30 of them are male and 5
Shape and zones of the impression are appeared in figure 1. female with different ages. numerical values for foot as Total
Zones 1, 2 and 3 relate to locales in contact with the surface Object, Hind foot, Mid foot, Fore foot, MH1, MH2, MH3,
when the foot is planted, these are called foremost heel, back MH4, MH5, Toes, Big toe, Second toe, Toes 345 and Toes
heel and isthmus separately. Zone 4 does not form part of the 2345 for each patient foot part get measuring the following 12
surface in contact and is called impression vault [21]. A basic attribute; [Force Time Integral, Max Force (N), Peak Pressure
strategy to acquire impressions is specifically venturing the (Kpa), Contact Area (cm²), Contact Time] mean and standard
inked foot onto a paper on the floor. Subsequent to acquiring deviation, Arch Index and grade. We use WEKA 3.6.9
the impressions, a specialist breaks down them and surveys in platform for classification step. Waikato Environment for
the event that they show pathologies. More often than not, in knowledge Analysis (Weka) is a popular suite of machine
the conclusion of these pathologies an instrument known as learning software written in java, developed at the University
podoscope is utilized to catch the impressions. A of Waikato, New Zealand.
straightforward computerized rendition of the podoscope in B. METHOD
light of a scanner has been proposed in [22]. Another essential
A ''highest quality level'' technique for deciding foot sort
instrument to get impressions is the pedobarograph. Present
has yet to be set up, and clinical perception remains the
day variations of the pedobarograph are proposed in [23, 24].
strategy frequently depended upon. [29] Measures showed
A technique to fragment impressions in optical shading
poor to great dependability in supporting clinical judgments
pictures of the sole by utilizing neural systems is proposed as
of foot sort, [30] yet the unwavering quality of these
a part of [25].
estimations in grouping foot sort has not been researched.
As of now, a specialist characterizes if a patient has an
Curve record values have been utilized to decide foot sort.
ordinary, surrender or level foot by a manual exam called
photopodogram. A photopodogram is a compound
photograph of the foot part supporting the heap. The master
Published By:
Blue Eyes Intelligence Engineering
219 & Sciences Publication Pvt. Ltd.
 International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249 – 8958, Volume-6 Issue-4, April 2017

[31] Arch record values ascertained by impression motivation behind this study was to decide interrater
examination have been acquired from compel plate, [31] assention for foot-sort arrangement. Strategy: Arch list was
carbon impression paper, [34] and photos brought with a ascertained from impression photos acquired through
reflected glass box. [32] [33]. [34] Arch list ascertained [35] reflected photograph box. Grouping as high-angled, typical,
solid for figuring curve list values, [32] and ok for both people or low-curved foot sort depended on curve list values.
of typical weight and those with stoutness. [36] Arch list Unwavering quality of the curve file was resolved with
additionally shifts with age, falling into the typical grown-up intra-class relationships; concurrence by walking sort
range by age [32] [37] Whether among corpulent or non-hefty arrangement was resolved utilizing quadratic weighted kappa
individuals, [36] school-matured kids or grown-ups, [37] or (kw).
men or ladies from various nations, [38] curve record values
fall into the distinctive scopes of curve list values proposed III. RESULTS AND DISCUSSION
and utilized as a potential technique to arrange high-angled, We will use data of foot measurements to compare which
ordinary, and low-angled foot sorts. [38] The unwavering classifier will be more suitable for automatic classification of
quality of grouping foot sort by this technique has not been the foot. We will use two type of classifiers as Decision Tree
examined, be that as it may. Foot plant pathologies and and Neural Network. Compare between them to detect which
pattern recognition diagnosis; It is conceivable to characterize of them are more accurate in the ability of differentiation of
a foot by its impression shape and measurements as an: foot categories from our dataset. We use 10-fold cross
ordinary, level or give in foot. Figure 1b demonstrates a validation technique.
picture of a level foot, figure 1c demonstrates a picture of a  Neural Network using weka classifiers functions
typical foot, and figure 1d demonstrates a picture of a give in Multilayer Perceptron as shown in figure 3.
foot.

Figure 1 Images of The Foot

Before characterization, the impression is confined from
whatever remains of the segments of the sole picture by
utilizing the technique proposed as a part of [1]. Figures 2a,
2b and 3c demonstrated the division of a level, a typical foot,
and a give in foot individually.
Figure 3 Neural Network for 12 Attribute Classified to
Get Three Different Grade of Foot.
 Decision Tree using weka classifiers trees J48
figure 4 present decision tree configuration.

Figure 2 Segmentation of Footprint

Subsequent to playing out the division, the impression is
spoken to by a vector containing the width in pixels of the
divided impression, without toes, by every segment in the
level course. Since each picture has a width vector with
various length, the vectors were standardized to have similar
length. Likewise the estimation of every component was
standardized to an esteem in the scope of 0 to 1.Figures 2a, 2b
and 2c demonstrate the standardized vectors of a level, a
typical and a surrender foot.
The absence of a dependable characterization standard for Figure 4 Decision Tree Example for Classify Data.
foot sort makes making inferences from existing exploration
and clinical choices troublesome, since various foot sorts may
move and react to treatment in an unexpected way. The
Published By:
Blue Eyes Intelligence Engineering
220 & Sciences Publication Pvt. Ltd.
 Foot Classification and Influence of Pattern Recognition

We will apply the sensitivity test in order to determine the

sensitivity of each classifier in order to distinguish between
foot categories. Sensitivity and specificity are widely ‎used
statistics to describe a diagnostic test. In ‎particular, they are
used to quantify how good ‎and reliable a test is.‫‏‬‎‫ ‏‬Sensitivity
evaluates how ‎good the test is at detecting a positive
disease. ‎Specificity estimates how likely patients ‎without
disease can be correctly ruled out.
The ‎ROC curve is a graphic representation of
the ‎relationship between both sensitivity and ‎specificity, and
it helps to decide the optimal ‎model. Accuracy measures how
correct a ‎diagnostic test identifies and excludes a
given ‎condition. The accuracy of a diagnostic test ‎can be
determined from sensitivity and ‎specificity with the presence
of prevalence [51]. ‎ There are several terms that are
Figure 5 ROC curve of Neural Network classifier for Left
commonly ‎used along with the description of
Leg Data
sensitivity, ‎specificity and accuracy.‫‏‬‎‫ ‏‬They are true ‎positive
(TP), true negative (TN), false negative ‎(FN), and false The model has different pattern for each categories we have
positive (FP). ‎ Low categories Sensitivity = 90.87%, Specificity = 100%,
Accuracy= 94.27%, Positive predictive value = 100%,
Sensitivity = TP/ (TP + FN) (1) ‎ Negative predicted value = 86.66% and false rate Discover =
0%; Normal categories Sensitivity = 94.47%, Specificity =
Specificity = TN/ (TN + FP) (2)‎ 91.41%, Accuracy= 92.43%, Positive predictive value =
)PT+PT+PT+PT(‎/)PT+PT(‎= uucarucc‎ )3(‎ 84.61%, Negative predicted value = 97.06% and false rate
Discover = 15.388%.; and High categories Sensitivity =
)PT+PT(‎/PT‎=Teuepeic‎taceeupce‎iricc‎ )4(‎
100%, Specificity = 98.08%, Accuracy= 98.159%, Positive
)PT+PT(‎/PT‎=Tcurpeic‎taceeupce‎iricc‎ )5(‎ predictive value = 84.615%, Negative predicted value =
)PT+PT(‎/PT‎=c‎teuueica‎srpc‎Priu )6(‎‎ 97.06% and false rate Discover = 32.14%
b. Decision Tree function J48 for patients left leg
A confusion matrix contains information ‎about actual and data, building model time taken is 0.02 seconds,
predicted classifications ‎done by a classification system. confusion matrix in table 2 followed by ROC
Performance ‎of such systems is commonly evaluated Curve in figure 6 clarify model output numbers.
using ‎the data in the matrix. In the field of Correctly Classified Instances are 461 with
artificial ‎intelligence, a confusion matrix is a specific ‎table 94.274 % and Incorrectly Classified Instances
layout that allows visualization of the ‎performance of an are 28 with 5.726 %
algorithm, typically a ‎supervised learning one. ‎Each column
of the matrix represents the ‎instances in a predicted class, Table 2 Confusion Matrix represent left leg classification
while each row ‎represents the instances in an actual class [51].‎ with Decision Tree Model for 12 attribute.
 Classification for 12 attribute. Classified as LOW NORMAL HIGH
a. Neural Network function Multilayer Perceptron LOW 279 0 0
for patients left leg data, building model time
NORMAL 28 154 0
taken is 80.55 sec, confusion matrix in table 1
followed by ROC Curve in figure 5 clarify model HIGH 0 9 19
output numbers. Correctly Classified Instances is
452 with 92.4335 % and Incorrectly Classified
Instances are 37 with 7.5665 %.
Table 1 Confusion Matrix represent left leg classification
with Neural Network Model for 12 attribute.

Classified as LOW NORMAL HIGH

LOW 279 0 0

NORMAL 28 154 0

HIGH 0 9 19
Figure 6 ROC curve of Neural Network classifier for Left Leg
Data

Published By:
Blue Eyes Intelligence Engineering
221 & Sciences Publication Pvt. Ltd.
 International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249 – 8958, Volume-6 Issue-4, April 2017

The model has different pattern for each categories we have 475 with 96.9388 % and Incorrectly Classified
Low categories Sensitivity = 90.87%, Specificity = 100%, Instances are 15 with 3.0612 %.
Accuracy= 94.27%, Positive predictive value = 100%, Table 4 Confusion Matrix represent Right leg
Negative predicted value = 86.66% and false rate Discover = classification with Neural Network Model for 12
0%; Normal categories Sensitivity = 94.47%, Specificity = attribute.
91.41%, Accuracy= 92.43%, Positive predictive value =
84.61%, Negative predicted value = 97.06% and false rate classified as LOW NORMAL HIGH
Discover = 15.388%.; and High categories Sensitivity = LOW 308 0 0
100%, Specificity = 98.08%, Accuracy= 98.159%, Positive
predictive value = 84.615%, Negative predicted value = NORMAL 1 167 0
97.06% and false rate Discover = 32.14% HIGH 0 14 0
c. Decision Tree function J48 for patients left leg
data, building model time taken is 0.02 seconds,
confusion matrix in table 2 followed by ROC
Curve in figure 6 clarify model output numbers.
Correctly Classified Instances are 461 with
94.274 % and Incorrectly Classified Instances
are 28 with 5.726 %
Table 3 Confusion Matrix represent left leg classification
with Decision Tree Model for 12 attribute.
classified as LOW NORMAL HIGH
LOW 279 0 0
NORMAL 28 154 0
HIGH 0 0 28

Figure 8 ROC curve of Neural Network classifier for

Right Leg Data
The model has different pattern for each categories we have.
Low information Sensitivity = 99.67%, Specificity = 100%,
Accuracy= 99.79%, Positive predictive value = 100%,
Negative predicted value = 99% and false rate Discover = 0%,
Normal with Sensitivity = 92.265%, Specificity = 99.67%,
Accuracy= 96.93%, Positive predictive value = 99.4%,
Negative predicted value = 9565% and false rate Discover =
0.59%. And High data Sensitivity = 0%, Specificity =
97.14%, Accuracy= 100%, Positive predictive value = 0%,
Negative predicted value = 100% and false rate Discover =
100%.
Figure 7 ROC curve of Decision Tree classifier for Left Leg
e. Decision Tree function J48 for patients Right leg
Data data, building model time taken is 0.03 seconds,
The model has different pattern for each categories we have confusion matrix in table 4 followed by ROC
Low with Sensitivity = 90.87%, Specificity = 100%, Curve in figure 8 clarify model output numbers.
Accuracy= 94.27%, Positive predictive value = 100%, Correctly Classified Instances are with 100 %
Negative predicted value = 86.66% and false rate Discover = and Incorrectly Classified Instances are 0 %
0%; Normal with Sensitivity = 100%, Specificity = 91.64%, Table 5 Confusion Matrix represent Right leg
Accuracy= 96.93%, Positive predictive value = 94.27%, classification with Decision Tree Model for 12 attribute
Negative predicted value = 84.615% and false rate Discover =
15.38%. And High categories Sensitivity = 100%, Specificity classified as LOW NORMAL HIGH
= 100%, Accuracy= 100%, Positive predictive value = 100%,
Negative predicted value = 100% and false rate Discover = LOW 308 0 0
0%. NORMAL 0 168 0
d. Neural Network function Multilayer Perceptron
for patients Right leg data, building model time HIGH 0 0 14
taken is 0.63 sec, confusion matrix in table 3
followed by ROC Curve in figure 7 clarify model
output numbers. Correctly Classified Instances is

Published By:
Blue Eyes Intelligence Engineering
222 & Sciences Publication Pvt. Ltd.
 Foot Classification and Influence of Pattern Recognition

with 64.8262 % and Incorrectly Classified

Instances are 172 with 35.1738 %
Table 7 Confusion Matrix represent left leg classification
with Decision Tree Model for 11 attribute.
classified as LOW NORMAL HIGH
LOW 218 57 4
NORMAL 83 96 3
HIGH 15 10 3

h. Neural Network function Multilayer Perceptron

for patients Right leg data, building model time
taken is 0.61 sec, confusion matrix in table 7
clarify model output numbers. Correctly
Figure 9 ROC curve of Decision Tree classifier for Right Classified Instances is 336 with 68.5714 % and
Leg Data Incorrectly Classified Instances are 154 with
31.4286 %
The model has different pattern for each categories we have.
Low information data Sensitivity = 100%, Specificity = Table 8 Confusion Matrix represent Right leg
100%, Accuracy= 100%, Positive predictive value = 100%, classification with Neural Network Model for 11 attribute
Negative predicted value = 100% and false rate Discover =
Classified as LOW NORMAL HIGH
0%; Normal Sensitivity = 100%, Specificity = 100%,
Accuracy= 100%, Positive predictive value = 100%, LOW 294 14 0
Negative predicted value = 100% and false rate Discover = NORMAL 126 42 0
0%. And High Sensitivity = 100%, Specificity = 100%, HIGH 12 2 0
Accuracy= 100%, Positive predictive value = 100%,
Negative predicted value = 100% and false rate Discover =
i. Decision Tree function J48 for patients Right leg
0%.
data, building model time taken is 0.03 seconds,
Applying both model of pattern recognition Neural
confusion matrix in table 8 clarify model output
Network and Decision Tree to our data which consist from 12
numbers. Correctly Classified Instances are 363
attribute shows that decision tree gives an impressive result in
with 74.0816 % and Incorrectly Classified
identifying each category of the foot classification for both
Instances are 127 with 25.9184 %
legs information. These result encourage us to find out the
importance of arch index attribute and test if it has influence Table 9 Confusion Matrix represent Right leg
to the degree of classification or not. So we will exclude this classification with Decision Tree Model for 11 attribute.
attribute from our data and repeat the previous test. classified as LOW NORMAL HIGH
 Classification for 11 attribute.
LOW 279 28 1
We Repeat the previous test but without identification of Arch
Index information and observe deviation in the classifier NORMAL 82 81 5
result. HIGH 8 3 3
f. Neural Network function Multilayer Perceptron
for patients left leg data, building model time The results shows that Decision tree classifier still is the best
taken is 0.45 sec, confusion matrix in table 5 choice in classifying the foot categories but there are a big
clarify model output numbers. Correctly drop for accuracy of each classifier by 25% to 30% in both
Classified Instances is 261with 53.3742 % and Decision Tree even in Neural Network. This drop give arch
Incorrectly Classified Instances are 228 with index importance indication, this also prove that arch index is
46.6258 % critical attribute and its affect the classification.
Table 6 Confusion Matrix represent left leg classification
with Neural network Model for 11 attribute. IV. CONCLUSION
Using arch index values obtained from digital photography
classified as LOW NORMAL HIGH
using a photo-box to classify foot type as high arched, normal,
LOW 223 54 2
or low-arched, based on defined ranges, proved to be a
NORMAL 144 37 1 reliable method that led to excellent agreement between
HIGH 19 8 1 raters. Classifying feet to subdivide sample populations may
help differentiate mobility, gait, or treatment effects among
g. Decision Tree function J48 for patients left leg foot types in future research and clinical practice. What is
data, building model time taken is 0.03 seconds, already known on this topic a variety of foot measures,
confusion matrix in table 6 clarify model output including arch index,
numbers. Correctly Classified Instances are 317

Published By:
Blue Eyes Intelligence Engineering
223 & Sciences Publication Pvt. Ltd.
 International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249 – 8958, Volume-6 Issue-4, April 2017

Have been shown to be reliable and related to other static Transactions on Neural Systems and Rehabilitation Engineering, vol.
10, no. 1, pp. 30-37, 2002.
measures. Yet static measures have not correlated strongly 15. S. Haykin, "Neural networks: MacMillan College Publ," Co., New
with motion in gait, in part because some study samples have York, 1994.
included mixed foot types. 16. S. Armand, E. Watelain, E. Roux, M. Mercier and F.-X. Lepoutre,
What this study adds Classification of foot type based on "Linking clinical measurements and kinematic gait patterns of
toe-walking using fuzzy decision trees," Gait & posture, vol. 25, no. 3,
arch index values rather than clinical judgments demonstrated pp. 475-484, 2007.
excellent agreement. Standardized foot-type classifications 17. R. Mikut, J. Jäkel and L. Gröll, "Interpretability issues in data-based
allow sample subdivisions that may further our understanding learning of fuzzy systems," Fuzzy Sets and Systems, vol. 150, no. 2, pp.
179-197, 2005.
of different mobility characteristics or treatment effects 18. W. Pretkiewicz-Abacjew and W. Erdmann, "Kinematics of walking of
among high-arched, normal, and low-arched foot types in 6 year old children," Journal of Human Kinetics, vol. 3, pp. 115-130,
future research and clinical practice. 2000.
19. Valenti, "Orthotic Treatment of Walk Alterations," Panamerican
The approach to automatic human gait analysis Medicine,(in spanish), 1979.
demonstrates that decision trees are a powerful technique 20. Morsy and A. Hosny, "A new system for the assessment of diabetic
which could be successfully applied in biomechanics. A foot planter pressure," Engineering in Medicine and Biology Society,
2004. IEMBS'04. 26th Annual International Conference of the IEEE,
decision tree could manage vector of a many numbers of real
vol. 1, no. IEEE, pp. 1376-1379, 2004.
parameters as an input and point the values which are main 21. K. Nakajima, Y. Mizukami, K. Tanaka and T. Tamura,
discriminators. The advantage of employing decision trees is "Footprint-based personal recognition," IEEE Transactions on
the easiness of interpretation so it could be applied Biomedical Engineering, vol. 47, no. 11, pp. 1534-1537, 2000.
22. S. R. Shah and K. Patil, "Processing of foot pressure images and
successfully to clinics. Moreover, decision trees could display of an advanced clinical parameter PR in diabetic neuropathy,"
improve the understanding of human gait phenomena and in Conference Proceedings. 2nd International IEEE EMBS Conference
could lead to the selection of more suitable methods for on Neural Engineering, 2005., 2005.
23. M. Mora and D. Sbarbaro, "A robust footprint detection using color
human gait improvement. It is really important that the same images and neural networks," in Iberoamerican Congress on Pattern
procedure could be used in any data describing a human Recognition, 2005.
activity independent of the clinical problem. 24. N. J. I. Q. R. Chuckpaiwong B, "Correlation between static foot type
measurements and clinical assessments.," Foot Ankle Int., vol. 30, no.
3, pp. 205-212, 2009.
REFERENCES 25. M. Razeghi M, "Foot type classification: a critical review of current
1. M. C. J. D. S. a. L. P. Marco Mora, "Automatic Diagnosis of Foot Plant methods.," Gait Posture., vol. 15, no. 3, p. 282–91, 2002.
Pathologies: A Neural Networks Approach," in International 26. M. A. E. J. Leung AK, "Biomedical gait evaluation of the immediate
Symposium on Neural Networks, 2007. effect of orthotic treatment for flexible flat foot.," Prosthet Orthot Int. ,
2. M. Derlatka and M. Ihnatouski, "Decision tree approach to rules vol. 22, no. 1, pp. 25-34, 1998.
extraction for human gait analysis," in International Conference on 27. M. S. Menz HB, "Validity of 3 clinical techniques for the measurement
Artificial Intelligence and Soft Computing, 2010. of static foot posture in older people.," J Orthop Sports Phys Ther. , vol.
3. T. Chau, "A review of analytical techniques for gait data. Part 1: fuzzy, 35, no. 8, p. 479–86, 2005.
statistical and fractal methods," Gait & posture, vol. 13, no. 1, pp. 28. E. J. T. C. e. a. Villarroya MA, "Foot structure in overweight and obese
49-66, 2001. children.," Int J Pediatr Obes., vol. 3, no. 1, pp. 39-45, 2008.
4. R. Begg and J. Kamruzzaman, "A machine learning approach for 29. H. W. N. J. e. a. Mall NA, "The reliability and reproducibility of foot
automated recognition of movement patterns using basic, kinetic and type measurements using a mirrored foot photo box and digital
kinematic gait data," Journal of biomechanics, vol. 38, no. 3, pp. photography compared to caliper measurements.," J Biomech., vol. 40,
401-408, 2005. no. 5, pp. 1171-6, 2007.
5. M. Derlatka, "Application of Kernel principal component analysis in 30. M. McPoil TG, "Use of plantar contact area to predict medial
human gait," Journal of Vibroengineering , vol. 7, no. 3, pp. 27-30, longitudinal arch height during walking.," J Am Podiatr Med Assoc,
2005. vol. 96, no. 6, pp. 489-94, 2006.
6. M. Derlatka and J. Pauk, "Data Mining in Analysis of Biomechanical 31. S. J. M. B. Mickle KJ, "The feet of overweight and obese young
Signals," in Solid State Phenomena, 2009. children: are they flat or fat?," Obesity (Silver Spring)., vol. 14, no. 11,
7. S. Ghoussayni, C. Stevens, S. Durham and D. Ewins, "Assessment and pp. 1949-53, 2006.
validation of a simple automated method for the detection of gait 32. G. J. R. D. Bosch K, "Development of healthy children’s
events and intervals," Gait & Posture, vol. 20, no. 3, pp. 266-272, feet–nine-year results of a longitudinal investigation of plantar loading
2004. patterns.," Gait Posture., vol. 32, no. 4, p. 564–71, 2010.
8. S. Wolf, T. Loose, M. Schablowski, L. Döderlein, R. Rupp, H. J. 33. M. B. S. M. Igbigbi PS, "Arch index as a predictor of pes planus: a
Gerner, G. Bretthauer and R. Mikut, "Automated feature assessment in comparative study of indigenous Kenyans and Tanzanians.," J Am
instrumented gait analysis," Gait & posture, vol. 23, no. 3, pp. Podiatr Med Assoc. , vol. 95, no. 3, p. 273–6, 2005.
331-338, 2006. 34. G. R. W. C. e. a. Xiong S, "Foot arch characterization: a review, a new
9. J. Wu, J. Wang and L. Liu, "Kernel-based method for automated metric, and a comparison.," J Am Podiatr Med Assoc., vol. 100, no. 1,
walking patterns recognition using kinematics data," in International pp. 14-24, 2010.
Conference on Natural Computation, 2006. 35. G. R. A. J. e. a. Faria A, "The relationship of body mass index, age and
10. Yardimci, "Fuzzy logic based gait classification for hemiplegic triceps-surae musculotendinous stiffness with the foot arch structure of
patients," in International Symposium on Intelligent Data Analysis, postmenopausal women.," Clin Biomech (Bristol, Avon). , vol. 25, no.
2007. 6, p. 588–93., 2010;.
11. S. J. Olney, M. P. Griffin and I. D. McBride, "Multivariate 36. W. C. Chu, S. H. Lee, W. Chu, T.-J. Wang and M.-C. Lee, "The use of
examination of data from gait analysis of persons with stroke," arch index to characterize arch height: a digital image processing
Physical Therapy, American Physical Therapy Association, vol. 78, approach," IEEE Transactions on Biomedical Engineering, vol. 42, no.
no. 8, pp. 814-828, 1998. 11, pp. 1088-1093, 1995.
12. T. Chau, "A review of analytical techniques for gait data. Part 2: neural 37. J. L. McClelland, D. E. Rumelhart and G. E. Hinton, "The appeal of
network and wavelet methods," Gait & Posture, vol. 13, no. 2, pp. parallel distributed processing," MIT Press, Cambridge MA, pp. 3-44,
102-120, 2001. 1986.
13. H. Liu and H. Motoda, Feature selection for knowledge discovery and 38. F. D. Foresee and M. T. Hagan, "Gauss-Newton approximation to
data mining, vol. 454, Springer Science & Business Media, 2012. Bayesian learning," in Neural Networks, 1997. International
14. L. Wang and T. S. Buchanan, "Prediction of joint moments using a Conference on, 1997.
neural network model of muscle activations from EMG signals," IEEE

Published By:
Blue Eyes Intelligence Engineering
224 & Sciences Publication Pvt. Ltd.
 Foot Classification and Influence of Pattern Recognition

39. J. MacKay, "Bayesian interpolation," Neural computation, vol. 4, no. Ahmed El-Sayed Negm, Lecturer, The Higher
3, pp. 415-447, 1992. Institute of Engineering - Shorouk Academy,
40. J. Pauk and K. Jaworek, "Parametric identification of lower limbs Cairo, Egypt.
during walking of a man," WIT Transactions on Ecology and the Doctorate of philosophy in system and
Environment, vol. 57, 2002. Biomedical Engineering.
41. M. Romei, M. Galli, F. Motta, M. Schwartz and M. Crivellini, "Use of  (System and Biomedical
the normalcy index for the evaluation of gait pathology," Gait & Engineering Department- Faculty of
posture, vol. 19, no. 1, pp. 85-90, 2004. Engineering-Cairo University). In
(Automatic Segmentation & Classification of Acute Leukemia
cells in microscopic images), Medical Imaging. 2017
Biography
Master Degree in System and Biomedical Engineering.
Saleh S. AL-Tayyar, Assistant professor  (System and Biomedical Engineering Department- Faculty of
in the Department of Biomedical Engineering-Cairo University). In (Decision Support System for
Technology, College of Applied Medical Lymphoma Classification), Medical Imaging . 2009
Sciences, at King Saud University Ph.D B.Sc. in System and Biomedical Engineering.
Biomedical Engineering - Mississippi  (The Higher Institute of Engineering - Shorouk Academy). 2000
State University - Mississippi state MS.  Participate in “Automatic Segmentation of Acute Leukemia Cells
USA. International” Journal of Computer Applications (0975 – 8887)
Publications: Volume 133 – No.10, January 2016.
 Saleh S. Altayyar. “The Importance of  Participate in “Decision support system in lymphoma
Plantar Pressure Measurements and classification”, Current Medical Imaging Reviews Journal, Volume.
Appropriate Footwear for Diabetic Patients” Journal of Analytical & 12, No. 4, pp. 1-10, 2016.
Pharmaceutical Research. Volume 3 Issue 3. October18, 2016  Participate in “A Decision Support System for Acute Leukemia
 Saleh S. Altayyar. “Medical Devices and Patient Safety” Journal of Classification Based on Digital Microscopic Images”, Alexandria
Analytical & Pharmaceutical Research. Volume 2 Issue 5. July 5, Engineering journal AEJ, 2017 in publication.
2016  Participate in “Classifiers comparison in Acute Myeloid Leukemia
 Saleh S. Altayyar. “The Impact of Custom Made Insoles on the Plantar Classification of Digital Microscopic Images”, Journal of
Pressure of Diabetic Foot. Majmaah Journal of Health Sciences, Vol Engineering and Applied Science, 2017 in publication.
4, No.(1), May 2016-Sha’ban 1437.  Participated in 3D Medical Printing Special Session in BUILDING
 S.Tayyar, P.S.Weinhold, R.A.Butler, J.C.Woodara, L.D Zaadiackas, HEALTHCARE MIDDLE EAST Exhibition &Conference at 30th
K.R.ST.JOHN, J.M.Bledsoe, J.A.Gilbert. “Computer Simulation of May 2016.
Trabecular Bone Remodeling Using A Simplified Structural Model”.  Participated in Medical Equipment Planning in BUILDING
Bone Vol. 25, No.6 Dec.1999 733-739. HEALTHCARE MIDDLE EAST Exhibition &Conference at 30th,
Invited lectures & conference proceedings: 31st May 2016.
 Saleh S. Altayyar. "Impact of Physical Activities (Walking) On the  Participated in Healthcare Management Conference in BUILDING
Plantar Pressure of Diabetic And Non-Diabetic Subjects” HEALTHCARE MIDDLE EAST Exhibition &Conference at 1st June
Accepted for presentation at ESM 2016. Lisbon in July 27-30. 2016.
 Saleh S. Altayyar, Medical Devices & Patient Safety. 24 th World  Participated in Lab Management & Mega Project Investment Session
Congress on Medical Physics & Biomedical Engineering (IUPESM in BUILDING HEALTHCARE MIDDLE EAST Exhibition
2015). 7-12 June 2015. Toronto Canada. &Conference at 1st June 2016.
 Altayyar, S., Alromayan, S. “Saudi National Medical Devices Implants  Participated in PROMOTING THE QUALITY CULTURE: NEAR
Registry - Aspirations and Challenges” 3rd International Congress MISS MONITORING AS A HARBINGER OF BIGGER ISSUES
of Arthoplasty Registries (31 May - 2 June 2014. Boston, MA. USA). Session in BUILDING HEALTHCARE MIDDLE EAST Exhibition
 Al Tayyar, S ., Thabit, A. “Post Market Surveillance in Saudi Arabia”. &Conference at 1st June 2016.
Second Global Forum on Medical Devices (22 - 24 November  Attend ICME Workshop “The Importance of Design in Creating a
2013Geneva, Switzerland). Functional and Efficient Hospital ” –Building Healthcare 2016,
 Saleh S. Al Tayyar “Raising your BPM Maturity Level ; Saudi Dubai, 30th May 2016
Arabian Case Study” Building Business Capability Conference. (11  Attend ICME Workshop “The role of project management and site
– 15 November 2013. Las Vegas, N. , USA). supervision in delivering a project on time and budget” –Building
Achievements: Healthcare 2016, Dubai, 31st May 2016
 Vice Executive President for Medical Devices Sector at the Saudi Food &  Attend in CIBEC the 6th Cairo International Biomedical Engineering
Drug Authority (Sep 2008 – Sep 2014). Conference Dec 20-22, 2012.
 Established a state of the art medical devices regulatory system in Saudi  Attend the 6th Cairo International Biomedical Engineering Conference
Arabia that achieved the respect and recognition of the international workshop entitled CLINICAL ENGINEER on December 2012.
regulatory arena.  Attend workshop and conference about OLYMPUS technologies and
 Chair, Asian Harmonization Working Party (AHWP), an organization of products of telepathology in August 2009.
more than 23 member economies working toward harmonization of  Attend the 2nd Oncology Conference on “Lymphoma Updating,
medical devices in Asia, Africa, Middle East, and Latin America for Diagnosis & Management” at International Medical Center from
three years, 24th-25th December 2008.
 Chair, Riyadh Biomedical Engineering Club.  Attend First construction of telepathological experiment at
 Chair, Medical & Clinical Engineering Chapter at the Saudi Council of International Medical Center July 2007.
Engineers.  Attend in the 3rd Cairo International Biomedical Engineering
Board Memberships: Conference and workshops 2006:
 Board of Saudi Standard, Metrology and Quality Organization (SASO).  Visualization. Medical Equipment Specifications. Attend workshop on
 The Scientific Board for Applied Medical Sciences (SBAMS), Saudi Electron-Microscope at Central Medical Labs of Military of Defense
Commission for Health Specialties. 2006.
 The Board of International Medical Devices Regulators Forum  Attend the Electronic Health Records training course on the occasion
(IMDRF). of the ITI 4th International Conference 2006.
 Member of the advisory board of many colleges and universities in  Attend workshop on Immunohistochemistry and Morphometry at
Saudi Arabia. faculty of medicine-Ain Shams University from 3rd to 5th April 2005.
Refereed and co-authored innovation in healthcare, diabetic foot, and  Attend and organize the 1st International Biomedical Engineering
clinical engineering management. Conference & workshops in Safety and Hazards in Hospital and
Clinical Engineering Role in February 2003
Nationality: Egyptian , Born in: Egypt, Address: KSA, Riyadh, Mobile:
(+966) 508502054, (+2) 01220655595
E-mail:[email protected]

Published By:
Blue Eyes Intelligence Engineering
225 & Sciences Publication Pvt. Ltd.