Handwritten Processing for Pre Diagnosis of Alzheimer Disease

Donato Impedovo Ieee member

Giuseppe Pirlo

Donato Barbuzzi

Alessandro Balestrucci

and Sebastiano Impedovo

Department of Electrical and Electronic of the Bari Polytechnic, Bari, Italy

Department of Computer Science of "Aldo Moro", Bari University, Via Orabona 4, 70125 Bari, Italy

Keywords: Neuromuscular Transfer Function, Handwritten Data Analysis, Alzheimer Disease.

Abstract: Based on neuromuscular transfer function of the handwriting system, in this paper a non invasive pre

diagnosis system for Alzheimer disease alert is proposed. It is well known in fact, that writing originates

from spike trains produced within the Central Nervous System (CNS) and more specifically, inside the 4-th

and the 6-th regions of the Bradman's map and then transmitted through the first and second order axons to

the spinal cord to control the muscles involved in the handwriting as the arm, the forearm, the hand and the

pen or pencil utilized for the writing. More specifically, in this work is proposed a new method, not

invasive, for early diagnosis of degenerative disability, it can be also useful for monitoring activities related

to the progression of neuromuscular disease in order to evaluate the changing related also to the efficiency

of the therapies used. Benefit can be obtained not only for the medical field but also for the

pharmaceutical developments. Specifically in the paper, the results of some experiments have been focused

by considering a certain number of persons some of which affect by Alzheimer disease.

1 INTRODUCTION

Biometrics, from the Greek: Bios that means "life"

and Metros that means "measure", is the discipline

devoted to the study of variables related to the

physicality of individuals in order to measure their

value (Boldrini, M., 1934) and to use them to the

creation of applications that are used in several

fields i.e. finger prints generally used for social end

military goals to certificate the identity of a person.

See fig.1 (Medugno, V., Valentino, S., Acampora,

G., 1999).

Bioengineering is recognized worldwide as an

emerging discipline aimed at generating a better

understanding of biological phenomena and produce

technologies for health with benefit of the society.

The Bioengineering operates in several areas:

 Technological,

 Industrial,

 Scientific,

 Clinical,

 Hospital.

The goal which it arises is twofold:

 Improvement of knowledge about the functioning

of biological systems;

 Development of new methodologies and

diagnostic tools for treatment and rehabilitation

(Biondi, E., Cobelli, E., 2001).

Figure 1: Fingerprint, signature, personal handwritten

strokes.

In this paper specifically the handwriting is

considered as a biometric entity and it is investigated

in depth.

After this short introduction on the biometry, the

paper presents: in sec. 2 the human neuromuscular

193

Impedovo D., Pirlo G., Barbuzzi D., Balestrucci A. and Impedovo S..

Handwritten Processing for Pre Diagnosis of Alzheimer Disease.

DOI: 10.5220/0004900701930199

In Proceedings of the International Conference on Biomedical Electronics and Devices (BIODEVICES-2014), pages 193-199

ISBN: 978-989-758-013-0

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

system involved in the hand writing process; in sec.

3 the problem of the stability in handwriting

emphasizing the most stable part of a writing and

showing the influence of a disease on these parts.

Finally in sec. 4 some results are reported by

processing handwritten sample produced by some

Alzheimer illness affected patients by stressing the

relationship between stability regions and disease

insurgence.

2 ANATOMICAL AND

NEURO-PHYSIOLOGICAL

ASPECTS OF HANDWRITING

The handwriting as neuromuscular activity is

described by starting from the neurophysiologic

point of view and arriving to the mechanics of the

neuromuscular system that allows the handwriting.

During the years of his live, a writer acquires at

subconscious level some graph logic features that

are synthesized in graphic strokes made increasingly

clear, safe, fast and unique, here called personal

strokes, that remain unchanged over time.

By looking at handwriting, it must be said that

humans have different types of movements:

"automatic" and "volunteers"; writing movements

are a combination of this two type. The ability to

move is made possible by the neurons in the cerebral

cortex of the motor area (Brodmann's area 4, also

called the primary motor area) at the origin of the

pyramidal system, that provide at the voluntary

movements of the muscles and the planning of the

motor gesture. The motor area controls all voluntary

movements, between these areas relative to the fine

movements of the hand between which the

movements necessary for writing; the voluntary

movement interacts with the involuntary movement

which belongs to the pre-motor area (area 6) which

starts the extra-pyramidal system. For writing-motor

processes are crucial sensory information, properly

those of the sphere proprioceptive (kinaesthetic and

sensitivity) that allow a continuous and accurate

movements (Guyton, A.C., Hall, J.E., 2006),

(Cattaneo, L., 1989).

This information, as well as through the

mediation of the cerebellum and of the specific

sensory areas, also converge directly on neurons of

the areas of motor skills (motor neurons and

neuromuscular junctions) and then operate without

the intervention of conscious perceptions. The

cerebellum has an important influence on all

movements of the arm, as a consequence, the

Figure 2: Example of the use of neurons to sense, motion

and connection.

(source: http://www.analisidellascrittura.com/Image/

neurofisiologia1(1).jpg)

behaviour of handwriting, allowing to perform all

movements with precision and harmony.

Alzheimer's disease is the most common form of

irreversible disabling degenerative dementia with

onset senile, and his one of first and is considered

one of the most serious diseases to social impact

(Borri, M., 2012), (Meek, P.D., McKeithan, E.K.,

Schumock, G.T., 1998), (Zhu, C.W., Sano, M.,

2006). In the brain of the patient is evident brain

atrophy, a sign of the death of neurons, which does

reduce the overall volume of the brain, which is

filled with cerebrospinal fluid and it does enlarge the

ventricles (Vallar, G., Papagno, C., 2007). The

diagnosis is usually confirmed by specific

behavioural assessments and cognitive tests, often

followed from imaging magnetic resonance, but the

computer-images based diagnostics allows only and

solely to ascertain the presence and effects of

cortical atrophy.

The first symptoms that appear are: impaired

memory and not aware of the disease. The course of

the disease and the symptom modes may be different

for each individual patient.

Clinical features: Alzheimer's gets worse, new

symptom joins to the existing, can be added aphasia,

agnosia and apraxia.

Having been in his youth a large volume of

activity mental/cognitive represents a protective

factor against Alzheimer's. The disease can be

recognized only with the neuropsychological

examination and not with the instrumental diagnosis,

and the main diagnosis of Alzheimer's occurs only

through clinical evaluation actually (Waldemar, G.,

Dubois, B., Emre, M., et al, 2007).

A complete description of the neuromuscular

BIODEVICES2014-InternationalConferenceonBiomedicalElectronicsandDevices

194

system is reported in the former paper presented to

the CIBB 2013 Nice (F) (Impedovo, S., et al, 2013)

by looking the anatomic, physiologic and

mathematical aspects, also the data acquisition

devices used for investigation is over there

described. In this paper is instead treated the relation

between handwriting of suspected disease person,

affected by the Alzheimer pathology, and illness

persons.

3 STABILITY MEASURE

OF HANDWRITING

The database (Impedovo, S., 2013) developed and

used at Intelligent System Laboratory of the "Aldo

MORO" Bari University: (ISUNIBA), utilizes a

certain number of handwritten words representing

the "mamma" word, produced exactly by 15 subjects

affected by Alzheimer disease, 16 without any

pathology, 7 namely with senile deficits of various

kind and 5 affected by other neuromuscular disease,

for a total of 43 persons. Each person has written

the "mamma" word 5 times. In the database there

are 215 specimen.

All the images are described in “bmp” format

and represented in black and white. They are named

according the code: “xxx\m-xxx-yy.bmp” for

injured subjects and “xxx\s-xxx-yy.bmp” for illness

subject, where xxx is the number of the writer and

yy is the test number.

The goal of the research is to investigate the

stability of handwritten "mamma" word and it has

been done by comparing the results obtained by

hilliness and injured subjects.

Therefore the proposed approach is a way for a

non invasive analysis to investigate on the wealth

status in persons suspected of disease affections. It

must be said that the choice of the mamma word has

been due to the fact that it is one of the first learned

in speaking and writing and one of the last that is

forget in the human live.

For each specimen processing algorithms have

been used in order to compare each one whit the

others in the data base. In fact after a pre processing

for noise removal each specimen (Larkins, 2009) has

been normalized at a standard dimension.

The median noise removal algorithm has been

used just by coloring the 8 neighboring pixels with

different colors for each pixel and averaging the

value. The results of the techniques is reported in the

next images.

Figure 3: Image affected by noise.

Figure 4: Image with noise removed.

The word dimension normalization has been

obtained by using a specific software based namely

on the reduction of number of pixels along x and y

direction to a fixed number and the results are

reported in the two specimen in the following, the

first being the original and the second the

normalized one.

Figure 5: Original image.

Figure 6: Normalized image.

The Equimass Grid techniques has been used to split

each image in elementary parties. This technique, as

HandwrittenProcessingforPreDiagnosisofAlzheimerDisease

195

it is well known, consists in dividing the image in

non uniform sub regions obtained by dividing the

image with horizontal and vertical lines on the base

of the black pixels included in each region according

to the formula :

Ma

Where Ma is the number of black pixels included in

each horizontal or vertical direction of the specimen

area (Larkins, R.L., 2009). In the following is

reported the results of the algorithm application for a

given specimen consisting of 10 by 10 subregions.

Figure 7: Equimass Grid Algorithm Application Result

for a the given Specimen.

For each stroke into a sub region, the waiting

vector is computed. It consists of twenty values

computed on the base of the black pixels detected in

the containing sub region. In the next figure are

reported the results of the horizontal and vertical

division but also the results of the oblique divisions

along the main and secondary diagonals .

Figure 8: Horizontal and vertical division

Figure 9: Horizontal, vertical and principal and secondary

diagonal divisions.

In order to compare two vector the cosine

similarity has been used, more specifically it is

divided for the vector length in order to normalize

the results just according to the following formula.

CosSim



Vp,Vj





→→

Vp∗Vj

→→

∗



∑

Wip∗Wij







∑

Wip







∗



∑

Wij









Figure 10: Graphic illustration of the cosine similarity

By taking into account the former similarity,

computed region by region, the stability of the

writing is computed for each region. At this purpose

the similarity average value among all the regions is

used according the formula:

stability 

∑

Sij





where r represents the total region number, S is the

similarity, "i" the tract under consideration and "j"

the region name. As matter of fact the stability is a

decimal number represented by a color according

the code scale here reported:

In the next image by using this code are represented

the different stability region by region for the

spacemen under consideration.

The Yoshimura approach (Yoshimura, I.Isao,

Yoshimura, M.Mitsu, 1991) for classification is used

to train the learning phase and to measure the

disease by comparing the characteristic of the stroke

in the patient with that in the data base recorded in

the learning phase. The result is a Boolean value

that measure the disease of the patient.

BIODEVICES2014-InternationalConferenceonBiomedicalElectronicsandDevices

196

Figure 11: Different stable regions in a fragment of the

mamma words.

About the Yoshimura approaches it must be said

that:

for an unknown input specimen to the system, let it

be 

′

, the system compute the distance 1

′

,2

′

,3

′

between 

′

and the corresponding values in the

injured subjects 1,2,3inthedatabase.For this

operation are considered only the three must stable

regions. The distances 1′,2′,3′ between these

strokes and the reference strokes 1,2,3infact

are:

1′ distancebetween

′

and1,

2′ distancebetween

′

and2

3′ distancebetween

′

and3

The maxima distancedmax



among 1



,2



,3



and the minima distance dmin among 1′,2′,3′

are then considered to evaluate the subject under

test and it is considered as an injured one only if





 , and on the contrary he is estimated

as an hilliness subject if it results:



.

Figure 12: Yoshimura similarity approach example.

Obviously it will be the medical doctor that will

suggest new clinical test in the first case.

The Software Diagram in use is divided in

packages as in the figure here in the following:

Figure 13: Class diagram applied on each word.

Figure 14: Class Diagram.

In the following some experimental results about the

Stability are reported: firstly the 10 table for

Alzheimer injured patients, divided in sub region of

the grid and colored:

Patient 1

N Grid Color

Figure 15: Table for the first patient.

MA I N

SaveImage

Image

DbAccess

ViewImage

Erease

WeightVector

Transaction

Binarization

Similarity EquimassGrid

ColorStability

NoiseRemoval

Transforms

Normalization

GUI

HandwrittenProcessingforPreDiagnosisofAlzheimerDisease

197

similarly also the tables for the 2nd,3-rd,4-th,5-

th,6-th,7-th ,8-th,9-th and

the 10-th patient have been considered and

computed the related maps.

Also the 10 maps for illness subjects have been

computed and in the following there is one of them :

Hillness subject 1-st

N Grid Color

Figure16: Table for the first hilliness patient.

4 STATISTICS

By using the colored pie diagram, the statistics

have been obtained and some of the results are in the

following:

Upper part of the first illness subject.

Example of computation for the upper part light

violet:

13:2 = 100: x

X = 2*100÷13= 7,69

In the same way have been processed all the other

parts of all the data in the data base and the

experiments concluded that the middle part of the

word is much more stable than lower and upped

Color

Upper

part

Middle

part

Lower

part

Violet 15,38% 0% 0%

Dark green 7,69% 1,72% 0%

Brown 15,38% 12,07% 0%

Light violet 7,69% 1,72% 8,33%

Gray 7,69% 15,52% 8,33%

Cyan 23,08% 10,34% 41,67%

Orange 0% 10,34% 25%

Pink 7,69% 24,14% 8,33%

Blue 7,69% 8,62% 8,33%

Green 7,69% 13,80% 0%

Red 0% 1,72% 0%

parts in injured persons but in any case they are less

stable than in respect to illness subject.

Also two error type have been considered:

 I type error False Rejection Rate (FRR)

 II type error False Acceptance Error (FAR)

The first error type occurs if a stroke produced by a

illness subject is rejected and of the second type if a

stroke produced by an injured subject is accepted.

In the following the results obtained analyzing

only the stable parts of the word are reported.

FRR: 12,5%

FAR: 32%

instead by considering all the zones the results

obtained have been:

FRR: 0%

FAR: 14,7%

5 CONCLUSIONS

The experimental results shown that are the lower

and upper parts of the handwritten word mamma

that are more affected in stability by the Alzheimer

disease instead the middle part of the word remain

almost stable, obviously it is necessary investigate

more in dept in order to define specific parameters

for this personal disease but these evaluation are

encouraging to continue on this direction of the

research and also defining others similarity measure

as we are starting to do.

REFERENCES

Boldrini, M., 1934. Biometria e antropometria, Giuffrè.

Medugno, V., Valentino, S., Acampora, G., 1999. Sistemi

di elaborazione dell'informazione: Sicurezza su Reti,

University of Salerno. Italy.

BIODEVICES2014-InternationalConferenceonBiomedicalElectronicsandDevices

198

Biondi, E., Cobelli, E., 2001. Storia della Bioingegneria,

Pàtron.

Guyton, A.C., Hall, J.E., 2006. Fisiologia Medica,

Elsevier.

Cattaneo, L., 1989. Anatomia del sistema nervoso centrale

e periferico dell'uomo, Monduzzi.

Borri, M., 2012. Storia della malattia di Alzheimer, Il

Mulino. Bologna.

Meek, P.D., McKeithan, E.K., Schumock, G.T., 1998.

Economic considerations in Alzheimer's disease.

Pharmacotherapy, Pharmacotherapy Publications Inc.

Zhu, C.W., Sano, M., 2006. Economic considerations in

the management of Alzheimer's disease, Dovepress

Journal.

Vallar, G., Papagno, C., 2007. Manuale di

Neuropsicologia, Il Mulino.

Waldemar, G., Dubois, B., Emre, M., et al, 2007.

Recommendations for the diagnosis and management

of Alzheimer's disease and other disorders associated

with dementia, NCBI. USA.

Impedovo, S., Balestrucci, A., Rollo, A., Impedovo, D.,

Pirlo, G., Mangini, F. M., Barbuzzi, D., Sarcinella, L.,

O'Reilly, C., Plamondon, R., 2014. Writing generation

model for health care neuromuscular system

investigation, Springer-Verlag, in press. Nice(F).

Impedovo, S., 2013. ISUNIBA Database

http://www.di.uniba.it/~impedovo/ISUNIBA.htm.

Larkins, R.L., 2009. Off-line Signature Verification,

University of Waikato.

Yoshimura, I.Isao, Yoshimura, M.Mitsu, 1991. Off-line

writer verification using ordinary characters as the

object., PR(24), No. 9.

HandwrittenProcessingforPreDiagnosisofAlzheimerDisease

199