applied
sciences
A icle
Changes in Phona ion and Thei Rela ions wi h
P og ess o Pa kinson’s Disease
Zol an Galaz 1, Ji i Mekyska 1, Voj ech Z oncak 1, Jan Mucha 1, Tomas Kiska 1,
Zdenek Smekal 1, Ilona Eliaso a 2,3, Ma ina M acko a 2,3, Milena Kos alo a 3,4,
I ena Rek o o a 2,3,*, Ma cos Faundez-Zanuy 5, Jesus B. Alonso-He nandez 6
and Ped o Gomez-Vilda 7
1Depa men o Telecommunica ions, B no Uni e si y o Technology, Technicka 10,
616 00 B no, Czech Republic; [email p o ec ed].cz (Z.G.); [email p o ec ed].cz (J.M.);
[email p o ec ed] (V.Z.); [email p o ec ed].cz (J.M.); [email p o ec ed].cz (T.K.);
[email p o ec ed].cz (Z.S.)
2Fi s Depa men o Neu ology, S . Anne’s Uni e si y Hospi al, Peka ska 53, 656 91 B no, Czech Republic;
[email p o ec ed] (I.E.); [email p o ec ed] (M.M.)
3Applied Neu oscience Resea ch G oup, Cen al Eu opean Ins i u e o Technology, Masa yk Uni e si y,
Kamenice 5, 625 00 B no, Czech Republic; [email p o ec ed]
4Depa men o Neu ology, Facul y Hospi al and Masa yk Uni e si y, Jihla ska 20,
639 00 B no, Czech Republic
5Escola Supe io Poli ecnica, Tecnocampus, A da. E nes Lluch 32, 083 02 Ma a o, Ba celona, Spain;
[email p o ec ed]
6Ins i u e o Technological De elopmen and Inno a ion in Communica ions (IDeTIC), Uni e si y o Las
Palmas de G an Cana ia, 35001 Las Palmas de G an Cana ia, Spain; [email p o ec ed]
7Neu omo phic P ocessing Labo a o y (NeuVox Lab), Cen e o Biomedical Technology, Uni e sidad
Poli ecnica de Mad id, Campus de Mon egancedo, s/n, Pozuelo de Ala con, 28223 Mad id, Spain;
[email p o ec ed]
*Co espondence: i ena. ek o [email p o ec ed]; Tel.: +420-543-182-639
Recei ed: 17 Oc obe 2018; Accep ed: 19 No embe 2018; Published: 22 No embe 2018


Abs ac :
Hypokine ic dysa h ia, which is associa ed wi h Pa kinson’s disease (PD), a ec s se e al
speech dimensions, including phona ion. Al hough he scien i ic communi y has deal wi h
a quan i a i e analysis o phona ion in PD pa ien s, a complex esea ch e ealing p obable ela ions
be ween phona o y ea u es and p og ess o PD is missing. The e o e, he aim o his s udy is o
explo e hese ela ions and model hem ma hema ically o be able o es ima e p og ess o PD du ing
a wo-yea ollow-up. We en olled 51 PD pa ien s who we e assessed by h ee commonly used clinical
scales. In addi ion, we quan i ied eigh possible phona o y diso de s in i e owels. To iden i y
he ela ionship be ween baseline phona o y ea u es and changes in clinical sco es, we pe o med
a pa ial co ela ion analysis. Finally, we ained XGBoos models o p edic he changes in clinical
sco es du ing a wo-yea ollow-up. Fo wo yea s, he pa ien s’ oices became mo e ape iodic
wi h inc eased mic ope u ba ions o equency and ampli ude. Nex , he XGBoos models we e
able o p edic changes in clinical sco es wi h an e o in ange 11–26%. Al hough we iden i ied
some signi ican co ela ions be ween changes in phona o y ea u es and clinical sco es, hey a e
less in e p e able. This s udy sugges s ha i is possible o p edic he p og ess o PD based on he
acous ic analysis o phona ion. Mo eo e , i ecommends u ilizing he sus ained owel /i/ ins ead
o /a/.
Keywo ds:
phona ion; acous ic analysis; ollow-up s udy; hypokine ic dysa h ia; Pa kinson’s disease
Appl. Sci. 2018,8, 2339; doi:10.3390/app8122339 www.mdpi.com/jou nal/applsci
Appl. Sci. 2018,8, 2339 2 o 18
1. In oduc ion
Pa kinson’s disease (PD) is a equen neu odegene a i e diso de ha is associa ed wi h
a subs an ial educ ion o dopamine gic neu ons especially in subs ancia nig a pa s compac a [
1
].
The p ima y mo o symp oms o PD comp ise emo a es , muscula igidi y, b adykinesia, and
pos u al ins abili y [
1
]. Pa ien s wi h PD also de elop a a ie y o non-mo o symp oms [
2
] such
as sleep dis u bances, dep ession, cogni i e impai men , e c. To diagnose, a e and moni o mo o
and non-mo o symp oms o PD, a ious clinical a ing scales such as Uni ied Pa kinson’s Disease
Ra ing Scale (UPDRS) [
3
], F eezing O Gai Ques ionnai e (FOG-Q) [
4
], o Addenb ooke’s Cogni i e
Examina ion-Re ised (ACE–R) [
5
] ha e been de eloped. Ne e heless, eliabili y o he assessmen is
o en educed by in e - a e a iabili y [6].
Up o 90% [
7
] o pa ien s wi h PD de elop a mul i-dimensional speech diso de named
hypokine ic dysa h ia (HD) [
8
], which is mani es ed in phona ion, a icula ion, and p osody [
9
–
11
].
In he a ea o phona ion, insu icien b ea h suppo , educ ion in phona ion ime, inc eased acous ic
noise, ins abili y o a icula o y o gans, mic ope u ba ions o equency/ampli ude, and ha sh b ea hy
oice quali y has been obse ed [
9
,
12
]. HD leads o se ious complica ions in daily communica ion o
pa ien s wi h PD [13]. Gene ally, HD was ound o be mo e se e e in he ad anced s ages o PD [14].
As epo ed by he ecen s udies, acous ic analysis o HD can p o ide clinicians wi h
non-in asi e and eliable me hodology o PD diagnosis, assessmen and moni o ing [
9
,
15
]. Mo eo e ,
his me hodology has also been used o moni o he e iciency o PD ea men [
10
,
16
–
18
]. In he ield
o acous ic analysis o PD phona ion, he au ho s mos ly ocused on he sus ained owel /a/ [
9
].
Con en ional phona o y ea u es such as ji e , shimme , ha monic- o-noise a io, deg ee o un oiced
segmen s, and o man -based pa ame e s ex ac ed om his owel ha e been widely used o diagnose
PD [
12
,
19
–
23
]. Al hough Hazan e al. [
24
] employed analysis o sus ained phona ion o diagnosis
o PD e en in i s ea ly s age, based on he ecen e iew [
9
], mos o he esea che s ind ele an
applica ions o he phona o y analysis especially in mode a e o se e e s ages o his diso de .
Fo example, he analysis o sus ained phona ion has been u ilized du ing PD se e i y assessmen .
In 2010, Tsanas e al. [
15
] en olled 42 PD pa ien s and pa ame e ized hei sus ained phona ion o
owel /a/ by a se o con en ional ea u es ha we e consequen ly mapped o UPDRS, pa III (mo o
examina ion) and he o al sco e o his scale. Using classi ica ion and eg ession ees, hey es ima ed
he UPDRSIII sco e wi h MAE (mean absolu e e o ) equal o 5.95. The o al UPDRS sco e was
es ima ed wi h
MAE =
7.52. A pa ame ic e sion o his da ase has been made a ailable o esea ch
pu poses and o he esea ch eams u he dec eased he es ima ion e o [
25
–
27
]. Ano he wo k ha
deals wi h he au oma ic clinical sco es es ima ion was published by Mekyska e al. [
21
]. In his s udy,
hey acqui ed sus ained phona ion o owels /a/, /e/, /i/, /o/, /u/ in 84 PD pa ien s. Modeling
con en ional and ad anced ea u es by andom o es s p o ided he es ima ion o UPDRSIII wi h
MAE =
5.70. In addi ion, he au ho s es ima ed se e al o he clinical sco es such as UPDRS, pa IV
(complica ions o he apy) wi h MAE =1.30 o Beck dep ession in en o y (BDI) wi h MAE =3.12.
E en hough HD is one o he mos p oblema ic aspec s o PD, he numbe o longi udinal
s udies in es iga ing he e olu ion o HD in PD o e ime (based on he acous ic analysis) is e y
limi ed [
28
–
31
]. I we ocus speci ically on longi udinal moni o ing o sus ained phona ion, hen, in ac ,
we can iden i y only one s udy, which is published by Skodda e al. [
31
]. In his wo k, he au ho s
epea edly (wi h a e age ime in e al 32.50 mon hs) acqui ed sus ained owel /a/ in 32 emale and
48 male PD pa ien s (age in session 1: 66.28
±
8.11 yea s; PD du a ion in session 1: 6.10
±
4.63 yea s;
UPDRSIII in session 1: 20.16 ±10.96; UPDRS III in session 2: 19.58 ±8.29). The oice was quan i ied
by ji e , shimme , noise- o-ha monic a io, and mean undamen al equency. Based on he pai ed
- es , he au ho s iden i ied signi ican changes in shimme and noise- o-ha monic a io. In bo h
cases, he alues o hese pa ame e s inc eased. Ano he in e es ing inding is ha , al hough some
phona o y ea u es signi ican ly changed, UPDRS III was held widely s able o e ime. The au ho s
p o ide wo possible explana ions: (1) oice impai men could be he esul o an escala ion o axial
dys unc ion oo sub le o be mi o ed by UPDRSIII; (2) al e a ions o speech pa ame e s could be
Appl. Sci. 2018,8, 2339 3 o 18
comple ely independen o mo o pe o mance ha may be based upon non-dopamine gic mechanisms.
Inconsis encies in e ms o he L-dopa e ec on HD a e u he discussed in B abenec e al. [9].
To sum i up, al hough he scien i ic communi y equen ly add esses phona ion in associa ion
wi h HD (especially when diagnosing o assessing PD), o he bes o ou knowledge, he e is only one
s udy ha ocuses on HD phona o y diso de s om a longi udinal pe spec i e. Mo eo e , he wo k
deals wi h he analysis o phona ion jus pa ially, i conside s only he sus ained owel /a/, and i
does no explo e a possibili y o PD p og ess p edic ion based on a combina ion o acous ic analysis
and machine lea ning. The e o e, in he ame o ou wo-yea ollow-up s udy, we a e going much
u he wi h he ollowing aims:
1. o iden i y phona o y acous ic ea u es a baseline ha a e signi ican ly co ela ed wi h changes
in a ious clinical a ing scales,
2.
o in es iga e ela ionship be ween changes in he phona o y acous ic ea u es and he clinical
a ing scales a e he wo-yea ollow-up,
3. o es ablish ma hema ical models ha will es ima e he change in clinical a ing scales based on
he change in acous ic measu es,
4. o compa e esul s based on i e owels: /a/, /e/, /i/, /o/, /u/.
The es o his a icle is o ganized as ollows: Sec ion 2desc ibes a da ase o PD pa ien s as well
as me hodology in e ms o acous ic analysis, s a is ical analysis and machine lea ning. Resul s a e
epo ed in Sec ion 3and consequen ly discussed in Sec ion 5. Finally, conclusions a e gi en in
Sec ion 4.
2. Ma e ials and Me hods
2.1. Da ase
In his wo k, we en olled 51 pa ien s wi h idiopa hic PD. All o hem a e Czech na i e speake s
(17 emales and 34 males; age: 65.47
±
7.46 yea s; PD du a ion: 7.61
±
4.01 yea s; mean LED (L-dopa
equi alen daily dose) [
32
]: 1033.67
±
567.96 mg/day) a he Fi s Depa men o Neu ology, S . Anne’s
Uni e si y Hospi al in B no, Czech Republic. A e wo yea s, he pa ien s we e e-examined (age:
67.61
±
7.38 yea s; PD du a ion: 9.57
±
4.50 yea s; mean LED: 1115.11
±
484.38 mg/day). All pa ien s
signed an in o med consen o m ha has been app o ed (including he s udy) in 14 Ma ch 2016 by
he Resea ch E hics Commi ee o Masa yk Uni e si y ( e . no.: EKV-2016-004, p ojec i le: E ec s
o non-in asi e b ain s imula ion on hypokine ic dysa h ia, mic og aphia, and b ain plas ici y in
pa ien s wi h Pa kinson’s disease, in es iga o : P o . MD. I ena Rek o o a, PhD.).
None o he pa ien s had a disease a ec ing he cen al ne ous sys em o he han PD. All pa ien s
we e examined on hei egula dopamine gic medica ion app oxima ely 1 h a e he L-dopa [
32
]
dose. The ollowing a ing scales we e used o e alua e he clinical symp oms o PD: UPDRSIII
and UPDRSIV [
3
], FOG-Q [
4
], REM sleep beha io diso de sc eening ques ionnai e (RBDSQ) [
33
],
and ACE-R [
5
]. The ull clinical cha ac e is ics o he da ase , ie., mean
±
sd alues o he clinical
a ing scales in session 1, session 2, and session
∆
(
session 2 −session 1
) can be seen in Table 1.
Mo eo e , o iden i y s a is ically signi ican di e ences, he able epo s p- alues o he Wilcoxon
signed- ank es be ween he da a acqui ed in session 1 (baseline examina ion) and session 2 ( wo-yea
ollow-up examina ion) oo.
The clinical da a om he
∆
session we e also used o gene a e desc ip i e isualiza ions
(i.e., his og ams, eg ession and esidual plo s) o he change in selec ed clinical a ing scales,
mo e speci ically: LED, UPDRS III, UPDRSIV, FOG-Q, RBDSQ, ACE-R, see Figu e 1. Wi h his app oach,
i is possible o assess he imp o emen and/o decline in mo o and non-mo o de ici s associa ed
wi h PD in he ho izon o wo yea s as well as a ela ionship be ween he change in each o he scales
ela i e o o he scales in he selec ed se .
Appl. Sci. 2018,8, 2339 4 o 18
Table 1. Clinical cha ac e is ics o he pa ien s.
Scale Mean ±sd (s1) Mean ±sd (s2) Mean ±sd (∆)p(Wilcoxon)
LED 917.61 ±544.78 1129.92 ±477.50 212.31 ± −67.28 0.188
UPDRSIII 22.49 ±13.47 27.45 ±12.68 4.96 ± −0.79 0.000
UPDRSIV 2.82 ±2.58 3.44 ±2.94 0.62 ±0.36 0.632
FOG-Q 6.57 ±5.40 8.33 ±5.97 1.76 ±0.57 0.000
RBDSQ 3.98 ±3.25 3.78 ±2.28 −0.2 ± −0.98 0.522
ACE-R 87.92 ±7.62 85.89 ±9.48 −2.03 ±1.86 0.000
s1— i s session; s2—second session;
∆
—del a session (
session 2 −session 1
); p(Wilcoxon) — p- alue
o Wilcoxon signed- ank es (pai ed samples); LED—L-dopa equi alen daily dose (mg/day) [
32
];
UPDRSIII—Uni ied Pa kinson’s Disease Ra ing Scale, pa III: e alua ion o mo o unc ion [
3
],
UPDRSIV—Uni ied Pa kinson’s Disease Ra ing Scale, pa IV: e alua ion o complica ions o
he apy [
3
]; FOG-Q—F eezing o gai ques ionnai e [
4
]; RBDSQ—The REM sleep beha io diso de
sc eening ques ionnai e [33]; ACE-R—Addenb ooke’s Cogni i e Examina ion-Re ised [5].





















 






  

 

  

 

  

Figu e 1.
Desc ip i e s a is ical g aphs o clinical cha ac e is ics o he PD pa ien da ase : on he
main diagonal, his og ams a e isualized. Nex , he uppe iangula pa o he g aph-g id shows
sca e plo s wi h he i ed lines o linea eg ession models. Finally, he lowe iangula pa o he
g aph-g id is used o display esiduals o he models shown in he uppe g id. Colo no a ion: he blue
colo ep esen s da a o session 1, and he g een colo ep esen s da a o session 2.
Appl. Sci. 2018,8, 2339 5 o 18
2.2. Vocal Tasks
To quan i y he de e io a ion o phona ion in pa ien s wi h PD, we used a sus ained phona ion
o owels: /a/, /e/, /i/, /o/, /u/ as a basis o ou expe imen s. The eason behind using all o
he i e owels is o employ he analysis wi h he emphasis on quan i ying all posi ions o a ongue
du ing phona ion. Fo mo e in o ma ion, see he Hellwag ( owel) iangle [
34
]. In ou iew, using
only a sus ained phona ion o he owel /a/ is no ully jus i ied as he e is e y li le o no eason
o assume ha his pa icula posi ion o he ongue can p o ide mo e in o ma ion abou phona o y
diso de s. In ac , as shown by p e ious s udies, he analysis o o he owels is impo an o a mo e
obus desc ip ion o HD [19–21,23,24,35–37].
Sus ained phona ion o a owel is a s anda d measu e used o assess quali y o phona ion [
9
].
Du ing his pa icula ocal ask, a speake is asked o sus ain phona ion o a owel, a emp ing o
main ain s eady equency and ampli ude a a com o able le el [
38
]. The ad an age o his ask
in compa ison wi h o he commonly used ocal asks is i s independence o a icula o y and o he
linguis ic con ounds [
38
]. Mo eo e , i is also p esen in mos o he da abases and he e o e he
expe imen s p oposed in ou wo k a e compa able wi h o he commonly used da abases [39,40].
The sus ained phona ion ask used in his s udy is a pa o a speech acquisi ion p o ocol de i ed
om he s anda dized 3F Dysa h ia P o ile [
41
]. Du ing he da a acquisi ion, a la ge capsule ca dioid
mic ophone M-AUDIO No a (Cumbe land, RI, Uni ed S a es) moun ed o a boom a m RODE PSA1
(Sil e wa e , Aus alia) and posi ioned a a dis ance o app oxima ely 20cm om he pa ien ’s
mou h was used o he eco ding. Consequen ly, he signals we e digi ized by audio in e ace
M-AUDIO Fas T ack P o (Cumbe land, RI, Uni ed S a es) wi h he sampling equency o 48kHz
(16-bi esolu ion) and checked by a ained acous ic enginee wi hou ha ing seen he pa ien ’s clinical
da a. Finally, he signals we e pa ame e ized using P aa [
42
] so wa e as well as a se o MATLAB
(MATLAB 9.4, Ma hWo ks, Na ick, MA, Uni ed S a es) pa ame iza ion unc ions [
43
] de eloped a
he B no Uni e si y o Technology.
2.3. Acous ic Fea u es
To desc ibe a a ie y o phona o y diso de s associa ed wi h HD, we quan i ied he
ollowing: (a) mic ope u ba ions in equency o oice using pe iod pe u ba ion quo ien (PPQ);
(b) mic ope u ba ions in in ensi y o oice using ampli ude pe u ba ion quo ien (APQ); (c) i egula
pi ch luc ua ions using coe icien o a ia ion o undamen al equency (F0 (CV)); (d) i egula
ampli ude luc ua ions using coe icien o a ia ion o Teage –Kaise ope a o (TKEO (CV)); (e) emo
o a icula o y o gans (such as jaw, ongue and lips), coe icien o a ia ion o 1s o man (F1 (CV)),
coe icien o a ia ion o 2nd o man (F2 (CV)), coe icien o a ia ion o 3 d o man (F3 (CV));
( ) inc eased acous ic noise using median o ha monic- o-noise a io (HNR (Q2)), median o ene gy
a io (ER (Q2), ene gy a io o bands 2000–4000Hz and 70–900Hz)), median o glo al- o-noise
exci a ion a io (GNE (Q2)), median o no malized noise ene gy (NNE (Q2)); (g) i egula acous ic
noise luc ua ions using s anda d de ia ion o ha monic- o-noise a io (HNR (SD)), coe icien o
a ia ion o ene gy a io (ER (CV)), s anda d de ia ion o glo al- o-noise exci a ion a io (GNE (SD)),
s anda d de ia ion o no malized noise ene gy (NNE (SD)); and (h) ape iodici y o oice using ac ion
o locally un oiced ames (FLUF). All o hese ea u es a e s anda d and clinically in e p e able
dysphonic measu es and we e selec ed based on a ecommenda ion gi en in ou ecen e iew on
acous ic analysis o oice/speech signals in pa ien s su e ing om HD [
9
]. Fo mo e in o ma ion
abou he oice/speech pa ame iza ion, see [43].
2.4. S a is ical Analysis
Be o e desc ibing he analy ical se up applied in his wo k, i is impo an o men ion ha he
da ase did no con ain any missing alues, and he e o e all da a samples we e used. Fu he mo e,
e en hough we used six clinical a ing scales when desc ibing he da ase (see Sec ion 2.1), only ou

Appl. Sci. 2018,8, 2339 6 o 18
o hese scales we e used o he analysis, speci ically: UPDRSIII, UPDRS IV, RBDSQ, and FOG-Q.
The eason is ha p e ious s udies ha e al eady shown ha non-mo o mani es a ions o PD a e
no linked wi h he phona o y aspec s o HD, bu a he wi h he impai men s o p osody and
a icula ion [
44
] ha a e commonly being quan i ied using a sen ence eading ask, ee speech
(monologue), e c. Since his s udy is ocused on he phona o y aspec s o HD, clinical a ing scales
desc ibing only mo o symp oms o PD we e used.
To e eal and assess he s eng h o a ela ionship be ween he compu ed acous ic ea u es and
pa ien s’ clinical da a (UPDRSIII, UPDRSIV, RBDSQ, and FOG-Q), Spea man’s co ela ion coe icien
was compu ed ( he s a is ical assump ions o Spea man’s co ela ion coe icien we e sa is ied as:
(a) he acous ic ea u es as well as he he clinical da a a e bo h a iables ha a e measu ed on a
leas an o dinal scale, and (b) he e is a mono onic ela ionship be ween he wo a iables). Since age,
gende , and p obably L-dopa, a e mani es ed in a oice o PD pa ien s [
9
], o he pu pose o his wo k,
we employed pa ial Spea man’s co ela ion con olling o he e ec o he ollowing con ounding
ac o s (also known as co a ia es): pa ien s’ age, gende [
29
,
45
], and dopamine gic medica ion [
32
,
46
].
The signi icance le el o co ela ion was se o 0.05. Mo e speci ically, wo co ela ion scena ios we e
conside ed: (a) co ela ion be ween he acous ic ea u es a he baseline and he change in alues
o he selec ed clinical a ing scales, and (b) co ela ion be ween he change in he acous ic ea u es
and he change in he alues o he selec ed clinical a ing scales. Wi h his app oach, we aimed
a iden i ying hose acous ic ea u es ha a e signi ican ly co ela ed wi h he speci ic mo o and
non-mo o symp oms assessed by he selec ed clinical a ing scales in bo h scena ios.
Nex , o e alua e he powe o he acous ic ea u es a he baseline o p edic he change o he
pa ien s’ clinical da a in he ho izon o wo yea s, we used he acous ic ea u es compu ed o he
eco dings acqui ed in session 1 (baseline examina ion) and buil ma hema ical models p edic ing
he change in he selec ed clinical a ing scales (
∆
). Fo his pu pose, we employed G adien Boos ed
T ees (mo e speci ically, he amous XGBoos algo i hm [
47
]) in a supe ised lea ning se up: 10- old
c oss- alida ion wi h 20 epe i ions [
48
]. The XGBoos algo i hm belongs o he s a e-o - he-a in
machine lea ning, which is suppo ed by he ac ha i has been ecen ly used o win compe i ions on
Kaggle. I wo ks well e en on small da ase s (whe e i ou pe o ms deep lea ning app oaches), i is
obus o ou lie s and i is able o model complex in e dependencies. Fo hese easons, i has been
used by many esea che s in a ious biomedical ields, e.g., [49–51], e c.
The pe o mance o he models (p ecision o he p edic ions) was e alua ed by MAE and
es ima ion e o a e (EER). These measu es a e de ined as:
MAE =1
n
n
∑
i=1
|yi−ˆ
yi|,
EER =1
n·
n
∑
i=1
|yi−ˆ
yi| · 100 [%],
whe e
yi
s ands o he ue label o
i
- h obse a ion,
ˆ
yi
ep esen s he p edic ed label o he
i
- h
obse a ion,
n
deno es he numbe o obse a ions, and inally
s ands o he ange o alues in he
p edic ed clinical a ing scale (no he ange ha can be heo e ically eached, bu he ac ual ange o
he alues in he da ase ). As can be seen, EER he e o e desc ibes a pe cen age o e o p edic ions
wi h espec o s a is ical p ope ies o he da ase , which is pa icula ly use ul o easy in e p e a ion
o he esul s.
3. Resul s
The alues o 16 acous ic ea u es ex ac ed om bo h sessions, as well as alues o hei di e ences
(session 2
−
session 1), a e epo ed in Table 2. Based on he Wilcoxon signed- ank es , we can obse e
ha none o he ea u es ex ac ed om owel /a/ signi ican ly changed a e wo yea s. Rega ding
owel /e/, we iden i ied signi ican ly inc eased mic ope u ba ions in in ensi y o oice and also
Appl. Sci. 2018,8, 2339 7 o 18
inc eased ape iodici y. The same signi ican changes we e iden i ied in owel /i/ and /u/. In he
case o owel /u/, in addi ion, we moni o ed he inc ease o mic ope u ba ions in equency o oice.
The epea ed acquisi ion o owel /o/ was associa ed wi h inc eased ape iodici y and mo e dominan
mic ope u ba ions in equency o oice.
Table 2. S a is ical desc ip ion o acous ic ea u es o all ocal asks.
Fea u e Mean ±sd (s1) Mean ±sd (s2) Mean ±sd (∆)p(Wilcoxon)
owel /a/
PPQ 1.31 ±1.28 1.89 ±2.52 0.58 ±1.24 0.069
APQ 10.69 ±4.23 12.81 ±6.66 2.12 ±2.42 0.063
FLUF 3.87 ±5.28 4.77 ±5.61 0.90 ±0.32 0.386
HNR (Q2) 13.35 ±2.90 12.81 ±3.73 −0.54 ±0.83 0.390
HNR (SD) 4.17 ±0.86 4.18 ±0.89 0.02 ±0.03 0.928
F1 (CV) 0.15 ±0.06 0.18 ±0.10 0.02 ±0.04 0.180
F2 (CV) 0.22 ±0.15 0.23 ±0.15 0.00 ±0.00 0.913
F3 (CV) 8.25 ±40.44 4.50 ±21.67 −3.75 ± −18.77 0.575
ER (Q2) 6.98 ±33.97 7.40 ±36.43 0.42 ±2.46 0.954
ER (CV) 0.57 ±0.26 0.57 ±0.30 0.01 ±0.05 0.880
F0 (CV) 0.30 ±0.74 0.36 ±1.15 0.06 ±0.41 0.763
GNE (Q2) −0.46 ±1.85 −0.46 ±1.85 0.00 ± −0.00 0.997
GNE (SD) 0.23 ±0.74 0.30 ±0.97 0.07 ±0.23 0.679
TEO (CV) −0.34 ±1.71 −0.31 ±1.75 0.03 ±0.04 0.931
NNE (Q2) −1.45 ±7.41 −1.33 ±7.08 0.13 ± −0.34 0.932
NNE (SD) 1.80 ±0.76 1.79 ±0.78 −0.01 ±0.02 0.955
owel /e/
PPQ 1.31 ±1.17 1.80 ±3.06 0.50 ±1.89 0.269
APQ 11.21 ±6.82 15.05 ±9.95 3.84 ±3.13 0.036
FLUF 2.62 ±3.58 5.31 ±6.02 2.69 ±2.44 0.007
HNR (Q2) 14.36 ±3.92 13.75 ±4.40 −0.61 ±0.48 0.394
HNR (SD) 4.19 ±0.96 4.32 ±1.18 0.13 ±0.21 0.533
F1 (CV) 0.62 ±0.23 0.56 ±0.24 −0.05 ±0.01 0.173
F2 (CV) 0.19 ±0.07 0.19 ±0.09 −0.00 ±0.02 0.965
F3 (CV) 12.41 ±86.28 7.06 ±48.94 −5.36 ± −37.34 0.709
ER (Q2) 3.46 ±23.30 3.39 ±22.83 −0.07 ± −0.47 0.989
ER (CV) 0.59 ±0.28 0.61 ±0.37 0.02 ±0.10 0.702
F0 (CV) 0.42 ±1.47 0.19 ±0.50 −0.24 ± −0.97 0.293
GNE (Q2) −0.36 ±1.77 −0.17 ±1.18 0.19 ± −0.59 0.536
GNE (SD) 0.08 ±0.36 0.06 ±0.30 −0.02 ± −0.07 0.726
TEO (CV) −0.07 ±0.50 −0.19 ±1.31 −0.11 ±0.80 0.571
NNE (Q2) −0.19 ±1.35 −0.21 ±1.44 −0.01 ±0.10 0.961
NNE (SD) 1.67 ±0.56 1.83 ±0.61 0.17 ±0.05 0.055
owel /i/
PPQ 1.26 ±1.68 1.92 ±3.08 0.66 ±1.40 0.196
APQ 10.49 ±5.31 14.83 ±9.65 4.35 ±4.34 0.005
FLUF 2.13 ±3.73 4.60 ±6.89 2.46 ±3.15 0.013
HNR (Q2) 17.16 ±3.38 16.17 ±5.32 −0.98 ±1.95 0.212
HNR (SD) 4.51 ±1.12 4.53 ±1.40 0.02 ±0.28 0.921
F1 (CV) 0.67 ±0.48 0.58 ±0.40 −0.09 ± −0.08 0.170
F2 (CV) 0.14 ±0.07 0.16 ±0.08 0.02 ±0.00 0.088
F3 (CV) 0.09 ±0.21 38.22 ±207.82 38.14 ±207.61 0.205
ER (Q2) 0.09 ±0.08 8.35 ±33.15 8.27 ±33.07 0.087
ER (CV) 0.61 ±0.37 0.56 ±0.44 −0.05 ±0.07 0.490
F0 (CV) 0.36 ±1.39 0.35 ±1.17 −0.01 ± −0.22 0.961
GNE (Q2) −0.30 ±1.49 −0.30 ±1.48 0.01 ± −0.01 0.980
GNE (SD) 0.07 ±0.32 0.12 ±0.40 0.06 ±0.08 0.449
TEO (CV) 0.78 ±0.55 0.36 ±2.47 −0.42 ±2.47 0.238
NNE (Q2) −0.01 ±0.18 −0.20 ±11.53 −0.19 ±1.15 0.228
NNE (SD) 1.44 ±0.55 1.40 ±0.70 −0.05 ±0.15 0.727
Appl. Sci. 2018,8, 2339 8 o 18
Table 2. Con .
Fea u e Mean ±sd (s1) Mean ±sd (s2) Mean ±sd (∆)p(Wilcoxon)
owel /o/
PPQ 1.14 ±1.06 1.68 ±2.06 0.54 ±0.99 0.047
APQ 11.09 ±4.41 14.16 ±9.94 3.06 ±5.53 0.051
FLUF 2.64 ±4.32 5.63 ±7.24 2.99 ±2.91 0.008
HNR (Q2) 15.49 ±3.30 15.28 ±4.79 −0.22 ±1.49 0.773
HNR (SD) 4.93 ±1.14 4.57 ±1.36 −0.36 ±0.21 0.149
F1 (CV) 0.24 ±0.19 0.31 ±0.22 0.06 ±0.03 0.091
F2 (CV) 0.14 ±0.10 0.14 ±0.10 0.00 ±0.01 0.867
F3 (CV) 8.75 ±35.36 23.86 ±96.34 15.11 ±60.97 0.317
ER (Q2) 8.17 ±32.52 8.07 ±32.42 −0.10 ± −0.10 0.989
ER (CV) 0.64 ±0.39 0.61 ±0.36 −0.03 ± −0.02 0.705
F0 (CV) 0.40 ±0.91 0.63 ±1.81 0.24 ±0.89 0.434
GNE (Q2) −0.91 ±2.72 −1.25 ±3.11 −0.35 ±0.39 0.582
GNE (SD) 0.30 ±0.75 0.39 ±0.90 0.09 ±0.15 0.580
TEO (CV) −0.74 ±3.27 −0.28 ±1.30 0.46 ± −1.97 0.376
NNE (Q2) −0.19 ±0.79 −0.11 ±0.47 0.09 ± −0.31 0.530
NNE (SD) 1.62 ±0.83 1.58 ±0.91 −0.04 ±0.08 0.843
owel /u/
PPQ 1.35 ±1.17 2.60 ±3.07 1.26 ±1.90 0.009
APQ 12.66 ±5.60 17.03 ±9.50 4.37 ±3.91 0.007
FLUF 2.77 ±5.13 8.52 ±9.83 5.75 ±4.69 0.001
HNR (Q2) 15.28 ±4.23 14.32 ±5.22 −0.96 ±0.99 0.270
HNR (SD) 5.40 ±1.56 5.08 ±1.46 −0.32 ± −0.10 0.252
F1 (CV) 0.69 ±0.44 0.71 ±0.34 0.03 ± −0.10 0.667
F2 (CV) 0.17 ±0.09 0.18 ±0.09 0.01 ± −0.00 0.468
F3 (CV) 10.28 ±52.21 15.04 ±104.80 4.76 ±52.58 0.779
ER (Q2) 8.27 ±33.82 3.63 ±25.29 −4.64 ± −8.53 0.453
ER (CV) 0.67 ±0.39 0.74 ±0.43 0.06 ±0.04 0.431
F0 (CV) 0.23 ±0.67 0.18 ±0.10 −0.05 ± −0.57 0.586
GNE (Q2) −0.19 ±1.31 0.00 ±0.00 0.19 ± −1.31 0.323
GNE (SD) 0.09 ±0.40 0.10 ±0.70 0.01 ±0.30 0.966
TEO (CV) −0.51 ±2.69 −0.07 ±0.51 0.43 ± −2.18 0.275
NNE (Q2) −1.27 ±7.09 −0.31 ±2.190 0.96 ± −4.91 0.373
NNE (SD) 1.57 ±0.73 1.68 ±0.49 0.11 ± −0.24 0.379
The esul s o Spea man’s pa ial co ela ion be ween he baseline acous ic ea u es (session 1)
and change in clinical da a (
∆
) can be seen in Table 3. None o he ea u es signi ican ly co ela ed
wi h UPDRS, pa III. On he o he hand, in he case o pa IV, we can obse e nega i e co ela ion
wi h ape iodici y (FLUF, owels /e/, /i/, /o/, /u/), i.e., low ape iodici y a he baseline esul ed in
inc eased complica ions wi h he apy. Simila ly, we iden i ied nega i e co ela ion wi h emo o jaw
(F2 (CV), owel /a/), bu posi i e co ela ion wi h he emo o lips (F3 (CV), owel /o/). Ano he
posi i e co ela ions we e obse ed wi h median o ene gy a io ( owels /o/, /u/), i egula pi ch
luc ua ions (F0 (CV), owel /a/), and a iabili y o oice quali y (GNE (SD), owel /a/). Change
in UPDRSIV nega i ely co ela ed wi h i egula ampli ude luc ua ions (TEO (CV), owel /u/),
acous ic noise (NNE (Q2), owel /u/) and i s a ia ion (NNE (SD), owel /a/). Resul s linked wi h
he acous ic noise quan i ied by he median GNE a e no consis en .
RBDSQ signi ican ly and posi i ely co ela ed wi h mic ope u ba ions in equency o oice (PPQ,
owel /u/) and mic ope u ba ions o i s in ensi y (APQ, owel /a/), i.e., inc eased mic ope u ba ions
in equency/ampli ude a he baseline esul ed in de e io a ion o sleep. In addi ion, RBDSQ
nega i ely co ela ed wi h he a ia ion o oice quali y (HNR (SD), owel /o/).
Rega ding gai di icul ies, as assessed by FOG-Q, we can obse e wo posi i e co ela ions wi h
emo o jaw (F1 (CV), owel /i/) and i egula pi ch luc ua ions (F0 (CV), owel /a/). The o al sco e
o his ques ionnai e nega i ely co ela es wi h a ia ion o acous ic noise (NNE (SD), owel /o/).
Appl. Sci. 2018,8, 2339 9 o 18
Table 3. Spea man’s co ela ion coe icien s be ween baseline acous ic ea u es and ∆o clinical da a.
Fea u e /a/ /e/ /i/ /o/ /u/ /a/ /e/ /i/ /o/ /u/
UPDRSIII UPDRSIV
PPQ −0.07 −0.08 0.26 −0.09 −0.10 −0.11 −0.26 −0.08 −0.17 −0.08
APQ −0.06 −0.05 0.17 −0.10 −0.09 0.08 0.12 −0.01 0.06 −0.09
FLUF 0.10 0.08 0.11 0.10 0.16 −0.23 −0.47 ** −0.33 * −0.34 * −0.32 *
HNR (Q2) 0.16 0.07 −0.05 0.17 0.09 −0.02 −0.08 0.07 0.11 0.16
HNR (SD) 0.10 0.11 −0.04 −0.05 −0.15 0.07 −0.02 0.20 −0.22 −0.09
F1 (CV) 0.04 0.10 −0.17 −0.02 0.19 −0.27 −0.07 0.08 −0.32 * 0.08
F2 (CV) −0.15 0.20 −0.10 −0.11 0.11 −0.37 * −0.11 −0.12 −0.23 0.04
F3 (CV) −0.24 0.25 −0.23 −0.17 0.11 0.22 0.28 0.10 0.32 * 0.10
ER (Q2) −0.25 0.25 −0.14 −0.17 0.17 0.16 0.28 0.12 0.34 * 0.32 *
ER (CV) 0.12 −0.09 0.22 −0.03 0.11 −0.17 −0.14 −0.22 −0.18 −0.06
F0 (CV) 0.28 −0.12 0.04 −0.08 0.17 0.33 * −0.04 −0.24 −0.16 −0.21
GNE (Q2) −0.14 0.11 −0.15 0.06 −0.17 −0.35 * 0.25 0.15 0.27 0.33 *
GNE (SD) −0.16 −0.08 −0.12 −0.22 0.24 0.30 * −0.01 −0.04 0.22 0.11
TEO (CV) 0.28 −0.25 0.11 0.15 −0.25 −0.23 −0.28 −0.28 −0.23 −0.30 *
NNE (Q2) 0.28 −0.25 −0.09 0.14 −0.23 −0.15 −0.28 −0.22 −0.26 −0.31 *
NNE (SD) 0.09 −0.25 0.12 0.11 −0.26 −0.43 ** −0.17 −0.28 −0.12 −0.12
RBDSQ FOG-Q
PPQ 0.15 0.11 0.21 −0.13 0.33 * −0.20 −0.09 0.15 −0.09 0.08
APQ 0.29 * 0.27 0.21 0.27 0.20 −0.18 −0.14 −0.16 −0.16 −0.12
FLUF −0.06 0.06 −0.22 −0.17 −0.24 0.20 0.15 0.13 0.17 0.23
HNR (Q2) −0.20 −0.17 −0.19 −0.27 −0.20 0.09 0.03 −0.03 0.09 −0.05
HNR (SD) −0.17 −0.16 0.21 −0.36 * −0.27 0.07 0.24 0.21 0.18 0.12
F1 (CV) 0.08 −0.02 0.24 0.09 0.11 0.16 −0.22 0.33 * −0.20 −0.20
F2 (CV) −0.08 −0.09 −0.26 −0.17 −0.27 0.18 −0.09 −0.03 −0.13 0.26
F3 (CV) −0.10 −0.04 −0.11 0.07 −0.08 0.03 0.22 0.14 0.08 −0.11
ER (Q2) −0.19 −0.21 −0.26 −0.12 −0.16 −0.13 0.04 0.25 0.10 −0.05
ER (CV) 0.08 0.28 0.20 −0.28 −0.11 0.04 −0.13 −0.11 0.10 0.06
F0 (CV) 0.21 0.13 0.19 0.25 0.10 0.37 * 0.20 −0.03 −0.10 0.19
GNE (Q2) −0.16 −0.17 −0.28 −0.18 −0.16 −0.29 −0.25 0.05 −0.19 −0.19
GNE (SD) −0.05 −0.09 −0.25 0.09 −0.20 0.08 0.14 −0.05 0.12 0.15
TEO (CV) 0.05 0.04 0.27 −0.17 0.15 −0.05 −0.22 −0.07 −0.25 −0.01
NNE (Q2) 0.24 0.28 0.18 0.20 0.21 −0.04 0.21 0.17 0.16 0.20
NNE (SD) −0.27 −0.05 −0.19 −0.16 0.11 −0.23 −0.12 0.16 −0.22 ** 0.05
*—
p
- alue o Spea man’s co ela ion coe icien <0.05; **—
p
- alue o Spea man’s co ela ion
coe icien <0.01.
The esul s o Spea man’s pa ial co ela ion be ween he change o baseline acous ic ea u es
(
∆
) and he change in clinical da a (
∆
) can be seen in Table 4. Rega ding he change o UPDRSIII,
i nega i ely co ela ed wi h he change o mic ope u ba ions in equency o oice (PPQ, owel /i/),
ape iodici y (FLUF, owels /e/, /o/), emo o ongue (F1 (CV), owels /a/, /u/), emo o jaw (F2
(CV), owel /e/), i egula pi ch luc ua ions (F0 (CV), owels /a/, /u/), and a ia ion o acous ic
noise (NNE (SD), owel /i/). Signi ican posi i e co ela ions we e iden i ied wi h he change o lips
emo (F3 (CV), owel /a/), acous ic noise (ER (Q2), owel /a/), and a ia ion o oice quali y (GNE
(SD), owel /e/).
In he case o UPDRSIV, we iden i ied se en signi ican posi i e co ela ions wi h he change
o mic ope u ba ions in equency o oice (PPQ, owel /e/), emo o jaw (F2 (CV), owel /a/),
i egula ampli ude luc ua ions (TEO (CV), owels /a/, /u/), and acous ic noise (NNE (Q2), owels
/o/, /u/). The change in UPDRS IV signi ican ly nega i ely co ela ed wi h he change o acous ic
noise (ER (Q2), owel /u/), and i s a ia ion (ER (CV), owel /e/).
Changes in RBDSQ signi ican ly nega i ely co ela ed wi h he change o mic ope u ba ions
in equency o oice (PPQ, owel /u/), mic ope u ba ions o i s in ensi y (APQ, owels /e/, /i/,
/u/), emo o lips (F3 (CV), owel /o/), acous ic noise (NNE (Q2), owel /e/), and i s a ia ion (ER
(CV), owel /e/). Posi i e co ela ions we e iden i ied wi h he change in oice quali y (HNR (Q2),
all owels) and i s a iabili y (HNR (SD), owels /e/, /o/, /u/). The simila esul s can be obse ed
when assessing he quali y by GNE ( owel /e/).
Appl. Sci. 2018,8, 2339 16 o 18
11.
El ma ko a, N.; Gajdos, M.; M acko a, M.; Mekyska, J.; Mikl, M.; Rek o o a, I. Impac o Pa kinson’s disease
and le odopa on es ing s a e unc ional connec i i y ela ed o speech p osody con ol. Pa kinsonism Rela .
Diso d. 2016,22 (Suppl. 1), S52–S55. [C ossRe ] [PubMed]
12.
Gómez-Vilda, P.; Mekyska, J.; Fe ández, J.M.; Palacios-Alonso, D.; Gómez-Rodella , A.; Rodella -Bia ge, V.;
Galaz, Z.; Smekal, D.; Rek o o a, I.; Eliaso a, I.; e al. Pa kinson Disease De ec ion om Speech A icula ion
Neu omechanics. F on . Neu oin o m. 2017,11, 56. [C ossRe ] [PubMed]
13.
Li ani-Sil a, C.; Mou ão, L.F.; Gobbi, L.T.B. Dysa h ia and Quali y o Li e in neu ologically heal hy elde ly
and pa ien s wi h Pa kinson’s disease. CoDAS 2015,27, 248–254. [C ossRe ] [PubMed]
14.
Ho, A.K.; Iansek, R.; Ma igliani, C.; B adshaw, J.L.; Ga es, S. Speech Impai men in a La ge Sample o
Pa ien s wi h Pa kinson’s disease. J. Beha . Neu ol. 1999,11, 131–137. [C ossRe ]
15.
Tsanas, A.; Li le, M.; McSha y, P.; Ramig, L. Accu a e elemoni o ing o Pa kinson’s Disease p og ession by
nonin asi e speech es s. IEEE T ans. Bio-Med. Eng. 2010,57, 884–893. [C ossRe ] [PubMed]
16.
Ha el, B.T.; Cannizza o, M.S.; Cohen, H.; Reilly, N.; Snyde , P.J. Acous ic cha ac e is ics o Pa kinsonian
speech: A po en ial bioma ke o ea ly disease p og ession and ea men . J. Neu olinguis .
2004
,17, 439–453.
[C ossRe ]
17.
Rusz, J.; Cmejla, R.; Ruzicko a, H.; Klempi , J.; Maje o a, V.; Picmauso a, J.; Ro h, J.; Ruzicka, E.
E alua ion o speech impai men in ea ly s ages o Pa kinson’s disease: A p ospec i e s udy wi h he
ole o pha maco he apy. J. Neu al T ansm. 2013,120, 319–329. [C ossRe ] [PubMed]
18.
Skodda, S.; G önhei , W.; Schlegel, U.; Südmeye , M.; Schni zle , A.; Woj ecki, L. E ec o sub halamic
s imula ion on oice and speech in Pa kinson’s disease: Fo he be e o wo se? F on . Neu ol.
2014
,4, 218.
[C ossRe ] [PubMed]
19.
O ozco-A oya e, J.R.; Hönig, F.; A ias-Londoño, J.D.; Va gas-Bonilla, J.F.; Daq ouq, K.; Skodda, S.; Rusz, J.;
Nö h, E. Au oma ic de ec ion o Pa kinson’s disease in unning speech spoken in h ee di e en languages.
J. Acous . Soc. Am. 2016,139, 481–500. [C ossRe ] [PubMed]
20.
Mekyska, J.; Smekal, Z.; Galaz, Z.; Mzou ek, Z.; Rek o o a, I.; Faundez-Zanuy, M.; López-de Ipiña, K.
Pe cep ual ea u es as ma ke s o Pa kinson’s Disease: he issue o clinical in e p e abili y. In Recen Ad ances
in Nonlinea Speech P ocessing; Sp inge : New Yo k, NY, USA, 2016; pp. 83–91.
21.
Mekyska, J.; Galaz, Z.; Mzou ek, Z.; Smekal, Z.; Rek o o a, I.; Eliaso a, I.; Kos alo a, M.; M acko a, M.;
Be anko a, D.; Faundez-Zanuy, M.; e al. Assessing p og ess o Pa kinson’s disease using acous ic analysis
o phona ion. In P oceedings o he 2015 4 h In e na ional Wo k Con e ence on Bioinspi ed In elligence
(IWOBI), San Sebas ian, Spain, 10–12 June 2015; pp. 111–118.
22.
A o a, S.; Venka a aman, V.; Zhan, A.; Donohue, S.; Biglan, K.; Do sey, E.; Li le, M. De ec ing and moni o ing
he symp oms o Pa kinson’s disease using sma phones: A pilo s udy. Pa kinsonism Rela . Diso d.
2015
,
21, 650–653. [C ossRe ] [PubMed]
23.
Villa-Cañas, T.; O ozco-A oya e, J.; Va gas-Bonilla, J.; A ias-Londoño, J. Modula ion spec a o au oma ic
de ec ion o Pa kinson’s disease. In P oceedings o he 2014 XIX Symposium on Image, Signal P ocessing
and A i icial Vision (STSIVA), A menia, Colombia, 17–19 Sep embe 2014; pp. 1–5.
24.
Hazan, H.; Hilu, D.; Mane i z, L.; Ramig, L.O.; Sapi , S. Ea ly diagnosis o Pa kinson’s disease ia machine
lea ning on speech da a. In P oceedings o he 2012 IEEE 27 h Con en ion o Elec ical & Elec onics
Enginee s in Is ael (IEEEI), Eila , Is ael, 14–17 No embe 2012; pp. 1–4.
25.
Eskide e, Ö.; E a¸s, F.; Hanilçi, C. A compa ison o eg ession me hods o emo e acking o Pa kinson’s
disease p og ession. Expe Sys . Appl. 2012,39, 5523–5528. [C ossRe ]
26.
Cas elli, M.; Vanneschi, L.; Sil a, S. P edic ion o he Uni ied Pa kinson’s Disease Ra ing Scale assessmen
using a gene ic p og amming sys em wi h geome ic seman ic gene ic ope a o s. Expe Sys . Appl.
2014
,
41, 4608–4616. [C ossRe ]
27.
Na anjo, L.; Pé ez, C.J.; Ma ín, J. Add essing oice eco ding eplica ions o acking Pa kinson’s disease
p og ession. Med. Biol. Eng. Compu . 2017,55, 365–373. [C ossRe ] [PubMed]
28.
Skodda, S.; Rinsche, H.; Schlegel, U. P og ession o dysp osody in Pa kinson’s disease o e ime—A
longi udinal s udy. Mo . Diso d. 2009,24, 716–722. [C ossRe ] [PubMed]
29.
Skodda, S.; Flasskamp, A.; Schlegel, U. Ins abili y o syllable epe i ion as a ma ke o disease p og ession in
Pa kinson’s disease: A longi udinal s udy. Mo . Diso d. 2011,26, 59–64. [C ossRe ] [PubMed]
30.
Skodda, S.; G önhei , W.; Schlegel, U. Impai men o owel a icula ion as a possible ma ke o disease
p og ession in Pa kinson’s disease. PLoS ONE 2012,7, e32132. [C ossRe ] [PubMed]

Appl. Sci. 2018,8, 2339 17 o 18
31.
Skodda, S.; G onhei , W.; Mancinelli, N.; Schlegel, U. P og ession o Voice and Speech Impai men in he
Cou se o Pa kinson’s Disease: A Longi udinal S udy. Pa kinson’s Dis.
2013
,2013, 389195. [C ossRe ]
[PubMed]
32.
Lee, J.Y.; Kim, J.W.; Lee, W.Y.; Kim, J.M.; Ahn, T.B.; Kim, H.J.; Cho, J.; Jeon, B.S. Daily dose o dopamine gic
medica ions in Pa kinson’s disease: clinical co ela es and a pos e io i equa ion. Neu ol. Asia
2010
,
15, 137–143.
33.
S iasny-Kols e , K.; Maye , G.; Scha e , S.; Mulle , J.C.; Heinzel-Gu enb unne , M.; Oe el, W.H. The REM
sleep beha io diso de sc eening ques ionnai e—A new diagnos ic ins umen . Mo . Diso d.
2007
,
22, 2386–2393. [C ossRe ] [PubMed]
34.
Mol, H. Loss ee Twin-Tube Resona o and he Vowel T iangle o Hellwag. J. Acous . Soc. Am.
1965
,37, 1186.
[C ossRe ]
35.
O ozco-A oya e, J.R.; Hönig, F.; A ias-Londoño, J.D.; Va gas-Bonilla, J.; Skodda, S.; Rusz, J.; Nö h, E.
Au oma ic de ec ion o Pa kinson’s disease om wo ds u e ed in h ee di e en languages. In P oceedings
o he Fi een h Annual Con e ence o he In e na ional Speech Communica ion Associa ion, Singapo e,
14–18 Sep embe 2014; pp. 1573–1577.
36.
Rusz, J.; Cmejla, R.; Tykalo a, T.; Ruzicko a, H.; Klempi , J.; Maje o a, V.; Picmauso a, J.; Ro h, J.; Ruzicka, E.
Imp ecise owel a icula ion as a po en ial ea ly ma ke o Pa kinson’s disease: E ec o speaking ask.
J. Acous . Soc. Am. 2013,134, 2171–2181. [C ossRe ] [PubMed]
37. Rusz, J.; Cmejla, R.; Ruzicko a, H.; Ruzicka, E. Quan i a i e acous ic measu emen s o cha ac e iza ion o
speech and oice diso de s in ea ly un ea ed Pa kinson’s disease. J. Acous . Soc. Am.
2011
,129, 350–367.
[C ossRe ] [PubMed]
38. Ti ze, I.R. P inciples o Voice P oduc ion; P en ice Hall: Englewood Cli s, NJ, USA, 1994.
39.
Ha a , P.; Alonso-He nandezy, J.B.; Mekyska, J.; Galaz, Z.; Bu ge , R.; Smekal, Z. Voice Pa hology De ec ion
Using Deep Lea ning: A P elimina y S udy. In P oceedings o he 2017 In e na ional Con e ence and
Wo kshop on Bioinspi ed In elligence (IWOBI), Funchal, Po ugal, 10–12 July 2017; pp. 45–48.
40.
Ha a , P.; Galaz, Z.; Alonso-He nandez, J.B.; Mekyska, J.; Bu ge , R.; Smekal, Z. Towa ds obus oice
pa hology de ec ion. Neu al Compu . Appl. 2018. [C ossRe ]
41.
Kos alo a, M.; M acko a, M.; Ma ecek, R.; Be anko a, D.; Eliaso a, I.; Janouso a, E.; Roubicko a, J.;
Bedna ik, J.; Rek o o a, I. The 3F Tes Dysa h ic P o ile—No ma i e Speach Values in Czech. Ceska Slo enska
Neu ologie Neu ochi u gie 2013,76, 614–618.
42. Boe sma, P.; Weenink, D. P aa , a sys em o doing phone ics by compu e . Glo In . 2002,5, 341–345.
43.
Mekyska, J.; Janouso a, E.; Gomez-Vilda, P.; Smekal, Z.; Rek o o a, I.; Eliaso a, I.; Kos alo a, M.;
M acko a, M.; Alonso-He nandez, J.B.; Faundez-Zanuy, M.; e al. Robus and complex app oach o
pa hological speech signal analysis. Neu ocompu ing 2015,167, 94–111. [C ossRe ]
44.
Rek o o a, I.; Mekyska, J.; Janouso a, E.; Kos alo a, M.; Eliaso a, I.; M acko a, M.; Be anko a, D.;
Necaso a, T.; Smekal, Z.; Ma ecek, R. Speech p osody impai men p edic s cogni i e decline in Pa kinson’s
disease. Pa kinsonism Rela . Diso d. 2016,29, 90–95. [C ossRe ] [PubMed]
45.
A ias-Ve ga a, T.; Vásquez-Co ea, J.C.; O ozco-A oya e, J.R. Pa kinson’s Disease and Aging: Analysis o
Thei E ec in Phona ion and A icula ion o Speech. Cogn. Compu . 2017,9, 731–748. [C ossRe ]
46.
Rusz, J.; Tykalo a, T.; Klempi , J.; Cmejla, R.; Ruzicka, E. E ec s o dopamine gic eplacemen he apy
on mo o speech diso de s in Pa kinson’s disease: Longi udinal ollow-up s udy on p e iously un ea ed
pa ien s. J. Neu al T ansm. 2016,123, 379–387. [C ossRe ] [PubMed]
47.
Chen, T.; Gues in, C. Xgboos : A scalable ee boos ing sys em. In P oceedings o he 22nd ACM
SIGKDD In e na ional Con e ence on Knowledge Disco e y and Da a Mining, San F ancisco, CA, USA,
13–17 Augus 2016; pp. 785–794.
48.
B eiman, L.; F iedman, J.H.; Olshen, R.A.; S one, C.J. Classi ica ion and Reg ession T ees; Wadswo h and
B ooks: Mon e ey, CA, USA, 1984.
49.
To lay, L.; Pe one-Be olo i, M.; Thomas, E.; Baciu, M. Machine lea ning—XGBoos analysis o language
ne wo ks o classi y pa ien s wi h epilepsy. B ain In o m. 2017,4, 159. [C ossRe ] [PubMed]
50.
Chen, Y.; Wang, X.; Jung, Y.; Abedi, V.; Zand, R.; Bikak, M.; Adibuzzaman, M. Classi ica ion o sho single
lead elec oca diog ams (ECGs) o a ial ib illa ion de ec ion using piecewise linea spline and XGBoos .
Physiol. Meas. 2018,39, 104006. [C ossRe ] [PubMed]
Appl. Sci. 2018,8, 2339 18 o 18
51.
Zhong, J.; Sun, Y.; Peng, W.; Xie, M.; Yang, J.; Tang, X. XGBFEMF: An XGBoos -based F amewo k o
Essen ial P o ein P edic ion. IEEE T ans. NanoBiosci. 2018,17, 243–250. [C ossRe ] [PubMed]
52.
Palme , J.L.; Coa s, M.A.; Roe, C.M.; Hanko, S.M.; Xiong, C.; Mo is, J.C. Uni ied Pa kinson’s Disease
Ra ing Scale-Mo o Exam: In e - a e eliabili y o ad anced p ac ice nu se and neu ologis assessmen s.
J. Ad . Nu s. 2010,66, 1382–1387. [C ossRe ] [PubMed]
53.
Baggio, J.A.O.; Cu a elli, M.B.; Rod igues, G.R.; Tumas, V. Validi y o he B azilian e sion o he eezing o
gai ques ionnai e. A qui os de Neu o-Psiquia ia 2012,70, 599–603. [C ossRe ] [PubMed]
54.
San os, D.G.; Macías, M.A. In e - a e a iabili y in mo o unc ion assessmen in Pa kinson’s disease
be ween expe s in mo emen diso de s and nu ses specialising in PD managemen . Neu ologia
2017
.
[C ossRe ]
55.
Janko ic, J. Pa kinson’s disease: Clinical ea u es and diagnosis. J. Neu ol. Neu osu g. Psychia y
2008
,
79, 368–376. [C ossRe ] [PubMed]
56.
Mekyska, J.; Galaz, Z.; Kiska, T.; Z oncak, V.; Mucha, J.; Smekal, Z.; Eliaso a, I.; Kos alo a, M.; M acko a, M.;
Fiedo o a, D.; e al. Quan i a i e Analysis o Rela ionship Be ween Hypokine ic Dysa h ia and he F eezing
o Gai in Pa kinson’s Disease. Cogn. Compu . 2018. [C ossRe ]
57.
Be g, D.; Pos uma, R.B.; Adle , C.H.; Bloem, B.R.; Chan, P.; Dubois, B.; Gasse , T.; Goe z, C.G.; Halliday, G.;
Joseph, L.; e al. MDS esea ch c i e ia o p od omal Pa kinson’s disease. Mo . Dis.
2015
,30, 1600–1611.
[C ossRe ] [PubMed]
c
2018 by he au ho s. Licensee MDPI, Basel, Swi ze land. This a icle is an open access
a icle dis ibu ed unde he e ms and condi ions o he C ea i e Commons A ibu ion
(CC BY) license (h p://c ea i ecommons.o g/licenses/by/4.0/).