From the bony structuto soft tissues of the

Louis-Jean Boë 1, Denis Autesserre 2, Laurent Cavazzana 3, J1 ICP Université Stendhal, INPG, CNR

2 LPL, CNRS, Aix-en-Provence, France, 34 Musée de l’Homme, Paris, 5 Polycliniq

E-mail: boe@icp.inpg.fr, denis.autesserre@lpl.univ-aix.fr, laurent@cavaz

ABSTRACT

This work is part of a project aiming at defining possible speech prerequisites in the geometry, the musculature and the control of the vocal tract. In this paper we intend to reconstruct anatomical and geometrical landmarks of the vocal tract from the bony structure of modern man. Thereafter, we will try to extrapolate our procedure to fossilized skull and mandible. We constituted a head and neck xeroradiographic database for 10 speakers. Using bony references and positional angles we intend to predict two dimensions: palatal distance and larynx height. These dimensions can be used to quantify growth differences (from birth to adulthood) and gender differences, as well as differences between species. They are important in that they lead to the prediction of maximal acoustic space of the vocal tract. Then we evaluated the range of accuracy of these predictions. This work has been carried out by a multidisciplinary team comprised of specialists in radiology, anthropology, informatics and speech.

1. INTRODUCTION

Mankind acquired capability to speak as a result of anatomical and cognitive evolution. This work is part of a project aiming at defining possible speech prerequisites in the geometry, the musculature, and the control of the vocal tract [1-2]. Do apes have a vocal tract disability or a cognitive incapability or both? If vocal tract configuration plays an important role for speech emergence, can the study of skulls and mandibles of fossils enlighten us about speech potentialities of our predecessors? In order to answer these questions one should be able to reconstruct geometrical limits of vocal tract from bony architecture (figure 1). Using bony references and positional angles, we intend to predict palatal distance and laryngeal height. These two dimensions are important in that they reflect differences of sex, growth, stage of evolution [3], and lead to the prediction of maximal acoustic space of the vocal tract. We evaluated the accuracy of this kind of reconstruction using head and neck xeroradiographic images. This kind of images permit to visualize the bony structure of the head and neck as well as soft tissues of the vocal tract. Thereafter, we will try to extrapolate our procedure to fossilized skull and mandible.

Fia

Vocalligammanusourcetissuearchit

Figure

re of the head vocal tract

ean-Louis Heim 4,, Fleur Letellier-Willemin 5 S, Grenoble, France ENSIMAG, Grenoble ue de l’Essonne, Evry zana.com, heim@mnhn.fr, f.letellier.willemin@free.fr

gure 1. Reconstruction of the vocal tract from the bony rchitecture (anatomical drawing from Pernkopf [4]).

2. THE BONY STRUCTURES AND SOFT TISSUES

tract is essentially composed of muscles, membranes, ents, fibrous lamella and adipose tissues (figure 2). In als dealing with speech production, the laryngeal and the vocal tract are described in terms of soft

s involved without relating them to the bony ecture.

2. Apart from teeth vocal tract is constituted of soft tissues (Sophie Jacopin's original drawing).

The cervical vertebra (C1-C7) and hyoid bone which are visible on RX are only referred to when it comes to describing the position of the larynx. But hyoid bone is unique in that it is not directly attached to the any other bone in the skeleton. Upper incisors constitute the only visible bony landmark for the upper part of the vocal tract as are the lower incisors for mandible position. Goldstein’s thesis [5] exemplified for the first time the possibilities to predict vocal tract dimensions from bony landmarks using data gathered from medical literature. Her model integrated data concerning influence of sex and growth on the vocal tract.

3. SELECTED LANDMARKS

Using magnetic resonance imaging (MRI) on modern human subjects, Honda and Tiede [6] showed that larynx position can be predicted from biometric measures of the cranio-mandibular geometry. Their basic idea was that the facial geometry of humans (and that of primates) and larynx position are correlated. They used three distances, palatal distance (PD), laryngeal height (LH), and oral cavity height (OCH) as follows: PD is the distance between the anterior nasal spine (ANS) and the posterior nasal wall (PNW). PNW is defined as the intersection point of a standard palatal line (specified by ANS and the posterior nasal spine PNS) and the posterior nasopharyngeal wall. LH is the distance from the arytenoid apex to the palatal line. LH, therefore, represents the pharyngeal cavity length and the vertical position of the larynx. OCH is defined as the distance from the menton to the palatal line (figure 3). The results indicate that when LH and OCH are normalized by PD, using a laryngeal height index (LHI = LH/PD) and oral cavity index (OCI = OCH/PD), the two indices exhibit a high degree of correlation. These two indices can be used, as articulatory parameters, to quantify growth differences (from birth to adulthood) and gender differences, as well as differences between species (homo sapiens, Neandertal, and apes) [3].

Figure 3. (1) anterior nasal spine, (2) posterior nasal spine,

(3) nasopharyngeal wall, (4) menton (5) apex of the arytenoid tissue.

The indices proposed by Honda and Tiede have been adapted to be used in this work. With speaking subjects jaw position can be variable. NPW is not always easy to determine by means of ANS and PNS. Therefore we prefer the most anterior part of the atlas which is a more reliable landmark. We prefer to localize glottis, instead of apex of

arytendimenin anforens

Atlala

Gonth

Incide

Infrpr

Manth

Menpl

Nasthpl

Pogsa

Prospr

Sellcrof

For tharchitfollow

Cerval

Corcelo

Gloofon

Hyora

JawLips

anPhaTon

th

oid tissue. For our prediction of the vocal tract sions we have selected the following landmarks used thropology [7, 8], radiology [5], orthodontics [9], ic analysis of the skull [10] (figure 4): s (the most anterior point of the tubercle of the): ndmark of the first cervical. ion: the most inferior, posterior and lateral point of e external angle of mandible. sors (upper and lower): landmarks for the limitation of the anterior part of the vocal tract. adental: the point of the superior alveolar process that oject most anteriorly in the midline of the maxilla. dibular plane: a tangential plane to the lower part of

e mandible. ton: the most inferior point of the chin in the sagittal

ane (sometimes confused with gnathion). ion: the depression at the root of the nose indicating e position of the frontal suture in the midsagittal ane. onion: the most anterior point of the chin in the gittal plane. thion: the point of the inferior alveolar process that oject most anteriorly in the midline of the maxilla a turcica (center of the): a transverse depression ossing the midline on the superior surface of the body the sphenoid bone.

Figure 4. Anthropological landmarks.

e purpose of reconstruction of soft tissues from bony ecture, we added speech related landmarks as s: ical vertebra (center of the seven C1-C7): landmarks

lowing to locate the position of the larynx. poreal line: tangential line to the anterior part of rvical vertebra on which the pharyngeal wall is cated. ttis: landmark for the delimitation of the inferior part the vocal tract. In fact, the glottis is not identifiable RX, only laryngeal ventricles are visible. id bone (the most anterior point of the tubercle): diographic landmark related to larynx position. angle: an estimation of the jaw opening. (the most anterior point of the upper and lower lips): terior extremity of the vocal tract. ryngeal wall: the most posterior part of the pharynx. gue shape: the contour drawn from radix to apex in e midsagittal plane.

4. XERORADIOGRAPHIC DATABASE

Figure 5 . Xeroradiograpy of a 10 year old child

Xeroradiography was well suited to visualize bony and cartilaginous elements. At the same time, soft tissue contours were well defined (figure 5). It should be noted particularly that hyoid bone, pharyngeal and laryngeal regions, as well as lip contours were clearly shown.

Xeroradiography was replaced by scanners and later by MRI. A xeroradiographic database was constituted by Denis Autesserre in 1978, thanks to Professor Chevrot and Dr. Sarrat, from the Service Central de Radiologie Adulte of the CHU de la Timone, Marseille. It consisted of images of 15 speakers pronouncing French vowels and consonants. A grid was used to come up with the scale. The database contained a total of 89 images of high quality [11]. Xeroradiographies were digitally scanned at ICP in 2002.

Autesserre acquired landmarks and contours using Craniomat software.

5. SOFTWARE

We developed Craniomat, a specialized software for radiographic measurements. It was conceived and realized with the help of some students [12], tested and improved thanks to the help of radiologists and anthropologists and speech specialists. Using this software, experts can make measurements on photographic, radiographic, and MRI images. Depending on the nature of images available, the software shows the experts a list of landmarks which is modifiable. Some landmarks are automatically determined (center of the sella turcica from three points, vertex and opisthocranion from a spline), others are semi-automatically acquired (gnation, menton, pogonion).

Contours (hard palate, soft palate, mandible, tongue, corporeal line, pharyngeal wall) are smoothed by cubic spline interpolations. A zoom function permitted to greatly increase the accuracy and the reliability of all the points (a standard deviation of 0.2 mm for repeated measurements of the same landmark for a given expert). The software produce two files containing the results as follows: the first one contains coordinates of landmarks, distances, angles and indices, the second one, an image file, shows how the expert made the measurements (figure 6). In total the software can produce 41 landmark coordinates, 40

distanplanes

6

Figureof hyoiand loinciso

In thepositibony througvisiblmandinfradintersangle

To locfact lathe m

ces between landmarks, 17 angles between reference , and 5 indices (ratios between distances).

Figure 6. Landmarks and contours (DA expert).

. PREDICTION AND EVALUATION OF SOFT TISSUES LANDMARK

7. Prediction of the position of glottis (Gl), of the position d bone (Hy), of the pharyngeal wall, of extremities of upper wer lips (uL, lL) from nasion (Na), sella turcica (ST), upper r (uI), atlas (At), infradental (iD), gonion (Go), pogonion

(Po), menton (Me) landmarks.

first place, we predicted glottis and hyoid bone ons, pharyngeal wall and external lip extremities from landmarks. We selected as reference: (1) the line h nasion – and sella turcica NSL, which is clearly

e on all RX images; (2) an x-y axis constituted by ibular plane and an orthogonal line through ental. To locate the pharyngeal wall we calculated (1) ection XPh of its extrapolation with NST line; (2) αPh between these two lines.

ate hyoid bone (its most anterior part) and glottis (in ryngeal ventricles), we used information concerning

andible size which depends on sex, age and specie as

predicting factors: (1) the height of the mandibular symphysis as measured from the highest and most prominent point on the lower alveolar arch (infradental) to the lowest point on the symphysis (menton); (2) distance between gonion and pogonion; (3) jaw opening angle (figure 7).

We calculated all predictions with multilinear regressions from 10 images corresponding to the vowel /a/ pronounced by 10 speakers (1 woman and 9 men). The results indicated 7,0 mm as mean differences between data and predictions for the position of XPh, and 5.6° for αPh angle, 7,6 mm for the vertical position of the glottis (figure 8), and 3,3 mm for its horizontal position. The three latter estimations can be improved by taking into account differences in head inclination. The most anterior part of the atlas can be selected as center of rotation. Lip extremities are less predictable, and this is not surprising: their dimensions are not correlated with bony structure. It is possible to use standards presented for the thickness of soft tissues at different parts of the face recommended for specialists in craniofacial forensic identification [13-14]. For the oral dimension of the vocal tract we used upper incisor (or prosthion) as anterior landmark.

Figure 8. Prediction of pharyngeal wall and glottis position.

5. CONCLUSION

The first step permitted us to show that it is possible to predict the positions of soft tissues given only the bony structure. Further steps listed below remain to be taken in order to corroborate our approach: (1) enlarging our training database: images corresponding

to vowels other than /a/; (2) validating the predictions for data other than the

training database: about fifty subjects have already been X-rayed).

(3) applying the predictions to fossils and verify that such an application brings out plausible results: we have in our possession radiographs of skull fossils from Musée de l’Homme, Paris.

CraniothanksBrouh(radioland imDiversContrôhumai(manathe Orprojec(2) OrprimatDemolof Thefunded

[1] Jcsd

[2] JL2

[3] LNh

[4] EABC

[5] UtC

[6] KboP

[7] DU

[8] JfA

[9] GRA

[10] Ms

[11] DDlI1

[12] PR

[13] G“M

[14] Jo

ACKNOWLEDGEMENTS

mat, the software for acquisition of data, was developped to Dimitry Obolensky, Philippe Jaunais, Christophe de la e, Thomas Mazur, and Servan Riaud. Roger Lichtenberg ogist) Patricia Soto-Heim (anthropologist) helped us to test prove this software. Thanks a lot to Pierre Badin and

on Mzemba for their assistance. This research is a part of (1) le oro-facial dans la communication chez les primates

ns et non humains : Neandertal, le singe et l’Homme project ged by Jean-Luc Schwartz, ICP, since 2001), being part of igine de l’Homme du Langage et des Langues (OHLL)

t (managed by Jean-Marie Hombert), funded by the CNRS; ofacial control in communication in human and non human es, EUROCORES program (principal investigator, Didier in, Phonology Laboratory, Brussels, since 2002), being part Origin of Man, Language and Languages (OMLL) project by European Community.

REFERENCES .L. Schwartz, “Contrôle oro-facial dans la communication hez les primates humains et non-humains: Néandertal, le inge et l’Homme,” In Origine de l’Homme, du Langage et es Langues, J.M. Hombert Ed. , 191-198, 2002. .M. Hombert, “Origine de l’Homme, du Langage et des angues,” Rapports scientifiques de fin de deuxième année 001-2002. Paris: CNRS, 2002. .J. Boë, J.-L Heim., K. Honda, S. Maeda, “The potential eandertal vowel space was as large as that of modern umans,” J. Phonetics, 30, 465-484, 2002. . Pernkopf, Atlas of Topographical an Applied Human natomy, Volume I: Head and Neck, Ed. By H. Ferner, altimore: Urban & Schwarrzenberg, W.B. Saunders omlpany, 1980. .G. Goldstein, An articulatory model for the vocal tract of

he growing children, Thesis of Doctor of Science, MIT, ambridge, Massachusett, 1980. . Honda, M.K. Tiede, “An MRI study on the relationship etween oral cavity shape and larynx position,” Proceedings f the 5th International Conference on Spoken Language rocessing, 2, 437-440, 1998. .R. Brothwell, Digging up bones, Ithaca: Cornell niversity Press, 1994.

.E. Buikstra, D.H. Ublaker, Sandards for data collection rom human skeletal remains, Arkansas Fayetteville: rcheological Survey, series NO. 44, 1994. .B. Scheideman, W.H. Bell, H.L. Legan, R.A. Finn, J.S. eisch, “Cephalometric analysis of dentofacials normals,” m. J. Orthodontics, 78, 4, 404-420, 1980. . Yasar Iscan, R.P. Helmer, Ed., Forensic analysis of the

kull, New York: Wiley-Liss. Inc., 1993. . Autesserre, M. Rossi, P. Sarrat, G. Giraud, R. Visquis, R. emange, L. Chevrot, “Contrôle radiologique de

'oropharynx, de l'hypopharynx et du larynx en phonation,” n Larynx et Parole, 47-73, Grenoble: Institut de Phonétique, 980. . Jaunais, Acquisition de données craniométriques, apport de stage, ENSERG, Grenoble, 2002. .V. Lebedinskaya, T.S. Balueva, E.V. Veselovskaya,

Principles of facial reconstruction,” In Forensic analysis, . Yasar Iscan and R.P. Helmer Ed., 183-198, 1993.

.N. Vignal, Les reconstitutions faciales assistées par rdinateur, Paris: Artcom, 1999.