552
|
Jou nal o Ana omy. 2022;241:552–564.wileyonlinelib a y.com/jou nal/joa
1 | INTRODUCTION
Th ee- dimensional (3D) compu e g aphics a e inc easingly
used o communica ing ana omical in o ma ion and knowledge.
Compu e ized, 3D ep esen a ions o ana omy can be isualized and
manipula ed dynamically and hus ep esen a aluable ool o ana -
omy lea ning, as well as o su gical planning and simula ion (Assa
& Pas e nak, 2008; Fang e al., 2020; Hemminge e al., 2005; Li
e al., 2017; Mu gi oyd e al., 2015; P eim & Saal eld, 2018; Sole
e al., 2014; S epan e al., 2017; T iepels e al., 2020; Yammine &
Viola o, 2015). Fo digi ally ende ed ana omical scenes o be isu-
ally ealis ic, no only geome y needs o be ep esen ed in adequa e
de ail bu ideally also colo a ibu es o issues and o gans should be
displayed as close o hei na u al appea ance as possible. A ecen
sys ema ic analysis o a ange o so wa e applica ions de o ed o
3D isualiza ion o human ana omy app aised he 3D models ea-
u ed in cu en applica ions wi h an a e age sco e o 3.04 o 5 in
he ealism dimension (Zil e schoon e al., 2019). This was a modes
sco e deno ing ha he ealism le el o mos 3D ana omical mod-
els is la gely limi ed o showing some de ail in isual/colo cla i y
Recei ed: 19 Oc obe 2021
|
Re ised: 18 Janua y 2022
|
Accep ed: 16 Feb ua y 2022
DOI: 10.1111/joa.13647
ORIGINAL ARTICLE
T ue- colo 3D ende ing o human ana omy using
su ace- guided colo sampling om cada e c yosec ion
image da a: A p ac ical app oach
Jon Ja su Azkue
This is an open access a icle unde he e ms o he C ea i e Commons A ibu ion-NonComme cial-NoDe i s License, which pe mi s use and dis ibu ion in
any medium, p o ided he o iginal wo k is p ope ly ci ed, he use is non-comme cial and no modi ica ions o adap a ions a e made.
© 2022 The Au ho . Jou nal o Ana omy published by John Wiley & Sons L d on behal o Ana omical Socie y.
Depa men o Neu osciences, School o
Medicine and Nu se y, Uni e si y o he
Basque Coun y, UPV/EHU, Leioa, Spain
Co espondence
Jon Ja su Azkue, Depa men o
Neu osciences, School o Medicine
and Nu se y, Uni e si y o he Basque
Coun y, UPV/EHU, Leioa, Spain.
Email: jonja [email protected]
Abs ac
Th ee- dimensional compu e g aphics a e inc easingly used o scien i ic isualiza ion
and o communica ing ana omical knowledge and da a. This s udy p esen s a p ac-
ical me hod o p oduce ue- colo 3D su ace endi ions o ana omical s uc u es.
The p ocedu e in ol es ex ac ing he su ace geome y o he s uc u e o in e es
om a s ack o cada e c yosec ion images, using he ex ac ed su ace as a p obe o
e ie e colo in o ma ion om c yosec ion da a, and mapping sampled colo s back
on o he su ace model o p oduce a ue- colo endi ion. O gans and body pa s can
be ende ed sepa a ely o in combina ion o c ea e cus om ana omical scenes. By ed-
i ing he su ace p obe, s uc u es o in e es can be ende ed as i hey had been p e-
iously dissec ed o p epa ed o ana omical demons a ion. The p ocedu e is highly
lexible and nondes uc i e, o e ing new oppo uni ies o p esen and communica e
ana omical in o ma ion and knowledge in a isually ealis ic manne . The echnical
p ocedu e is desc ibed, including eely a ailable open- sou ce so wa e ools in ol ed
in he p oduc ion p ocess, and examples o colo su ace ende ings o ana omical
s uc u es a e p o ided.
KEYWORDS
3D g aphics, ana omy educa ion, medical isualiza ion, su ace ende ing
|
553
AZKUE
and nonsimplis ic shapes, hence a he a om he highes sco e
co esponding o ealis ic models bo h in shape and isual de ails
(Zil e schoon e al., 2019).
The wo majo classes o app oaches o 3D isualiza ion, ha is,
su ace ende ing and olume ende ing, p ocess and display colo
di e en ly. In olume ende ing, a 2D p ojec ion is simula ed by
compu ing he abso p ion and emission o ligh ays cas h ough a
3D da ase composed o olume ic pixels o oxels gene a ed om
a s ack o sec ional image da a (D ebin e al., 1988; Kau man, 1996;
Le oy, 1988). G ay scale o colo da a con ained in oxels a e usually
mapped in o p ede e mined opaci y and colo alues using ans e
unc ions (D ebin e al., 1988; Ney e al., 1990). Since olume en-
de ings o medical isualiza ion commonly a ge ana omical g ay
scale da a such as hose om CT and MRI scans, issues and o gans
a e usually ep esen ed using a i icially c ea ed colo s. S ill, s ik-
ingly ealis ic endi ions o he human body may s ill be gene a ed
e en in he absence o colo in o ma ion in he sou ce image da a.
Fo ins ance, he cinema ic- ende ing app oach p oduces hype eal-
is ic endi ions ou o CT imaging da a by compu ing mul iple pa hs
o isible pho ons h ough he body issues (Engel, 2016; Glemse
e al., 2018; Paladini e al., 2015). In su ace ende ing, on he o he
hand, a s uc u e is ep esen ed only by i s su ace— mos com-
monly a polygon mesh o isosu ace— and made isible by showing
i s appea ance upon ex e nal illumina ion wi h a i ual ligh sou ce.
Al hough mos implemen a ions o su ace ende ing use a bi a ily
assigned solid colo s (Cline e al., 1987; Cook e al., 1983; He man &
Liu, 1979), e ices o aces o a polygon mesh can ha e associa ed
scala da a ca ying colo in o ma ion, and hus a su ace can also
be ende ed wi h a mul icolo appea ance. Fo example, in pho o-
g amme y, an ana omical specimen is pho og aphed om mul iple
iewpoin s and hen a 3D poin cloud o ma ching poin s is compu ed
ha e lec s he objec ’s 3D geome y and o which colo a ibu es
om he o iginal pho og aphs a e hen mapped o mimic he objec ’s
ex e nal isual aspec (Pe iceks e al., 2018; Schenk, 2000). In addi-
ion, a 3D polygon mesh can be w apped in a bi map image displaying
colo s and ex u es in o de o p o ide he ende ed s uc u e a li e-
like appea ance (P eim & Saal eld, 2018; Zil e schoon e al., 2017).
The use o cada e c yosec ion images as sou ce da a o 3D
econs uc ions allows o e ie e colo in o ma ion o issues
and o gans, p o iding an a enue o p oduce 3D endi ions o any
ana omical s uc u e wi h i s ue colo appea ance as ound in
he ozen cada e . Collec ions o se ial, high- esolu ion c yo-
sec ion images o human cada e specimens bo h o whole bod-
ies and sepa a e body pa s ha e been made a ailable, including
he Visible Human P ojec om he Na ional Lib a y o Medicine
(Acke man, 1998; Acke man e al., 2001; Spi ze e al., 1996), he
Chinese Visible Human (Zhang e al., 2003, 2006), he Visible
Ko ean Human (Kim e al., 2002; Pa k e al., 2005), and he Visible
Ea da ase s (Sø ensen e al., 2002). High- esolu ion mu ine c yo-
sec ion da a a e also a ailable (Roy e al., 2009; Wilson e al., 2008).
In he olume ende ing modali y, opaci y ans e unc ions can be
used o map ue colo p ope ies o he issues in o opaci y le els
in o de o manipula e he isibili y o speci ic o gans. This s a egy,
e med alpha blending o alpha ende ing (Kah s & Labadie, 2013),
allows o highligh speci ic issues based on hei colo p ope -
ies as well as o emo e unwan ed componen s om he scene,
o example, supe icial issues o he embedding gel (Ga gesha
e al., 2009, 2011; Kah s & Labadie, 2013). A p ac ical limi a ion o
his app oach is ha speci ic ans e unc ions mus be se in o de
o highligh di e en issues o o gans om each sou ce da ase . In
addi ion, issues wi h simila colo p ope ies a e di icul o single
ou based solely on opaci y ans e unc ions. This di icul y can be
o e come by ci cumsc ibing he ende ed scene o subse s o he
o iginal olume ic da a p e iously segmen ed om ei he he o ig-
inal c yosec ion image s ack (Heng e al., 2006) o om egis e ed
CT scan da a (Robb & Hanson, 2006). In e es ingly, o gans o body
pa s segmen ed om olume ic da a can be su ace ende ed in
ue colo by mapping colo a ibu es om he c yosec ion image
s ack back on o he ex ac ed su ace model. This app oach is ad-
an ageous in ha i educes bo h he s o age equi emen s and
he amoun o da a ha has o be ende ed as compa ed wi h he
o al numbe o oxels ha en e in o compu a ion in olume en-
de ing (Udupa e al., 1991). Endea o s in his di ec ion ha e been
epo ed in he scope o biomedical image isualiza ion. Robb and
Hanson (2006) p o ided a ew examples o endi ions o g oss seg-
men s o he Visible Human Male abdomen a e sec ioning in se -
e al planes. An analogous app oach was used o p oduce su ace
ende ings o he ana omy o he pel is econs uc ed om he
Visible Human P ojec da ase in he con ex o The Vesalius (TM)
P ojec (Venu i e al., 2004).
Despi e he po en iali ies o ealis ic colo ende ing o ana om-
ical isualiza ion, he me hods and ools in ol ed in p oducing ue-
colo su ace models ha e no ye been desc ibed o a e no a ailable
in he public domain (Dai e al., 2012; Robb & Hanson, 2006; Venu i
e al., 2004). The aim o his wo k was o demons a e a simple and
e sa ile p ocedu e o c ea e high- quali y, ue- colo su ace endi-
ions o human ana omical s uc u es by ex ac ing colo in o ma ion
om c yosec ion cada e image da a. Digi al models c ea ed using
his echnique can also be ende ed as i hey had been p e iously
dissec ed o p epa ed in a a ie y o ways o ana omical demon-
s a ion, and easily sa ed and combined. A se o open- sou ce digi al
ools a e p oposed, all o which a e a ailable in he public domain,
and examples o a a ie y o digi al dissec ions ha may be use ul
o p esen ing ana omical in o ma ion and knowledge a e p o ided.
2 | MATERIALS AND METHODS
The echnical p ocedu e in ol ed wo majo s eps (gene al wo k low
desc ibed in Figu e 1). The i s was c ea ing a su ace ep esen a-
ion o he a ge s uc u e om a s ack o cada e se ial c yosec ion
images, which was accomplished using semiau oma ic segmen a ion
and he ma ching cubes algo i hm o isosu ace ex ac ion. The sec-
ond s ep in ol ed e ie ing colo in o ma ion om he subse o
oxels ep esen ing he su ace laye o he a ge s uc u e, using
he su ace mesh gene a ed in he p e ious s ep as a p obe.
554
|
AZKUE
2.1 | C yosec ion image da a
High- esolu ion axial plane c yosec ion images o i e cada e s om
wo sepa a e collec ions we e used. Da ase s and a ge s uc u es
we e chosen wi h a c i e ion o con enience, in o de o illus a e
he gene al p ope ies o he me hod while p o iding examples o a
a ie y o di e en looking o gans and issues.
Axial sec ions o bo h he male and emale cada e s om he
Visible Human P ojec da abase (Na ional Lib a y o Medicine,
Be hesda, Ma yland; Acke man, 1998; Acke man e al., 2001;
Spi ze e al., 1996) had 1 mm and 0.33 mm spacing in he z- axis, e-
spec i ely, and images we e 2048 × 1216 pixels, whe e each pixel is
de ined by 24 bi s o colo (RGB, one by e each). C yosec ion images
o he o malin- p ese ed head o a 72- yea - old male dono om
he same da abase we e also used (0.147 mm in e als, wi h image
dimensions o 1056 × 1528 pixels and 24 bi s o colo ), whose blood
essels had been illed wi h a aldi e- F (Ra iu e al., 2003). In addi ion,
c yosec ion images o he Visible Head and Visible Male da ase s
om he Visible Ko ean Human collec ions (Kim e al., 2002; Pa k
e al., 2005) we e used (0.1 mm and 0.2 in e als and image dimen-
sions o 4368 × 2912 and 2468 × 1407 pixels, espec i ely).
2.2 | Segmen a ion and su ace pos p ocessing
Each se ies o c yosec ion images con aining a gi en s uc u e
o in e es was c opped o he co esponding a ge o gan using
GNU Image Manipula ion P og am (GIMP) so wa e, con e ed
FIGURE 1 Flow diag am o he su ace- guided colo sampling p ocedu e. C yosec ion images a e con e ed in o a olume ic da a ile,
which can hen be esliced in any o hogonal plane o segmen a ion. Segmen a ion o he submandibula gland (sb, delimi ed in ed colo )
om he Visible Human Male is shown. The ma ching cubes algo i hm gene a es a 3D polygon mesh ep esen ing he ex e nal su ace o
he segmen ed a ge objec . A his s age, a su ace mesh can be edi ed o p oduce a modi ied e sion o he a ge s uc u e ha will be
ende ed as ha ing been dissec ed o p epa ed. The colo sampling ope a ion e ie es colo in o ma ion (shades o b own) o hose oxels
ep esen ing he ex e nal bounda ies o he a ge s uc u e, using he su ace mesh ou pu ed by he p e ious s ep as a p obe ( ed), and
hen assigns colo a ibu es o he co esponding polygon e ices. A schema ic ep esen a ion o his ope a ion in a subg oup o oxels
is shown. Finally, he VTK ende e c ea es colo on polygon aces by in e pola ion o e ex colo s and ende s he su ace. b: Body o
mandible; c: Ca o id a e ies; d: Digas ic muscle; : Facial ein; h: Hyoglossus muscle; hb: Hyoid bone; j: In e nal jugula ein; m: Masse e
muscle; n: Deep ce ical lymph nodes; p: Pala opha yngeus and pha yngeal cons ic o muscles; pl: Pla ysma; p : Medial p e ygoid muscle;
sb: Submandibula gland; sg: S yloglossus muscle; sm: Submen al ein; s : S e nocleidomas oid muscle; : Tongue
|
555
AZKUE
TABLE 1 So wa e ools used o segmen a ion, colo sampling, and pos p ocessing o su ace models
Subp ocess So wa e ool Pla o m A ailabili y Websi e
C yosec ion image c opping GIMP MS Windows, MacOS, GNU Linux F ee, open sou ce h ps://www.gimp.o g/
Con e sion o 2D image s acks
in o olume ic iles
Slice 3D MS Windows, MacOS, GNU Linux F ee, open sou ce h ps://www.slice .o g/
3D segmen a ions ITK- SNAP MS Windows, MacOS, GNU Linux F ee, open sou ce h p://www.i ksn ap.o g
Edi ion o polygon meshes Blende MS Windows, MacOS, GNU Linux F ee, open sou ce h ps://www.blend e .o g/
Colo sampling, isualiza ion,
and ile expo a ion
The Visualiza ion
Toolki
MS Windows, MacOS, GNU Linux F ee, open sou ce h ps:// k.o g/
Tex u e baking MeshLab MS Windows, MacOS, GNU Linux F ee, open sou ce h ps://www.meshl ab.ne
FIGURE 2 Implemen a ion o he colo sampling ope a ion using VTK. Inpu da a impo a ion, colo sampling, ende ing, and ile sa ing s eps
a e shown, also indica ing he in ol ed il e s o unc ions in he VTK p ocessing pipeline. The kP obeFil e , co e o he sampling ope a ion,
akes he sou ce olume ic da a and he polygon mesh ex ac ed by segmen a ion as inpu s. The il e ou pu s a copy o he inpu mesh and
appends he ex ac ed colo da a om scala alues o polygon e ices. A sepa a e s ep con e s scala alues in o colo o isualiza ion.
A mappe con e s he polygon mesh in o a i ual model o su ace ende ing. The esul ing model can be s o ed as a ile. A py hon sc ip
ha pe o ms his ope a ion and ende s he esul ing su ace is p o ided in Appendix A
556
|
AZKUE
in o a single olume ic da a ile in Nea ly Raw Ras e Da a o -
ma (NRRD; h p:// eem.sou c e o ge.ne /n d/) using 3D Slice
so wa e ( . 4.10.2 28257; h ps://www.slice .o g/), and hen
impo ed o ITK- SNAP Medical Image Segmen a ion Tool so -
wa e ( 3.4.0; Yushke ich e al., 2006) o segmen a ion (Table 1).
Volume ic da a can hen be e- sliced in he x- , y- , and z- axes in ITK-
SNAP o gene a e sagi al, co onal, and ans e se plana iews o
con enien isualiza ion du ing segmen a ion. Segmen a ion e-
e s he e o he p ocess o singling ou he subse o oxels wi hin a
olume ic ile ha ep esen he s uc u e o in e es . This allows
o compu e a 3D su ace ep esen ing he ex e nal bounda ies o
he a ge s uc u e. All segmen a ions we e done he e semiau o-
ma ically using he buil - in ac i e con ou algo i hm in ITK- SNAP
(Kass e al., 1988), as desc ibed p e iously (Azkue, 2021b). Ac i e
con ou s a e compu e - gene a ed cu es ha p opaga e wi hin
he olume ic da a, h ough a se ies o i e a ions, un il hey adap
o he bounda ies o he a ge objec . ITK- SNAP hen uses he
ma ching cubes algo i hm (Lo ensen & Cline, 1987) o gene a e
a 3D polygon mesh ep esen ing he ex e nal su ace o he a -
ge objec . Polygon meshes we e expo ed as su ace 3D iles in
S anda d Tessella ion Language (STL) o ma and hen subjec ed o
Laplacian smoo hing using Blende 2.90.0 so wa e. This p oce-
du e emo es noise locally by smoo hing he posi ion o a gi en
e ex o he mesh based on in o ma ion abou i s immedia e
neighbo s, while p ese ing he gene al shape o he o iginal model
(So kine e al., 2004). He e, 20– 35 i e a ions and a lambda ac o
o 0.2– 0.3 we e used o smoo hing. Addi ional manual smoo hing
was applied as needed.
FIGURE 3 Success ul colo sampling is dependen on he accu acy o he su ace p obe. A su ace endi ion o he Visible Human Female
head is shown using co ec ly sized (100% scale) and dis o ed p obes. Inc easing o dec easing p obe size by 2%– 3% esul s in sampling
colo s om oxels ou side he body (embedding blue gel) o ep esen ing subcu aneous issues. A c oss- sec ional image shows he ela i e
posi ions o he h ee p obes o e e ence
|
557
AZKUE
The abo e p ocess ou pu s a 3D polygon mesh ha ep esen s
he ex e nal shape o he ana omical s uc u e ha will be ul ima ely
ende ed. In addi ion o he ob ious possibili y o ende ing he a -
ge s uc u e as a whole, a polygon mesh may be edi ed digi ally a
his s age by cu ing o disc e e po ions o he model in o de o
p oduce wha would be ende ed as a dissec ed specimen. All such
p epa a ions we e done he e by applying Boolean sub ac ion ope -
a ions using Blende .
2.3 | Su ace- guided colo sampling
Colo in o ma ion was e ie ed om hose oxels a posi ions de-
e mined by he spa ial coo dina es o he polygon mesh yielded by
he p eceding s ep. The sampling ope a ion was compu ed he e using
he Visualiza ion Toolki (VTK, 9.0.3), an open- sou ce so wa e sys-
em o 3D compu e g aphics and scien i ic isualiza ion (Sch oede
e al., 1998; h ps:// k.o g/). The VTK p o ides a a ie y o so- called
il e s o elemen s ha ecei e da a om o he componen s in a isu-
aliza ion pipeline, modi y hose in a a ie y o ways (e.g., ex ac , sub-
sample, in e pola e, me ge, o spli inpu da a), and hen ou pu he
modi ied da a o be handled by o he elemen s o subp ocesses. The
p esen app oach u ilized he kP obeFil e , a ool ha can e ie e
scala da a con ained in oxels, such as RGB da a, using a polygon
mesh as a p obe. RGB alues (1 by e each) e ie ed om sampled
oxels a e hen mapped back o he polygon mesh as colo a ibu es.
In o de o p oduce a inal su ace endi ion wi h he ex ac ed colo s,
he VTK isualiza ion pipeline assigns colo o polygon aces by in-
e pola ing e ex colo s. The da a p ocessing pipeline is shown in
Figu e 2, and a sc ip in Py hon language ha pe o ms he sampling
ope a ion and ende s he esul ing su ace is p o ided in Appendix A.
2.4 | Addi ional p ocessing
The abo e- desc ibed p ocedu e is su icien o p oduce a 3D endi-
ion o he a ge ana omical s uc u e wi h i s ue colo appea -
ance. O en, howe e , sa ing he model wi h he a ached colo
in o ma ion is also desi ed. The X3D ile o ma , an XML- based
o ma o ep esen ing 3D in o ma ion de eloped as an imp o ed
e sion o he Vi ual Reali y Modeling Language (VRML) was used
he e o expo p oduced models wi h he appended colo a ibu es
(Py hon sc ip also p o ided in Appendix A).
FIGURE 4 The le el o de ail o ende ed colo s and ex u es depends on he esolu ion o he sou ce c yosec ion images. Recons uc ions o
he ce ical spinal co d om he Visible Head om he Visible Ko ean da ase (le ) and he Visible Human Male om he NLM ( igh ) a e
shown, and he maximum wid h o he spinal co d in he co esponding o iginal images is indica ed. Do sal and en al ho ns a e eadily
iden i iable in he highe esolu ion model, and ela i ely blu ed in he lowe esolu ion one
FIGURE 5 Kidney in en i e y and a e edi ing he su ace p obe.
Su ace endi ion o he whole le kidney and a ached u e e om
he Visible Human Male a e shown, as well as i ually dissec ed
e sions o he kidney, u e e , and ascula supply a e sec ioning
in he axial and co onal planes. The mo phology and in e nal
o ganiza ion o he enal py amids a e clea ly disce nible. No e an
accesso y enal a e y en e ing he in e io pole
558
|
AZKUE
Op ionally he model can also be subjec ed o ex u e baking o
he ans e ing o colo in o ma ion o a bi map image, which allows
o handle geome y and colo da a sepa a ely and educe polygon
coun — and hus ile size— while p ese ing colo and ex u e qual-
i y. The o me was accomplished he e using he Pa ame iza ion and
Ve ex Colo o Tex u e il e s in MeshLab 2020.07 so wa e. The
esul ing model was expo ed in OBJ o ma , which also gene a ed
a bi map image ile in PNG o ma and an MTL o ma ile linking
he p eceding wo iles. Mesh simpli ica ion was ca ied ou using
Blende .
3 | RESULTS
T ue- colo su ace endi ions shown in Figu e 1 and Figu es 3– 8 il-
lus a e he main p ope ies o he me hod and p o ide examples o
a a ie y o di e en looking o gans and body pa s.
3.1 | Gene al cha ac e is ics o he me hod
The ende ed s uc u es exhibi shapes and colo ex u es as hey
a e ound in he cada e image da ase , and he e o e hey may
display nonidealized colo s o pa icula i ies ha a e usually no
ep esen ed in medical illus a ions, o example, colo imp in s
made by nea by s uc u es such as blood essels (Figu e 1) o bile
imp egna ion o he a su ounding he gall bladde (no shown).
In addi ion, possible colo modi ica ions by e ec o cada e ixa-
ion o p ocessing a e also cap u ed, such as pe mea ion o he
embedding gel (Figu es 3 and 7), eddish imp egna ion by a e ial
in usion o a aldi e (Figu e 6), o issue ab asion du ing sec ioning
(Figu e 6).
The le el o de ail o he ende ed colo s and ex u es is di-
ec ly ela ed o he esolu ion o he c yosec ion images and hus
o he amoun o in o ma ion a ailable om he sou ce da ase . An
example showing su ace ende ings o spinal co ds econs uc ed
FIGURE 6 The b ain as a whole and a e edi ing he su ace p obe. Su ace endi ions o he b ain om he isola ed head da ase in he
Visible Human image collec ion a e shown, bo h as a whole ( op le ) and ollowing sec ioning in he axial and co onal planes o exposu e o
he le insula co ex (bo om igh ) a e emo al o he o e lying on al, pa ie al, and empo al co ices. No e whi ish discolo a ion on he
pos e io sho gy us and long gy us o he insula due o ab asion du ing cada e sec ioning
|
559
AZKUE
om c yosec ions o di e en o iginal esolu ions is p o ided in
Figu e 4.
A key aspec o he echnical p ocedu e is ha success ul colo
sampling is c i ically dependen on accu a ely de ining he bounda -
ies o he a ge s uc u e by he su ace p obe. I is shown ha in-
c easing o educing he size o he su ace p obe by 2%– 3% esul s
in e ie ing colo da a om oxels ep esen ing he embedding blue
gel o subcu aneous issues, espec i ely (Figu e 3).
3.2 | Cap u ing ex e nal and in e nal colo ea u es
T ue- colo su ace endi ions o a numbe o ana omical s uc u es
we e p oduced, including a sali a y gland (Figu e 1), skin and sub-
cu aneous issue (Figu es 3 and 7), spinal co d (Figu e 4), he kidney
wi h a ached u e e and blood supply (Figu e 5), he b ain (Figu es 6
and 7), bone (Figu es 7 and 8), and muscle (Figu e 8).
Edi ing he su ace p obe o a gi en s uc u e p io o colo
sampling allowed he gene a ion o a ange o e sions o he
model ende ed as i dissec ed in a a ie y o possible ways. Fo
example, Figu e 5 shows su ace endi ions o he kidney exhibi -
ing he spa ial disposi ion o enal py amids a e c oss sec ioning
in he axial and co onal planes. In addi ion o plana sec ioning, a
mo e complex dissec ion o he b ain was simula ed o expose he
insula co ex by digi ally emo ing he o e lying co ical egions
(Figu e 6). A di e en edi ing app oach is shown in Figu e 7, which
p esen s a su ace endi ion o he Visible Ko ean male a e laye -
wise emo al o so issues and skull bone o expose he unde -
lying b ain.
3.3 | Tex u e baking and polygon educ ion
Tex u e baking p ese ed he gene al isual appea ance o models
and allowed expo a ion in OBJ ile o ma o subsequen mesh
FIGURE 7 S ipping laye s o he model. A su ace endi ion o
he head o he Visible Male om he Visible Ko ean da ase is
shown, using a su ace p obe om which wo pieces ep esen ing
po ions o he skin and skull o e lying he b ain had been
emo ed. No e he enuous pu ple imp egna ion o he skin due o
he embedding gel
FIGURE 8 Reducing mesh complexi y wi hou losing isual ealism.
The igh empo alis om he Visible Human Female is shown
a ached o he co onoid p ocess, be o e ex u e baking ( op). Two
e sions o he muscle model alone a e shown a e ex u e baking
and educing he e ex coun by ca. 86% o by ca. 97% (bo om).
The e ex coun could be educed d ama ically ( igu es p o ided
in Table 2) wi hou no iceably a ec ing he o e lying bi map
ex u e. The magni ica ion boxes show he s uc u e o he polygon
mesh om app oxima ely he same egion a he a ious le els o
simpli ica ion
560
|
AZKUE
simpli ica ion. As shown in Figu e 8 and Table 2 using a empo alis
model, mesh simpli ica ion can d ama ically educe ile size while en-
i ely p ese ing ex e nal appea ance.
4 | DISCUSSION
4.1 | Con ibu ion o he s a e o he a
An ad an age o olume ende ing in compa ison wi h su ace en-
de ing is di ec isualiza ion wi hou he need o segmen a ion.
Howe e , olume ende ing indi idual o gans wi hou p io segmen-
a ion is challenging, and e en mo e so is making hem appea as
dissec ed o p epa ed o ana omical demons a ion. On he o he
hand, olume- ende ed scenes as such canno be sa ed o disk and
s o ed o la e use. The me hod p esen ed he e allows o p oduce
ana omical models ha can be ende ed bo h as a whole and as i
p e iously dissec ed, can be expo ed o disk indi idually, and easily
combined o c ea e cus om ana omical scenes. Fu he , i is shown
ha colo - ex u ed models can be pos p ocessed o op imizing ile
size wi hou any disce nible loss o isual de ail.
Applica ion o bi map ex u es o su ace models, including
eal issue pho og aphs, is a use ul app oach o mimic he na u-
al appea ance o ana omical s uc u es (P eim & Saal eld, 2018;
Zil e schoon e al., 2017). Su ace scanning and pho og amme y
can be seen as e inemen s o his s a egy, whe e he 3D su ace
s uc u e is en iched wi h ac ual colo p ope ies om pho og aphs
o he same objec . These a e p obably ideal app oaches o digi-
izing p osec ed body pa s, as shown by Pe iceks e al. (2018), as
well as bone specimens (Azkue, 2021a), since 3D econs uc ions
using such echniques equi e cap u ing he geome y o he ob-
jec o in e es om all possible iewpoin s and consequen ly he
scanned objec needs o be de ached om he body. I is no e-
wo hy, howe e , ha he le el o de ail o econs uc ed models
is limi ed by p osec ion quali y, and he e a e a numbe o ac-
o s ha can comp omise he quali y o models such as di icul
angles, poo ligh ing, and occlusions (Pe iceks e al., 2018). The
colo sampling p ocedu e desc ibed he e o e s signi ican ad-
an ages o e p e ious app oaches. Fi s , his is a nondes uc i e
echnique ha allows o p oduce mul iple endi ions o he same
ana omical s uc u e in a a ie y o shapes and p epa a ions, wi h
he only limi o he abili y o edi he su ace p obe digi ally. F ail
o collapsible s uc u es, such as small essels, u e e s, o bile
duc s, ha become easily dis o ed o damaged du ing ac ual dis-
sec ion can be econs uc ed and ep esen ed accu a ely, and so
do smalle s uc u es di icul o access by dissec ion, such as he
inne ea , p o ided ha c yosec ion images o adequa e esolu ion
a e u ilized. Second, i digi al 3D models a e o se e as an educa-
ional esou ce in he con ex o in e ac i e so wa e (Mu gi oyd
e al., 2015; P eim & Saal eld, 2018; T iepels e al., 2020), i is im-
po an ha models o indi idual o gans and body pa s can be
combined and used in e ac i ely as sepa a e componen s wi hin a
scene, a he han p esen ed as a p osec ed egion o issue block.
Fu he mo e, he lexibili y o e ed by su ace 3D models can
be enhanced by including sec ional ana omy and egis e ed clin-
ical imaging da a such as CT o MRI in he scene (Liu e al., 2013;
Ma a - Ha amija e al., 2015; P a s- Galino e al., 2014; Robb &
Hanson, 2006; Shin & Pa k, 2016). Su ace models gene a ed om
cada e c yosec ion image da ase s a e ideally sui ed in his e-
ga d since mos a ailable collec ions also include clinical imaging
da a (Acke man, 1998; Acke man e al., 2001; Pa k e al., 2005;
Zhang e al., 2003, 2006).
4.2 | Rep esen ing human ana omy in a isually
ealis ic manne
Digi al ana omical models a e ex e nal ep esen a ions o he body,
whe e he co espondence be ween he ep esen ing and ep-
esen ed wo lds is es ablished by physical esemblance, based on
aspec s such as dimensionali y, numbe o s uc u es, spa ial ela-
ionships be ween s uc u es, and su ace de ails including ex u e
and colo de ail (Chan & Cheng, 2011; Palme , 1978). Li le pub-
lished in o ma ion is cu en ly a ailable as o he ex en o which
ealism o ana omical models in luences ana omy lea ning. Visual
ealism o models may be less c i ical o an unde g adua e medi-
cal s uden becoming acquain ed wi h he basic ana omy o a body
egion, who may also bene i om simpli ied, low- echnology mod-
els (Chan, 2015; Chan & Cheng, 2011). Howe e , a as majo i y o
ana omis s (80%) in a ecen su ey ag eed ha using ealis ic mod-
els o each ana omy is impo an o s uden s o e ain hei ana-
omical knowledge (Bal a e al., 2017). Con inued e o s o de elop
TABLE 2 Me ada a o a empo al muscle model subjec ed o ex u e baking and wo le els o polygon educ ion
Ve ex coun
File size
X3D ile OBJ ile MTL ile PNG ile
O iginally expo ed 288,546 43.8 MB
P io o simpli ica ion 288,546 90.6 MB 728 by es 53.1 MB
Fi s - le el simpli ica ion 38,234 5.9 MB 278 by es 2.5 MB
Second- le el simpli ica ion 9745 1.4 MB 282 by es 2.4 MB
No es: The su ace mesh was i s expo ed in X3D o ma suppo ing colo da a appended o polygon e ices. Following he ex u e baking
ope a ion, geome y and colo da a we e expo ed in OBJ ile o ma wi h associa ed MTL and PNG iles, and subjec ed o polygon educ ion.
