Resul O e iew
Resul s o AI-Suppo ed Ex ac ion o Func ional Tissue Uni
P ope ies o Human Re e ence A las Cons uc ion
Yongxin Kong1,2, * and Ka y Bö ne 1,*
1Depa men o In elligen Sys ems Enginee ing, Luddy School o In o ma ics, Compu ing, and Enginee ing, Indiana
Uni e si y, Blooming on, IN 47408, USA
2Cen e o Regene a i e Medicine and Heal h, Hong Kong Ins i u e o Science & Inno a ion, Chinese Academy o
Sciences, Hong Kong, 999077, China
The Human Re e ence A las (HRA) e o b ings oge he expe s om mo e han 25 in e na ional conso ia o cap u e
he mul iscale o ganiza ion o he human body— om la ge ana omical o gan sys ems (mac o) o he single-cell le el
(mic o). Func ional issue uni s (FTU, meso) in 10 o gans ha e been de ailed and 2D illus a ions ha e been c ea ed
by expe s. Comp ehensi e FTU p ope y cha ac e iza ion is essen ial o he HRA, bu manual e iew o he as
numbe o schola ly publica ions is imp ac ical. He e, we in oduce La ge-Model Re ie al-Augmen ed Gene a ion o
HRA FTUs (HRA u-LM-RAG), an AI-d i en amewo k o scalable and au oma ed ex ac ion o FTU- ele an
p ope ies om schola ly publica ions. This alida ed amewo k in eg a es La ge Language Models o ex ual
easoning, La ge Vision Models o isual in e p e a ion, and Re ie al Augmen ed Gene a ion o knowledge
g ounding, o e ing a balanced ade-o be ween accu acy and p ocessing e iciency. We e ie ed 244,640 PubMed
Cen al publica ions con aining 1,389,168 igu es o 22 FTUs and iden i ied 617,237 igu es wi h mic oscopy and
schema ic images. F om hese images and associa ed ex , we au oma ically ex ac ed 331,189 scale ba s and
1,719,138 biological en i y men ions, along wi h dono me ada a such as sex and age. All da a and code a e eely
a ailable a h ps://gi hub.com/cns-iu/h a- u- ag-suppo ing-in o ma ion in suppo o obus and scalable compu a ion
o schola ly e idence o he size and mo phological p ope ies and hei dependence on dono demog aphics—
acili a ing he design, e iew, and app o al o u u e FTU illus a ions du ing HRA cons uc ion.
Table desc ip ion
Table Name
A ibu e
Desc ip ion
u_publica ion
Pu pose
To map he unc ion issue uni s o he PubMed Cen al (PMC) publica ions.
Fields
Column
Da a ype
Desc ip ion
u
s ing
Func ion issue uni .
pmcid
s ing
The PubMed Cen al iden i ie o a publica ion.
publica ion_su
mma y
Pu pose
To s o e a concise summa y o key me ada a o PMC publica ions.
Fields
Column
Da a ype
Desc ip ion
pmcid
s ing
The PubMed Cen al iden i ie o a publica ion.
a icle_ i le
s ing
Ti le o he publica ion.
pmid
s ing
The PubMed iden i ie .
doi
s ing
The Digi al Objec Iden i ie .
abs ac
s ing
The abs ac ex o he publica ion.
pub_yea
s ing
The yea in which he publica ion was published.
jou nal_ i le
s ing
The name o he jou nal in which he publica ion appea ed.
ile_pa h
s ing
PMC ilesys em pa h o he downloaded ile o his
publica ion.
Rela ions
hips
Linked o " u_publica ion" able ia "pmcid".
image_summa
y
Pu pose
To s o e a summa y o each image associa ed wi h a PMC publica ion.
Fields
Column
Da a ype
Desc ip ion
pmcid
s ing
The PubMed Cen al iden i ie o a publica ion.
igid
s ing
The iden i ie o he igu e wi hin he publica ion.
label
s ing
The display label o he igu e.
g aphic
s ing
The g aphic objec iden i ie .
cap ion
s ing
The ull cap ion ex o he igu e.
ile_pa h
s ing
The PMC ilesys em pa h o he sou ce ile.
Rela ions
hips
Linked o " u_publica ion" able ia "pmcid".
image_ ype
Pu pose
To s o e he image ype o each image.
Fields
Column
Da a ype
Desc ip ion
pmcid
s ing
The PubMed Cen al iden i ie o a publica ion.
g aphic
s ing
The g aphic objec iden i ie .
ex
s ing
File ex ension o he image.
is_mic o
s ing
Indica es whe he he image is a mic oscopy image.
is_s a is
s ing
Indica es whe he he image is a s a is ical image.
is_schema
s ing
Indica es whe he he image is a schema ics image.
is_3d
s ing
Indica es whe he he image is a 3D s uc u e diag am.
is_chem
s ing
Indica es whe he he image is a chemical s uc u e diag am.
is_ma h
s ing
Indica es whe he he image is a ma hema ical exp ession.
Rela ions
hips
Linked o "image_summa y" able ia "pmcid".
Linked o "image_summa y" able ia "g aphic".
scale_ba
Pu pose
To s o e in o ma ion abou scale ba measu emen s associa ed wi h images in PMC
publica ions.
Fields
Column
Da a ype
Desc ip ion
u
s ing
Func ion issue uni .
pmcid
s ing
The PubMed Cen al iden i ie o a publica ion.
g aphic
s ing
The g aphic objec iden i ie .
alue
s ing
The nume ic leng h o he scale ba as epo ed in he igu e.
uni
S ing
The uni associa ed wi h he scale ba alue.
Rela ions
hips
Linked o "image_ ype" able ia "pmcid".
Linked o "image_ ype" able ia "g aphic".
dono _me ada
a
Pu pose
To s o e dono me ada a ex ac ed om PMC publica ions.
Fields
Column
Da a ype
Desc ip ion
pmcid
s ing
The PubMed Cen al iden i ie o a publica ion.
id
s ing
In e nal iden i ie wi hin he publica ion ha deno es he
loca ion o objec con aining he dono me ada a.
ype
s ing
Con ex ype indica ing whe e he me ada a was ound.
species_ ag
s ing
Tag o he dono ’s species.
de elop_ ag
s ing
De elopmen al s age ca ego y o p ena al and pos na al.
sex_ ag
s ing
Sex ag o he dono .
bmi_ ag
s ing
Body-mass index ca ego y.
age_yea old_ ag
s ing
Nume ic age in yea s.
Rela ions
hips
Linked o "publica ion_summa y" able ia "pmcid".
biological_s uc
u e
Pu pose
To s o e biological en i ies ex ac ed om PMC publica ions o 22 FTUs, along wi h
con ex ual in o ma ion abou whe e each en i y appea s (e.g., igu e cap ions, igu e
nodes).
Fields
Column
Da a ype
Desc ip ion
Rela ions
hips
pmcid
s ing
The PubMed Cen al iden i ie o a publica ion.
id
s ing
An in e nal iden i ie wi hin a single a icle ha deno es he
loca ion o objec con aining he en i y.
ype
s ing
The con ex ype indica ing whe e he en i y was ound (i.e.,
“cap ion”, “ igu e_node”).
En i y
s ing
The ex ac ed biological e m o name.
Rela ions
hips
Linked o "publcia ion_summa y" able ia "pmcid".
