BearWorks BearWorks
College of Arts and Letters
1-1-2021
IDCube lite: Free interactive discovery cube software for multi- IDCube lite: Free interactive discovery cube software for multi-
and hyperspectral applications and hyperspectral applications
Deependra Mishra
Helena Hurbon
John Wang
Steven T. Wang
Tommy Du
See next page for additional authors
Follow this and additional works at: https://bearworks.missouristate.edu/articles-coal
Recommended Citation Recommended Citation
Mishra, Deependra, Helena Hurbon, John Wang, Steven T. Wang, Tommy Du, Qian Wu, David Kim et al.
"IDCube Lite: Free Interactive Discovery Cube software for multi and hyperspectral applications." Journal
of Spectral Imaging 10 (2021).
This article or document was made available through BearWorks, the institutional repository of Missouri State
University. The work contained in it may be protected by copyright and require permission of the copyright holder
for reuse or redistribution.
For more information, please contact [email protected].
Authors Authors
Deependra Mishra; Helena Hurbon; John Wang; Steven T. Wang; Tommy Du; Qian Wu; David Kim; Shiva
Basir; Maria Gerasimchuk-Djordjevic; and For complete list of authors, see publisher's website.
This article is available at BearWorks: https://bearworks.missouristate.edu/articles-coal/312
Correspondence
Mikhail Berezin ([email protected])
Received: 22 January 2021
Revised: 8 April 2021
Accepted: 22 April 2021
Publicaon: 5 May 2021
doi: 10.1255/jsi.2021.a1
ISSN: 2040-4565
Citaon
D. Mishra, H. Hurbon, J. Wang, S.T. Wang, T. Du, Q. Wu, D. Kim,
S.Basir,Q.Cao,H.Zhang,K.Xu,A.Yu,Y.Zhang,Y.Huang,R.Garne,M.
Gerasimchuk-DjordjevicandM.Y.Berezin,“IDCubeLite:FreeInteracve
DiscoveryCubesowareformulandhyperspectralapplicaons”,
J. Spectral Imaging 10, a1 (2021). hps://doi.org/10.1255/jsi.2021.a1
© 2021 The Authors
This licence permits you to use, share, copy and redistribute the paper in
anymediumoranyformatprovidedthatafullcitaontotheoriginal
paper in this journal is given.
1
D. Mishra et al., J. Spectral Imaging 10, a1 (2021)
volume 1 / 2010
ISSN 2040-4565
IN THIS ISSUE:
spectral preprocessing to compensate for packaging film / using neural nets to invert
the PROSAIL canopy model
JOURNAL OF
SPECTRAL
IMAGING
JSI
Peer Reviewed Arcle
openaccess
IDCube Lite: Free Interactive Discovery
Cubesoftwareformulti-andhyperspectral
applications
Deependra Mishra,
a
Helena Hurbon,
a,d
John Wang,
a,d
Steven T. Wang,
a,d
Tommy Du,
a
Qian Wu,
a
David Kim,
a
Shiva Basir,
a
Qian Cao,
a
Hairong Zhang,
a
Kathleen Xu,
a
Andy Yu,
a
Yifan Zhang,
a
Yunshen Huang,
a
RomanGarne,
b
Maria Gerasimchuk-Djordjevic
c
and Mikhail Y. Berezin
a,d
*
a
DepartmentofRadiology,WashingtonUniversitySchoolofMedicine,4515McKinleyAve,StLouis,MO63110,USA
b
DepartmentofComputerScienceandEngineering,WashingtonUniversity,1BrookingsHall,StLouis,MO63110,USA
c
ArtandDesignDepartment,MissouriStateUniversity,901S.NaonalAve,Springeld,MO65897,USA
d
HSpeQLLC,4340DuncanAve,StLouis,MO63110,USA
Contacts
Deependra Mishra: [email protected]
hps://orcid.org/0000-0003-1578-8526
Helena Hurbon: [email protected]
hps://orcid.org/0000-0003-3181-6566
John Wang: [email protected]
Steven T. Wang: m.d@tascienst.com
hps://orcid.org/0000-0002-5630-9010
Tommy Du: [email protected]
Qian Wu: [email protected]om
David Kim: [email protected]om
Shiva Basir: shiv[email protected]
Qian Cao: [email protected]
hps://orcid.org/0000-0002-3900-0433
Hairong Zhang: [email protected]
Kathleen Xu: [email protected]
Andy Yu: [email protected]
YifanZhang:[email protected]
Yunshen Huang: [email protected]
RomanGarne:garne@wustl.edu
hps://orcid.org/0000-0002-0152-5453
Maria Gerasimchuk-Djordjevic: [email protected]
Mikhail Berezin: [email protected]
hps://orcid.org/0000-0002-2670-2487
Mul-andhyperspectralimagingmodaliesencompassagrowingnumberofspectraltechniquesthatndmanyapplicaonsingeospaal,bio-
medical,machinevisionandotherelds.Therapidlyincreasingnumberofapplicaonsrequiresconvenienteasy-to-navigatesowarethatcanbe
usedbynewandexperienceduserstoanalysedata,anddevelop,applyanddeploynovelalgorithms.Herein,wepresentourplaorm,IDCubeLite,
anInteracveDiscoveryCubethatperformsessenaloperaonsinhyperspectraldataanalysistorealisethefullpotenalofspectralimaging.
Thestrengthofthesowareliesinitsinteracvefeaturesthatenabletheuserstoopmiseparametersandobtainvisualinputfortheuserina
waynotpreviouslyaccessiblewithothersowarepackages.Theenresowarecanbeoperatedwithoutanypriorprogrammingskillsallowing
interacvesessionsofrawandprocesseddata.IDCubeLite,afreeversionofthesowaredescribedinthepaper,hasmanybenetscompared
toexisngpackagesandoersstructuralexibilitytodiscovernew,hiddenfeaturesthatallowuserstointegratenovelcomputaonalmethods.
Keywords: hyperspectral, mulspectral, spectral imaging, IDCube, segmentaon, geospaal, biomedical
2 IDCubeLite:FreeInteractiveDiscoveryCubeSoftwareforMulti-andHyperspectralApplications
Introducon
Mul-andhyperspectralimaging(HSI)modalieshave
emergedasanexcingopportunitytoexploreopcal
properesofobjectsanddiscoverhiddenfeaturesnot
accessible by other techniques. In contrast to the tradi-
onalspaalimagesproducedbyconvenonalcameras,
spectralimaginggenerates3Ddatasets(datacubes),
withspaalandspectraldimensions.Witheachpixel
containinginformaonontheenremediumorhigh-
resolution spectrum, spectral imaging provides abun-
dantinformationaboutindividualchromophoresand
theirinteraconsthatcontributetothelocaon,intensity
andalteraonoftheopcalsignal,signicantlybeer
thanmonochromacortradionalcolourcameras.
1,2
This
spectral imaging approach leads to a vastly improved
abilitytoclassifyanddierenatetheobjectsbasedon
theirspectralfeatures,enablingevensmall,otherwise
unnoceable,featurestobeamplied.
In the last decade, spectralimaginghasexpandedfrom
anarrownicheaccessibleonlytoahandfuloforganisa-
onsinacademicandgovernmentalresearchfaciliesto
abroadrangeofcommercialandclinicalinstuons.As
spectral imaging hardware (benchtop scanners, handheld
HSI cameras, drones etc.) have become more available,
3–5
thenumberofspectralimagingstudieshasincreased
tremendously,
6–11
reaching more than 25,000 publica-
onsin2019.
Meaningfulanalysisofdatacubesisthemostcrical
andme-consumingstepinmanycurrentapplicaons.
Thehighdimensionalityofspectralimagingdataand
theirlargedatasizes(oen>1 GB) gives an excellent
opportunity to learn more about the subject; however,
theextensiveanalyses(i.e.pretreatments,idenfying
appropriatealgorithmsetc.)ofthesedatasetspresentthe
strongestbarriertotheimagingworkow.Despitethe
progressincomputaonalspeedandalgorithmdevel-
opment,ecientcomputaonalapproachessuitablefor
generaluseapplicaonsarelacking.Well-knownsoware
packages,suchasENVI,
12
are comprehensive; however,
theyaremostlyusedforremotesensingapplicaons.
Ahandfuloffreehyperspectralsowarepackageswith
stronginnovavealgorithms(e.g.,Gerbil
13
orMulSpec
14
)
aremoresuitableforexpertsandarelimitedtogeospa-
alapplicaons.Introducedin2020,theHyperspectral
ToolboxfromMATLAB
15
hasvery few algorithms,is
limitedinthenumberofsupportedleformatsandis
largelycommand-linebased.Otherpackagesanalysing
spectralimagingdataareaachedtospecichardware
to visualise the acquired data and usually do not provide
thedesiredlevelofdataminingandprocessingcapability.
Assuch,thereisatangibleneedforamoreuniversal,
powerfulcomputaonalplaormthatenablescompre-
hensiveandrapiddataminingforavarietyofplaormsin
ordertooerreal-meeciencyandthroughputforthe
majorityofapplicaons.
Herein, we present IDCube, Interactive Discovery
Cube, Lite (hps://www.idcubes.com), a highly versa-
lesowarethatperformsalargenumberofessenal
operaonsinthespectralimagingdomainandenables
imageanalysisforusersacrossarangeoftechnicalpro-
ciencies.Thegoalofthesowareistomakespectral
imagingaccessibletonewandcurrentusersthatfocus
onobtainingusefulresultsratherthan(butnotexcluding)
developingalgorithms.Thestrengthoftheproposed
sowareliesinitsintuivedesignthatenablestheuser
toperformhigh-leveldataanalysisaswellasdevelop
theirownalgorithmsviaavisual,interacveinterface.
Builtaroundacolleconofspectralimagingalgorithms,
thesowarefacilitatesthesearchofhiddeninformaon
insidelargedatasetsprovidinganewexperienceofdata
analysis.Theenreprogramcanbeoperatedwithout
prior programming skills and allows almost any currently
useddataformatstobeprocessed.
TheoverallworkowofIDCubeLiteisshowninTable1.
Theprocessingstepsarecentredaroundthevisualisaon
module that presents the original and processed data in
a2Dformat.Thefrontendofthesowareisshownin
Figure 1.
ThecoreofthesowareiswrieninMATLABwithan
extensivenumberofbuilt-inimageprocessingfuncons.
Thecompiledsowarecanberunonanycomputerwith
aWindowsorMacOperangSystem,withouthavinga
MATLABlicense.Anenrelyfreeversionofthesoware
IDCubeLiteisavailablefordownloadfromasecurecloud
locaon,www.idcubes.com.Anextensivelibraryofvideo
tutorialsisavailablefromthelearningmoduleoftheso-
warewebsite.Belowwewillfocusonthetechnicalcapa-
biliesofIDCubeLiteanddemonstrateitsperformance
throughseveralexperimentsinthegeospaal,machine
visionandbiomedicalelds.Abriefdescriponofthe
datasets used in this paper is given in Table 2.
Features
IDCubeLiteenablesvisualanalysisofmanydatasetsfrom
avarietyofformats.Theplaormhasbeenconstantly
improving since its launch in September 2020 and now
includesmorethan50integratedalgorithmsfordata
D. Mishra et al., J. Spectral Imaging 10, a1 (2021) 3
processinganddatavisualisaon.Thesealgorithmsare
groupedintothefollowingcategories:Input/Output,
DataReducon,ImageVisualisaon,SpectralAnalysis,
PrincipalComponentAnalysis(PCA)/MaximumNoise
Fracon(MNF),SegmentaonandSpectralMatching.
Theselectedalgorithmsareoptimisedforprocessing
meanddonotexceedseveralsecondswhenprocessing
a 1 GBleonastandarddesktopcomputer.Thedescrip-
onsofthemajorfeaturesaregivenbelow.
Input/output module
IDCubeLitesupportsawiderangeofimageformats,
including Raw/HDR, DAT,TIFF, JPEG2k, PNG and
severalothers.Thelistoftheavailableformatsisgivenin
Table3.Theimportedles(datale+headerle)arerst
converted, then saved in the same directory as a single
MATLAB-typele.Thesinglelethatcombinesrawdata
andthechannelassignmentinformation(wavelength
vector)automacallyopensintheIDCubeLiteinterface
Modules Specicoperaons Usageexample
DataI/O Import, convert, export Import satellite and other images and
datasets
Visualisaon Pan, zoom Displays image
Preprocessing Cropping, binning Size decrease
Datareducon PCA,MNF Dimensionalreducon
Image enhancement Histogram-based contrast adjustment Enhanceobjectcontrastforvisualisaon
Spectral analysis Pixelselecon,SpectralMatching,end-
members etc.
Find pixels with similar spectral signatures
in image
Segmentaon Thresholding Spectralcorrelaonanddivergencemaps
Image algebra Addion,subtraconetc. Contrast enhancement
a
see SupplementaryInformaon describing the steps
Table1.ListoffunconaliesoeredbyIDCubeLite,specicfunconsandusagecase.Visually,thesefunconaliesare
mappedtopanelsshowninFigure1.
a
Figure1.FrontendoftheIDCubeLitehyperspectralsoware.A:import/export,datareducon,data
correcon;B:imagealgebraseleconwithpreselectedanduser-developedfuncons;C:wavelength
andbandwidthselecon,imageopmisaon;D:imagevisualiser,E:spectralanalysisofselectedROIs;F:
imageenhancementviahistogrammanipulaon;G:informaonabouttheimage,theconsumedcompu-
taonalresourcesfortheperformedtasks.
4 IDCubeLite:FreeInteractiveDiscoveryCubeSoftwareforMulti-andHyperspectralApplications
aertheconversioniscomplete.IDCubeLitesupports
imagingdatafromavarietyofbench-typeHSIsystems
thatuliseaRaw/HDRformataswellasfromsatellites
oeringdatainJP2andTIFF.InaddiontoHSIdata-
sets,thesowarecanalsotreatstandardRGBimages
andthuspresentsaninteresngopportunitytoapply
sophiscatedcomputaonalHSItechniquestocommon
pictures.
Thesowarecanopenuptotenhyperspectraldata-
setssimultaneouslyforalldatasetswiththesamedimen-
sions.Iftheimageshavedierentspaaldimensions,
theycanberstcroppedusingthespaalcroppingfunc-
on.Analysingmulpledatasetsinonesengenables
datacomparisonandthesameprocessingfordatafrom
longitudinal studies.
Preprocessing module
Thepreprocessingmoduleincludestheabilitytoperform
binninginthespatialdomaintodecreasethesizeof
thefileandreducenoise.Themoduleoffersinterac-
tivity through spatial cropping, spectral cropping and
ipping/transposion/rotaon.Allpreprocessingfunc-
onsareappliedgloballytotheenredataset.TheDATA
CORRECTIONtoolremovesundesirableartefactscaused
byscaeringoflightthatproducesundesirableartefacts.
Thecorreconalgorithmsincludemulplicavescaering
Dataset Applicaon Instrument
# Channels and
wavelengths
Size
(MB)
Random items Machine vision Benchtop, SWIR/pushbroom 510 channels
867–1700 nm
326
Human hand Biomedical Benchtop, SWIR/pushbroom 343channels
950–170 0 nm
217
Leaves Agriculture Benchtop, SWIR/pushbroom 350channels
940–170 0 nm
82
ViewofForest
Park in St Louis
Geospaal Satellite, Pleiades-1B/AIRBUS 4 channels
430–550 nm (blue),
490–610 nm (green),
60 0 –720 nm (red);
750–950 nm(NIR)
46
Table2.Descriponofthedatasetsusedasanexample.
Format Type of detectors Source Implementaon
JPEG,PNG Colour RGB cameras Many Yes
TIFF Mulandhyperspectralsatellites
Bench-type imagers
AVIRIS
AVIRISNG
a
OBTsatellites
Hyperion
Various
Yes
In process
Yes
Yes
Yes
hdr/raw/dat Mulandhyperspectralsatellites
Bench-type HSI imagers
Hyperion
Specim
Middleton
SpectralVision
CytoViva
Headwall
Yes
Yes
Yes
Yes
Yes
JP2
h5
Mulspectralsatellites
Mulspectralsatellites
Sennel-2
SuomiNPP
Yes
In process
a
AVIRISNextGeneraondatahavedierentsetsofwavelengthsfromtheAVIRIS
Table3.TypesoflesandformatssupportedbyIDCubeLite(othersourcescanalsoberecognised).
D. Mishra et al., J. Spectral Imaging 10, a1 (2021) 5
correcon(MSC)
16
andstandardnormalvariate(SNV)
17
funcons.Thisfeatureisespeciallyusefulforbiomedical
imaging;forexample,todecreasescaeringartefacts
fromskin.Otherpreprocessingstepstodecreasethesize
oftheimageleviaremovalofhighlycorrelangspectral
bandsandextracngendmembersignaturesfromhyper-
spectral data
18
willbeimplementedinfutureversions.
Visualisaonmodule
This module is central to image processing and pres-
entsthree-dimensional(3D)datasetsthroughasetof
two-dimensional (2D) images. The 2D images are gener-
ated through the three wavelength channels. The wave-
lengthseleconcanalsobeappendedwithapreselected
bandwidthratherthanthedefaultbandwidthof1.The
producedmonochromacimagecanbecolouropmised
by applying an appropriatelookuptable(LUT) from
morethan20availableLUTs.Thevisualisaonisfurther
adjustedviathe histogram tool.The BROADBAND
funconenablesvisualisingtheimagewithinaspecic
wavelengthrange.Whenthisfunconisselected,the
wavelengthsw1andw2willprovidetheboundariesfor
thespectralrange(i.e.theimageproducedbyselecon
w1 = 950 nm, and w2 = 1400 nm would correspond to an
imageacquiredbythecameraoverthe950–1400 nm
range).TheMATHEMATICSmodeinconjunconwith
theTWO-CHANNELmodesenablestheusertoconduct
imagealgebra.Presetfunconsincludedivisionofone
wavelengthchanneloveranother,subtracon,logical
functions etc.TheEXPRESSIONmoduleallowsthe
usertotypeacustommathemacalfunconorselect
fromthelistofimplementedfunconsspeciedinthe
MathemacsSheetforExpressionsavailablefromthe
sowarewebsite(hps://www.idcubes.com/tutorials).
TheSINGLECHANNEL,TWO-CHANNELandTHREE-
CHANNELtoolsallowuserstoscrolltheimagesthrough
the individual wavelengths. The RGB mode enables
the user to combine up to three wavelengths into a
pseudo-RGBimage.Eachwavelengthcanbeusedwith
thespecicbandwidth.TheHISTOGRAMopmisaon
andtheCONTRASTADJUSTMENTeldsenabletheuser
toimprovetheimagecontrast.Oneoftheuniquefeatures
ofIDCubeLiteistopresentthedatasetasamoviewhere
eachframerepresentsanimageataspecicwavelength
orwithaformulaapplied.Thisfeatureimplementedin
theFRAMEBYFRAMEdisplayfunconradicallymini-
misestheamountofuserinteraconandpreventsimage
processingfague.Theimagewiththeenredatasetcan
beipped,transposedandrotated.Theproduced2D
image can be also copied, zoomed, panned and saved.
Spectral analysis module
The spectral analysis module enables the user to visualise
andprocessspectralinformaonfromindividualpixels
andinteracvelyselectregionsofinterests(ROIs).Inthe
REAL-TIMEspectralmode,themoduleenablesvisual-
isaonofspectrafromindividualpixelsbymovingthe
cursorovertheimage.Themoduleautomacallyrecog-
nisesaspectralrangeandscalesthedimensionsofthe
spectralplot.IntheMULTI-SPECTRAmode,theuser
ispromptedtoselectoneormoreROIs.Thespectrum
foreachROIreectstheaveragespectrumacrossthe
selectedarea.SPECTRAMATHEMATICSenablesthe
usertoperformbasicmathfuncons,i.e.subtraconand
divisionofthespectra,spectranormalisaonandcalcu-
laonsoftherstandsecondderivaves.Theusercan
comparethespectraloutputfromtheselectedROIsusing
spectralcorrelaon,spectralinformaondivergence
19
or
spectral angle
20
funcons.Relavelyhighspectralcorre-
lation values, low divergence and low spectral angles
suggestregionswithsimilarspectralproperesindicang
theobjectsbelongtothesameclassofobjectswith
similaropcalprolesorsamematerials.Allspectracan
bezoomed,pannedandexportedtoExcelorotherdata
analysissoware.
Principal Component Analysis (PCA) module
The PCA module (Figure 2) computes associations
betweendatapointsandconvertsadatasetofpoten-
allycorrelatedvariablesintoanewsetoflinearlyuncor-
related principal components.
21,22
Uptothreeprincipal
components can be selected by the user to generate a
pseudo-colour RGB image, where the selected compo-
nents are assigned to three colours, red, green and blue.
Objectswiththesamecolourindicatehighsimilarity
betweentwosubjects.Forexample,acentrifugetube
andawrenchshowninFigure2areapparentlymadefrom
the same material, since their PCA-based pseudo-colours
arealmostidencal.Thepseudo-colourRGBimagecan
befurtheradjustedthroughchangingtheWEIGHTofthe
individualcomponent,adjusngtheCONTRASTtothe
wholeimageandapplyingtheGAMMACORRECTION.
Thegammacorrecon
23
helps to improve the contrast
iftheimageistoodarkortoobright.Thevalueofthe
gammacorreconcantakeanyvaluebetween0and
innity(upto10usingaslider,ortoanyvalueiftyped).If
the gamma is less than 1, the output image is brightened,
forgammagreaterthan1,theoutputimageisdarkened.
Ingeneral,thetotalnumberofprincipalcomponents
isequaltothenumberofwavelengthchannels.Since
mostoftheinformaonisintherstfewcomponents,
6 IDCubeLite:FreeInteractiveDiscoveryCubeSoftwareforMulti-andHyperspectralApplications
theIDCubeLiteversionlimitsthenumberofstoredprin-
cipalcomponentsto20.TheCOMPONENTSPECTRA
window enables the user to examine the eigenvectors
visuallytocheckifrelevantfeaturesmaybeextracted.
Themodulealsopresentsthecumulativefractionof
variance.Inthegivenexample,thersttwocomponents
carry>96 %ofthevariance.PCAtransformstheorig-
inaldatacubeintoanewdatacubewiththesamespaal
dimensionsandchangestheZ-axisfromwavelengths
to principal component scores. In that case, PCA can be
alsoconsideredasafunconthatdecreasesthesizeof
thedata,sinceonlyveryfewprincipalcomponents(rst
20,forexample)canbeused.Thenew,smallerdatacube
canbeexportedbacktotheVISUALISATIONpaneland
analysed using implemented algorithms.
AvariaonofthePCAmethodisanMNF.
24
TheMNF
transform has advantages over the PCA transform
becauseittakesthenoiseinformationinthespatial
domainintoconsideraon.Forexample,theshadows
seen in Figure 2A can be removed to some extent using
theMNFfunconimplementedinIDCubeLiteasshown
in Figure 2B.
Segmentaonmodule
Thesegmentationmodule in IDCubeLite(Figure3)
enablesminimallysupervisedclassicaonofadataset
fromanyofthepreprocessingalgorithms(spaal,spec-
tralcropping,binning,PCAetc.).Inatypicalworkow,
theuserselectsoneoftheclassicaonalgorithms[i.e.
based on the improved Spectral Angular Mapper (SAM)
currently implemented in IDCube Lite] and the metrics
(i.e. area, perimeter) then draws an area (class) on the
imagepassedfromtheVisualisaonmodule.Areaswith
similarspectralproperes(thathavelowvaluesofspectral
angle)arerepresentedbythesamecolourandquaned
according to the selected metrics. The example shown in
Figure3illustratesthismethodforclassicaon.Onecan
nocethatthemethodishighlysensivetoevensmall
spectral changes, where the vial and the wrench can be
separatedeventhoughtheirspectralcorrelaonvalue
is0.99(asmeasuredusingtheSPECTRALANALYSIS
module, see above).
Speed and resources
Duetothelargesizeofhyperspectraldatasets,manyof
thefunconsofthesowareareopmisedforworking
with large datasets exceeding several hundred mega-
bytes. Figure 4presentsthespeedofthemostdemanding
funconsintypicalhyperspectraldatasetanalysis.Ale
ofabout1 GB can be opened in less than 10 s on a stan-
dardhomePCandsignicantlyfasteronmorepowerful
computers.Mostofthevisualisaonandspectralanalysis
funconsareperformedalmostinstantaneously.PCA
isthemostme-consumingwiththeprocessingme
signicantlyandnon-linearlyincreasingwiththesizeof
thele,reachingmorethan2 minfora1-GBleinour
testPC(DellInc.,VostroDT5090,IntelCorei7-9700,8
Core, GB (1 × 8 GB) DDR4 2666 MHzUDIMM).
Examples
Allexamplesmenonedinthispaper(Table2)andother
datasetscanbedownloadedfromIDCubeLitedirectly
Figure2.Preprocessingtools.A.PrincipalComponentAnalysisofahyperspectraldataset.First,second
and third principal components are combined in a pseudo-colour RGB image. The module also presents
thespectraofselectedcomponentsandacumulavefraconofvariance.Theimagecanbeimproved
byadjusngthecontrastandgammacorrecon.B.MaximumNoiseFraconofthesamedatashowsthe
paralremovaloftheshadow.ExamplelecanbedownloadedthroughIDCubeLiteunderFile/Example
Files/PlascAndCoin.
D. Mishra et al., J. Spectral Imaging 10, a1 (2021) 7
(File–Downloadexamples)orfromthewebsitehps://
www.idcubes.com/examples.
Biomedicalapplicaons
With the development of clinically relevant hyper-
spectralimaginginstrumentaon,HSIhasemergedas
apowerfultoolfor investigatingcomplexbiological
systems.
1,25
Biologicalssuesinthevisiblerangeoendo
notprovidesucientcontrasttodisnguishthestruc-
turesofinterestand,therefore,requirecontrastagents
forcontrastenhancement.Hyperspectralimaging,with
itsinherentlyhighersensivitytominorchanges,can
replacesomeoftheelaboratestainingtechniquesand
significantlyacceleratethepathologicalpracticeboth
in vitro and in vivo. Clinical examples include histopa-
thology,
26
dermatology,
27
ophthalmology,
28
gastroen-
terology,
29
oncology
30,31
anddeepssueimagingwith
hyperspectralshortwaveinfrared(SWIR,900–2200 nm)
duetoahighpenetraonofSWIRphotonsthroughthe
skinandthessue.
32
Figure3.ImageClassicaonModule.IntheSpectralAngularMapping,theuserinteracvely
selectsdierentareas(classes)fromthe2Dimage.Althoughthereisnolimittothenumberof
classes,themoduleworksbestforoneortwoclasses.Thedarklinescorrespondto“badpixels”
thatareoenseenintheSWIRcameras.Thismaybeduetoadamagedpixelinthesensorarray.
Figure4.Timeforleopening,PCAprocessingandsegmentaonwithSAMwithtwo
classes.PC:DellVostro5090,IntelCorei7-9700CPU,3 GHz,RAM24 GB,Windows
10.
8 IDCubeLite:FreeInteractiveDiscoveryCubeSoftwareforMulti-andHyperspectralApplications
TheexampleshowninFigure5illustratestheulity
ofIDCubeLitetobeervisualisebloodvessels.First,
the acquired dataset was spectrally cropped to elimi-
nate the noisy wavelength channels. Since our imaging
systembasedontheInGaAsdetector(Ninox640,Raptor
Photonics)incombinaonwiththespectrographN17E
(Specim Inc.) used in this study typically has lower sensi-
vitybelow900 nmandabove1700 nm, these wave-
lengths were excluded. The RGB mode was then selected,
and the wavelengths were manually adjusted to visu-
alise the blood vessels. The image made with a conven-
onalcolourcameradoesnotshowthebloodvessels
(Figure5A).Thevisualisaonoftheresulngimagewas
furtherimprovedbyadjusngcorrespondingcolourband
histograms.TheresulngimageshowninFigure5Bpres-
ents blood vessels in a greater contrast than the visible
image.ThedatasetwasthentreatedwiththeMNFfunc-
onselectedfromtheANALYSEtab.SimilartothePCA
moduledescribedabove,thesowareenablestheuser
to select individual components and presents them in the
pseudo-RGBformat(Figure5C).Highcontrastforblood
vesselswasachievedbyusingcomponents#2and#3.
In addition to HSI, IDCube can handle other spec-
tral imaging modalities commonly used in preclinical
and clinical studies, such as Raman and Fourier trans-
forminfraredspectroscopies,anduorescence-lifeme
imaging microscopy.
33
Environmentalapplicaons
HSIofplantsprovidessoluonstoalargenumberof
challengesfrom identifying environmentalissues to
monitoring crops yield and diseases
34
andevendetecon
ofcontaminaons
35
andminerals.Theopcalsignature
ofplantsandespeciallyleavesisanimportantmonitoring
andpredicveparameterforavarietyofbiocandabioc
stresses.Figure6illustratesanapplicaonofIDCubeLite
onadatasetfromleaveswithdierentmoisturelevels.
Therightleafoneachimagebelongstoaplantgrown
undernormalcondions,andtheleleafwasexposed
to a drying element. This treatment was used to mimic a
droughtcondioninordertoshowcasetheeecveness
oftheindex.Figure6Ashowsanimagerecordedbya
convenonalvisiblecamerafromacellphonecamera.
Figure6B–Dshowprocessedimagesrecordedusing
aSWIRhyperspectralimagerinreeconmode.Low
signal/noise ratio bands were removed using a spec-
tralcroppingfuncon.Apseudo-RGBimagecomposed
fromthreewavelengthsshowsthedierencebetween
the two leaves (Figure 6B). The contrast between
twoleavescanbefurtherenhancedwithPCA.Three
selectedprincipalcomponents#1/2/3wereusedina
pseudo-RGBformatasred,greenandblue(Figure6C).
Combined in a single image, these components high-
lightthedierencebetweenthetwoleaves.Evenhigher
contrast can be achieved using a previously developed
indexofdroughtusingtheformula:I = (1529–1416 nm)/
(1519 + 1416 nm)
36
(Figure 6D). This can be achieved by
selecngMathemacsfromtheANALYSISsecon,then
selecngMichelsonRaoandnallyselecngtwowave-
lengths 1416 nmand1529 nm.
Geospaalapplicaons
Geospaalremotesensingisoneofthemoremature
applications of hyperspectral imaging due to its
Figure5.Hyperspectralimagingofahand.A:madebyaconvenonalvisiblecamera;B:usingahyper-
spectralSWIRimager.Pseudo-RGBimageat1070 nm (red), 1260 nm (green), 1320 nm(blue);C:MNF
funconappliedtotheHSIdataset.Pseudo-RGBimageat#3(red),#3(green),#2(blue)components.
ExamplelecanbedownloadedthroughIDCubeLiteunderFile/ExampleFiles/HandSWIR.
D. Mishra et al., J. Spectral Imaging 10, a1 (2021) 9
relativelylong history,beginning in the middle of
1970s.
37
Sincethen,alargenumberofplaormsbased
on satellites, planes and, recently, drones have been
developed.IDCubeLitecanbeusedonanyofthese
plaorms.Thecurrentversionofthesowareenables
dataprocessingfromhyperspectralandmultispec-
tralsatellitessuchasER-01Hyperion,
38
Sennel-2,
39
Orbitasatellites,
40
airbornesystemssuchasAVIRIS
41,42
andotherplaorms.Theuserrstdownloadsthele
fromtherelevantimageproviderwebsites.IDCubeLite
convertstheunzippeddataintotheIDCubeformat,
savesandautomacallyopenstheconvertedlefor
furtherprocessing.Anexampleofthisworkflowis
showninFigure7,wheretheoriginaldatasetwasrst
downloadedfromacommercialvendor(ApolloHunter),
convertedtotheIDCubeformatandprocessedto
produceanRGBimageusingthefirstthreebands
(Figure7A).ThedatasetwasthenprocessedbyPCA.
Forbettervisualisationoftheobjects ofinterest,
three principal components were used to construct a
pseudo-RGBimage(Figure7B).APCA-baseddatacube
wasfurtherclassiedusingaSAMmethodbyselecng
roadasendmemberspectratogenerateanimageof
roadsandstreets(Figure7C).
Futuredevelopment
OurcurrentversionofIDCubeLiteisdownloadable
freesowarewiththeperformancelimitedbytheuser’s
computaonalresources.ThefutureIDCubeplaorm
willaddresstheneedforaweb-accessibleplaormto
performcomplexandcomputaonallydemandingtasksin
real-me.Equippedwithadvancedimageprocessingand
machinelearningcapabilies,theweb–based,constantly
updatedIDCubeplaormwillenablegeospaal,biomed-
icalandotherscientistsandstakeholderstoperform
sophiscatedanalysiswithoutsignicantcomputaonal
resources,usingonlyconvenonaldesktopsorlaptops.
Conictofinterest
BerezinisthefounderandCEOofHSpeQLLC.
Figure6.HyperspectralimagingofleaveswithIDCubeLite.A:imageoftwosimilarleavesobtainedwith
aconvenonalcolourcamera;theleafonthelecamefromtheplantexposedtoadryingcondion;
B:contrastimprovementwithathree-bandapproachinapseudo-RGBimage:1421 nm(red),1351 nm
(blue)and1476 nm(green);C:PCAwiththreecomponents#1(red),#2(green),#3(blue)inapseudo-RGB
image;D:monochromacimagereecngadroughtindex(1529–1416 nm)/(1519 nm+1416 nm).Exam-
plelecanbedownloadedthroughIDCubeLiteunderFile/ExampleFiles/RoseLeaves.
Figure7.MulspectralimageoftheStLouisareawithfourbands:RGB+NIR:430–550 nm (blue),
490–610 nm(green);600–720 nm(red);750–950 nm(NIR).A:RGBimage;B:PrincipalComponent
Analysis(PCA);C:SpectralAngularMappingwithoneclassselected.SatelliteSensorPleiades-1B
(0.5 m)operatedbyAirbusDefence&Space.ThedatawereacquiredfromApolloHunter.Examplele
canbedownloadedthroughIDCubeLiteunderFile/ExampleFiles/StLouisarea.
10 IDCubeLite:FreeInteractiveDiscoveryCubeSoftwareforMulti-andHyperspectralApplications
Acknowledgements
The team would liketo acknowledge funding from
NaonalScienceFoundaon,NSF1827656(MB)and
NSF1355406(MB,RG),andMallinckrodtInstuteof
Radiology (HH, QC). The original standalone package
waslicensedbyHSpeQLLCfromWashingtonUniversity
wherethesowarewaspreviouslydevelopedandhas
beenmodiedbyHSpeQLLC.
References
1.
G. Lu and B. Fei, “Medical hyperspectral imaging: a
review”,J. Biomed. Opt. 19(1),10901(2014).h p s://
doi.org/10.1117/1.JBO.19.1.010901
2.
W.Jahr,B.Schmid,C.Schmied,F.O.Fahrbach
and J. Huisken, “Hyperspectral light sheet micros-
copy”,Nat. Commun. 6,7990(2015).hps://doi.
org/10.1038/ncomms8990
3.
N.Gat,“Imagingspectroscopyusingtunablelters:a
review”,Proc. SPIE 4056, Wavelet Applicaons VII, pp.
50–64(2000).hps://doi.org/10.1117/12.381686
4.
N.GuptaandV.Voloshinov,“Hyperspectralimaging
performanceofaTeO
2
acousto-opctunablelter
intheultravioletregion”,Opcs Le. 30(9),985–987
(2005). hps://doi.org/10.1364/OL.30.000985
5.
D. Bannon, “Hyperspectral imaging: cubes and
slices”,Nat. Photon. 3(11),627–629(2009).h p s://
doi.org/10.1038/nphoton.2009.205
6.
Y.Khouj,J.Dawson,J.CoadandL.Vona-Davis,
“HyperspectralimagingandK-meansclassicaon
forhistologicevaluaonofductalcarcinomain situ”,
Front. Oncol. 8,17 (2018). hps://doi.org/10.3389/
fonc.2018.00017
7.
E.L.Wisotzky,F.C.Uecker,P.Arens,S.Dommerich,
A.HilsmannandP.Eisert,“Intraoperavehyper-
spectraldeterminaonofhumanssueproperes”,
J. Biomed. Opt. 23(9),091409(2018).hps://doi.
org/10.1117/1.JBO.23.9.091409
8.
M.E.Marn,M.B.Wabuyele,K.Chen,P.Kasili,M.
Panjehpour,M.Phan,B.Overholt,G.Cunningham,
D.Wilson,R.C.DeNovoandT.Vo-Dinh,
“Developmentofanadvancedhyperspectralimaging
(HSI)systemwithapplicaonsforcancerdetecon”,
Ann. Biomed. Eng. 34(6),1061–1068(2006).hps: //
doi.org/10.1007/s10439-006-9121-9
9.
W.R. Johnson, D.W. Wilson, W. Fink, M.S. Humayun
and G.H. Bearman, “Snapshot hyperspectral imaging
inophthalmology”,J. Biomed. Opt. 12(1),014036
(2007).hps://doi.org/10.1117/1.2434950
10.
M. Wahabzada, M. Besser, M. Khosravani, M.T.
Kuska,K.Kersng,A.-K.MahleinandE.Stürmer,
“Monitoringwoundhealingina3Dwoundmodelby
hyperspectralimagingandecientclustering”PLOS
One 12,e0186425(2017).hps://doi.org/10.1371/
journal.pone.0186425
11.
M.Garcia,C.Edmiston,R.Marinov,A.VailandV.
Gruev,“Bio-inspiredcolor-polarizaonimagerfor
real-mein situimaging”,Opca 4(10),1263–1271
(2017).hps://doi.org/10.1364/OPTICA.4.001263
12.
ENVI-GeospaalSoware.L3HarrisTechnologies,
Inc. hps://www.l3harrisgeospaal.com/Soware-
Technology/ENVI
13.
J.Jordan,E.AngelopoulouandA.Maier,“A
novelframeworkforinteracvevisualizaonand
analysisofhyperspectralimagedata”,J. Electr.
Comput. Eng. 2016,2635124(2016).hps://doi.
org/10.1155/2016/2635124
14.
D.Landgrebe,“Hyperspectralimagedataanalysis”,
IEEE Signal Proc. Mag. 19(1),17–28(2002).hps: //
doi.org/10.1109/79.974718
15.
Hyperspectral Image Processing Toolbox. Mathworks,
Inc. hps://www.mathworks.com/help/images/
hyperspectral-image-processing.html
16.
A.Candol,R.DeMaesschalck,D.Jouan-Rimbaud,
P.HaileyandD.Massart,“Theinuenceofdata
pre-processinginthepaernrecognionofexcipi-
entsnear-infraredspectra”,J. Pharm. Biomed. Anal.
21(1),115–132(1999).hps://doi.org/10.1016/
S0731-7085(99)00125-9
17.
E.Grisan,M.Totska,S.Huber,C.KrickCalderon,
M.Hohmann,D.LingenfelserandM.Oo,
“DynamiclocalizedSNV,peakSNV,andparalpeak
SNV:Novelstandardizaonmethodsforpreprocess-
ingofspectroscopicdatausedinpredicvemodel-
ing”,J. Spectrosc. 2018,5037572(2018).hps://doi.
org/10.1155/2018/5037572
18.
C.-I.ChangandA.Plaza,“Afastiteravealgorithm
forimplementaonofpixelpurityindex”,IEEE
Geosci. Remote Sens. Le. 3(1),63–67(2006).h p s://
doi.org/10.1109/LGRS.2005.856701
19.
C.-I.Chang,“Aninformaon-theorecapproach
to spectral variability, similarity, and discrimina-
onforhyperspectralimageanalysis”,IEEE Trans.
Inform. Theory 46(5),1927–1932(2000).hps://doi.
org/10.1109/18.857802
20.
F.A.Kruse,A.Leo,J.Boardman,K.Heidebrecht,
A. Shapiro, P. Barloon and A. Goetz, “The spectral
imageprocessingsystem(SIPS)-interacvevisual-
izaonandanalysisofimagingspectrometerdata”,
D. Mishra et al., J. Spectral Imaging 10, a1 (2021) 11
AIP Conf. Proc. 283,192–201(1993).hps://doi.
org/10.1063/1.44433
21.
G.Zuendorf,N.Kerrouche,K.HerholandJ.C.
Baron,“Ecientprincipalcomponentanalysisfor
mulvariate3Dvoxel-basedmappingofbrain
funconalimagingdatasetsasappliedtoFDG-PET
andnormalaging”,Hum. Brain Mapp. 18(1),13–21
(2003).hps://doi.org/10.1002/hbm.10069
22.
C. Rodarmel and J. Shan, “Principal component anal-
ysisforhyperspectralimageclassicaon”,Survey.
Land Inform. Sci. 62(2), 115 (2002).
23.
M.S. Tooms, Colour Reproducon in Electronic Imaging
Systems: Photography, Television, Cinematography.
John Wiley & Sons (2016). hps://doi.
org/10.1002/9781119021780
24.
A.A. Green, M. Berman, P. Switzer and M.D. Craig,
“Atransformaonfororderingmulspectraldatain
termsofimagequalitywithimplicaonsfornoise
removal”,IEEE Trans. Geosci. Remote Sens. 26(1),
65–74(1988).hps://doi.org/10.1109/36.3001
25.
S.Kiyotoki,J.Nishikawa,T.Okamoto,K.Hamabe,M.
Saito, A. Goto, Y. Fujita, Y. Hamamoto, Y. Takeuchi,
S.SatoriandI.Sakaida,“Newmethodfordetecon
ofgastriccancerbyhyperspectralimaging:apilot
study”,J. Biomed. Opt. 18(2),026010(2013).h ps: //
doi.org/10.1117/1.JBO.18.2.026010
26.
G.Lu,D.Wang,X.Qin,S.Muller,J.V.Lile,X.Wang,
A.Y. Chen, G. Chen and B. Fei, “Histopathology
featureminingandassociaonwithhyperspectral
imagingforthedeteconofsquamousneoplasia”,
Sci. Rep. 9(1),17863(2019).hps://doi.org/10.1038/
s41598-019-54139-5
27.
A.Pardo,J.A.Guérrez-Guérrez,I.Lihacova,J.M.
López-HigueraandO.M.Conde,“Onthespectral
signatureofmelanoma:anon-parametricclassi-
caonframeworkforcancerdeteconinhyper-
spectralimagingofmelanocyclesions”,Biomed.
Opt. Express 9(12),6283–6301(2018).hps://doi.
org/10.1364/BOE.9.006283
28.
X. Hadoux, F. Hui, J.K.H. Lim, C.L. Masters, A.
Pébay,S.Chevalier,J.Ha,S.Loi,C.J.Fowler,C.
Rowe,V.L.Villemagne,E.N.Taylor,C.Fluke,J.-P.
Soucy,F.Lesage,J.-P.Sylvestre,P.Rosa-Neto,S.
Mathotaarachchi,S.Gauthier,Z.S.Nasreddine,
J.D.Arbour,M.-A.Rhéaume,S.Beaulieu,M.
Dirani,C.T.O.Nguyen,B.V.Bui,R.Williamson,J.G.
CrowstonandP.vanWijngaarden,“Non-invasivein
vivohyperspectralimagingoftherenaforpotenal
biomarkeruseinAlzheimer’sdisease”,Nat. Commun.
10,4227(2019).hps://doi.org/10.1038/s41467-
019-12242-1
29.
S.Ortega,H.Fabelo,D.K.Iakovidis,A.Koulaouzidis
andG.M.Callico,“Useofhyperspectral/mulspec-
tral imaging in gastroenterology. shedding some-
dierent-lightintothedark”,J. Clin. Med. 8(1),36
(2019).hps://doi.org/10.3390/jcm8010036
30.
S.J. Leavesley, M. Walters, C. Lopez, T. Baker, P.F.
Favreau, T.C. Rich, P.F. Rider and C.W. Boudreaux,
“Hyperspectralimaginguorescenceexcitaon
scanningforcoloncancerdetecon”,J. Biomed. Opt.
21(10),104003(2016).hps://doi.org/10.1117/1.
JBO.21.10.104003
31.
E.Kho,L.L.deBoer,K.K.VandeVijver,F.van
Duijnhoven,M.VranckenPeeters,H.Sterenborg
andT.J.M.Ruers,“Hyperspectralimagingforresec-
onmarginassessmentduringcancersurgery”,Clin.
Cancer Res. 25(12),3572–3580(2019).hps://doi.
org/10.1158/1078-0432.CCR-18-2089
32.
T. Du, D.K. Mishra, L. Shmuylovich, A. Yu, H.
Hurbon, S.T. Wang and M.Y. Berezin, “Hyperspectral
imagingandcharacterizaonofallergiccon-
tactdermasintheshort-waveinfrared”,J.
Biophoton. 13(9), e202000040 (2020). hps://doi.
org/10.1002/jbio.202000040
33.
M.Lukina,K.Yashin,E.Kisileva,A.Alekseeva,
V.Dudenkova,E.V.Zagaynova,E.Bederina,I.
Medyanik, W. Becker, D. Mishra, M.Y. Berezin,
V.I.ShcheslavskiyandM.Shirmanova,“Label-free
macroscopicuorescencelifemeimagingofbrain
tumors”,Front. Oncol. in press (2021). hps://doi.
org/10.3389/fonc.2021.666059
34.
A.Lowe,N.HarrisonandA.P.French,
“Hyperspectralimageanalysistechniquesforthe
deteconandclassicaonoftheearlyonsetof
plantdiseaseandstress”,Plant Meth. 13(1), 80
(2017).hps://doi.org/10.1186/s13007-017-0233-z
35.
P.V.Manley,V.Sagan,F.B.FritschiandJ.G.Burken,
“Remotesensingofexplosives-inducedstressin
plants:hyperspectralimaginganalysisforremote
deteconofunexplodedthreats”,Remote Sens.
11(15),1827(2019).hps://doi.org/10.3390/
rs11151827
36.
D.M. Kim, H. Zhang, H. Zhou, T. Du, Q. Wu,
T.C.MocklerandM.Y.Berezin,“Highlysensive
image-derivedindicesofwater-stressedplants
using hyperspectral imaging in SWIR and histo-
gramanalysis”,Sci. Rep. 5,15919(2015).hps://doi.
org/10.1038/srep15919
12 IDCubeLite:FreeInteractiveDiscoveryCubeSoftwareforMulti-andHyperspectralApplications
37.
A.F.Goetz,“Threedecadesofhyperspectralremote
sensingoftheEarth:apersonalview”,Remote
Sens. Environ. 113,S5–S16(2009).hps://doi.
org/10.1016/j.rse.2007.12.014
38.
J.S. Pearlman, P.S. Barry, C.C. Segal, J. Shepanski,
D. Beiso and S.L. Carman, “Hyperion, a space-
basedimagingspectrometer”,IEEE Trans. Geosci.
Remote Sens. 41(6),1160–1173(2003).hps://doi.
org/10.1109/TGRS.2003.815018
39.
M.Drusch,U.DelBello,S.Carlier,O.Colin,V.
Fernandez, F. Gascon, B. Hoersch, C. Isola, P.
LaberinandP.Marmort,“Sennel-2:ESA’sop-
calhigh-resoluonmissionforGMESoperaonal
services”Remote Sens. Environ. 120,25–36(2012).
hps://doi.org/10.1016/j.rse.2011.11.026
40.
OrbitaAerospace.hps://www.obtdata.com/en/
41.
R.O.Green,M.L.Eastwood,C.M.Sarture,T.G.
Chrien, M. Aronsson, B.J. Chippendale, J.A. Faust,
B.E.Pavri,C.J.ChovitandM.Solis,“Imagingspec-
troscopyandtheairbornevisible/infraredimaging
spectrometer(AVIRIS)”,Remote Sens. Environ. 65(3),
227–248(1998).hps://doi.org/10.1016/S0034-
4257(98)00064-9
42.
L.Hamlin,R.Green,P.Mouroulis,M.Eastwood,D.
Wilson, M. Dudik and C. Paine, “Imaging spectrom-
etersciencemeasurementsforterrestrialecology:
AVIRISandnewdevelopments”,2011 Aerospace
Conference,pp.1–7(2011).hps://doi.org/10.1109/
AERO.2011.5747395
