Faster super-resolution ultrasound imaging with a deep learning model for tissue decluttering and contrast agent localization

Katherine G. Brown, Scott Chase Waggener, Arthur David Redfern, Kenneth Hoyt

Research output: Contribution to journalArticlepeer-review

Abstract

Super-resolutionultrasound(SR-US)imagingallowsvisualizationofmicrovascularstructuresassmallas tensofmicrometersindiameter.However,useintheclinicalsettinghasbeenimpededinpartbyultrasound (US)acquisitiontimesexceedingabreath-holdandbytheneedforextensiveofflinecomputation.Deep learningtechniqueshavebeenshowntobeeffectiveinmodelingthetwomorecomputationallyintensive stepsofmicrobubble(MB)contrastagentdetectionandlocalization.Performancegainsbydeepnetworks overconventionalmethodsaremorethantwoordersofmagnitudeandinadditionthenetworkscanlocalize overlappingMBs.TheabilitytoseparateoverlappingMBsallowsuseofhighercontrastagentconcentrations andreducesUSimageacquisitiontime.Hereinweproposeafullyconvolutionalneuralnetwork(CNN) architecturetoperformtheoperationsofMBdetectionaswellaslocalizationinasinglemodel.Termed SRUSnet,thenetworkisbasedontheMobileNetV3architecturemodifiedfor3-Dinputdata,minimal convergencetime,andhigh-resolutiondataoutputusingaflexibleregressionhead.Also,weproposeto combinelinearB-modeUSimagingandnonlinearcontrastpulsesequencing(CPS)whichhasbeenshown toincreaseMBdetectionandfurtherreducetheUSimageacquisitiontime.Thenetworkwastrainedwithin silicodataandtestedoninvitrodatafromatissue-mimickingflowphantom,andoninvivodatafromtherat hindlimb(N = 3).ImageswerecollectedwithaprogrammableUSsystem(Vantage256,VerasonicsInc., Kirkland,WA)usinganL11–4vlineararraytransducer.Thenetworkexceeded99.9%detectionaccuracyon insilicodata.Theaveragelocalizationaccuracywassmallerthantheresolutionofapixel(i.e.l/8).The averageprocessingtimeonaNvidiaGeForce2080TiGPUwas64.5msfora128 × 128-pixelimage.

Original languageEnglish (US)
Article number065035
JournalBiomedical Physics and Engineering Express
Volume7
Issue number6
DOIs
StatePublished - Nov 2021
Externally publishedYes

Keywords

  • Contrast-enhanced ultrasound
  • Deep learning
  • Microbubbles
  • Plane-waves
  • Super-resolution ultrasound

ASJC Scopus subject areas

  • Biophysics
  • Bioengineering
  • Biomaterials
  • Physiology
  • Biomedical Engineering
  • Radiology Nuclear Medicine and imaging
  • Computer Science Applications
  • Health Informatics

Fingerprint

Dive into the research topics of 'Faster super-resolution ultrasound imaging with a deep learning model for tissue decluttering and contrast agent localization'. Together they form a unique fingerprint.

Cite this