Skip to content
DraconiticResearch
Search Sources API
203,948 studies31 sourcesUpdated 27 min ago
Peer-reviewedCC BY-NC-ND 4.0Cardiovascular Diagnosis and Therapy

A comparative study of the state-of-the-art artificial intelligence models for automated plaque measurement in a carotid ultrasound Japanese cohort

  • Nikhil Singh1,2
  • Pankaj K. Jain3,4
  • Rubeena Vohra2
  • Arun K. Dubey5
  • Mainak Biswas6
  • Ekta Tiwari7
  • Vinod Anand1
  • Inder M. Singh1
  • Esma R Isenovic8
  • Amer M. Johri9
  • Laura E. Mantella10
  • John R Laird11
13 more authors
  • Narendra N. Khanna12
  • Zoltan Ruzsa13
  • Rajesh Singh14
  • Sanjay Saxena15
  • Mustafa Al Maini16
  • Klaudija Viskovic17
  • Manudeep K Kalra18
  • Andrew Nicolaides19
  • George Kitas20
  • Vijay Viswanathan21
  • Janet N A Ajuluchukwu1
  • Luca Saba22
  • Jasjit S. Suri1,23,24
24 affiliations
  1. 1Stroke Monitoring and Diagnostic Division, AtheroPoint™, Roseville, CA, USA
  2. 2Department of Electronics and Communication Engineering, Bharati Vidyapeeth’s College of Engineering, New Delhi, India
  3. 3School of Artificial Intelligence, Sai University, Chennai, India
  4. 4School of Biomedical Engineering, IIT (BHU), Varanasi, India
  5. 5Department of Computer Science Engineering, Samrat Ashok Rajkiya Engineering College (SAREC), Uttar Pradesh, India
  6. 6School of Computer Engineering, Kalinga Institute of Industrial Technology, Bhubaneswar, India
  7. 7Department of Innovation, Global Biotechnologies, Roseville, CA, USA
  8. 8Department of Radiobiology and Molecular Genetics, “VINČA” Institute of Nuclear Sciences - National Institute of the Republic of Serbia, University of Belgrade, Belgrade, Serbia
  9. 9Division of Cardiology, Department of Medicine, Queen’s University, Kingston, ON, Canada
  10. 10Department of Biomedical and Molecular Sciences, Queen’s University, Kingston, ON, Canada
  11. 11Heart and Vascular Institute, Adventist Health St. Helena, St Helena, CA, USA
  12. 12Department of Cardiology, Indraprastha APOLLO Hospitals, New Delhi, India
  13. 13Invasive Cardiology Division, University of Szeged, Szeged, Hungary
  14. 14Department of Research and Innovation, Uttaranchal Institute of Technology, Uttaranchal University, Dehradun, India
  15. 15Department of CSE, IIIT, Bhubaneswar, India
  16. 16Allergy, Clinical Immunology and Rheumatology Institute, Toronto, ON, Canada
  17. 17University Hospital for Infectious Diseases, Zagreb, Croatia
  18. 18Department of Radiology, Massachusetts General Hospital, Boston, MA, USA
  19. 19Vascular Screening and Diagnostic Centre, University of Nicosia, Nicosia, Cyprus
  20. 20Department of Rheumatology, University of Manchester, Dudley, UK
  21. 21MV Diabetes Centre, Chennai, India
  22. 22Department of Neurology, University of Cagliari, Cagliari, Italy
  23. 23Department of ECE, Idaho State University, Pocatello, ID, USA
  24. 24Symbiosis Institute of Technology, Nagpur Campus, Symbiosis International (Deemed University), Pune, India
Read full textDOICite

Abstract

Background: Automated carotid intima-media thickness (cIMT) measurement from B-mode ultrasound is important for cardiovascular risk stratification. Direct comparisons of deep learning architectures on the same cohort under controlled conditions remain limited. This study aimed to compare the accuracy and computational cost of three deep learning architectures with the commercial AtheroEdge™ 2.0 (AE 2.0) system for carotid ultrasound analysis.

Methods: This study evaluated three deep learning architectures: carotid non-attention double-stage convolutional neural network (cNonAtt-dsCNN), carotid attention-based single-stage UNet (cAtt-ssUNet), and carotid attention-based double-stage UNet (cAtt-dsUNet) against the commercial AE 2.0 system on 400 carotid ultrasound images from 200 Japanese patients. All deep learning models were trained using identical preprocessing, augmentation, hyperparameters (Adam optimizer, learning rate =1e−4, batch =32), and 5-fold cross-validation. Evaluation metrics included Dice similarity coefficient (DSC), Jaccard index, lumen-intima error (LIe), media-adventitia error (MAe), intima-media thickness error (IMTe), and computational cost.

Results: Under identical conditions, cAtt-dsUNet achieved the highest accuracy [DSC: 91.2%, Jaccard index: 85.5%, correlation coefficient (CC): 0.98, LIe: 0.031 mm, IMTe: 0.022 mm], reaching commercial-grade overlap and correlation but at ~2.5× longer training (20.2 vs. 8 min) and ~1.7× more graphics processing unit (GPU) memory than cNonAtt-dsCNN (DSC: 81.5%, CC: 0.92, LIe: 0.09 mm, IMTe: 0.124 mm). cAtt-ssUNet was intermediate (DSC: 88.5%, CC: 0.95, LIe: 0.07 mm). AE 2.0 retained the lowest boundary errors (LIe: 0.012 mm, IMTe: 0.011 mm) with 6 ms inference.

Conclusions: A clear trade-off between accuracy-efficiency exists across the evaluated methods. cAtt-dsUNet is optimal for high-accuracy research settings with adequate GPU resources, while cNonAtt-dsCNN and AE 2.0 are preferable for speed-critical clinical screening. The single-cohort Japanese dataset limits generalizability; multi-center, multi-ethnic validation is recommended before clinical deployment.

Related

  1. Artificial Intelligence for Automated Analysis and Interpretation of Intravascular Ultrasound in Coronary Artery Disease: A Systematic Review
    Ashwin Umasankar, Heramba Vigneshwar Sachithananthan, Anusha Chitravanshi +1
    CureusPeer-reviewedCC BYFull textCardiovascular medicineEurope PMCEurope PMCPMC
  2. Semi-Automated Measurement of Central Retinal Artery Blood Flow Using Doppler Ultrasound Coupled With Electrocardiography
    Keren Wood, Alon Harris, Temitope Olayinka +7
    Translational Vision Science & TechnologyPeer-reviewedCC BY-NC-ND 4.0Full textMedical imaging & radiologyPMC
  3. Artificial Intelligence‐Assisted Automated Optical Coherence Tomography Analysis for Percutaneous Coronary Intervention Planning: From Plaque Analysis to Precision Lesion Preparation and Stent Optimization
    Kevin Mohee, Eirinaios Tsiartas, Jason Sangha +2
    Journal of the American Heart AssociationPeer-reviewedLicence not statedCardiovascular medicineCrossref
  4. Artificial intelligence-based prostate cancer risk stratification via large-scale ultrasound images paired with histopathology
    Qiong Zhang, Qiang Lu, Yu Liang +3
    Quantitative imaging in medicine and surgeryPeer-reviewedCC BY-NC-NDFull textMedical imaging & radiologyEurope PMC
  5. Validating Artificial Intelligence Guidance for Ultrasound Acquisition and Remote Interpretation
    T. Maldonado, S. Muluk, P. Rali +14
    medRxivPreprintAll rights reservedFull textMedical imaging & radiologymedRxiv

Cite

BibTeXRIS
BibTeX
@article{singh2026comparative,
  title = {{A comparative study of the state-of-the-art artificial intelligence models for automated plaque measurement in a carotid ultrasound Japanese cohort}},
  author = {Nikhil Singh and Pankaj K. Jain and Rubeena Vohra and Arun K. Dubey and Mainak Biswas and Ekta Tiwari and Vinod Anand and Inder M. Singh and Esma R Isenovic and Amer M. Johri and Laura E. Mantella and John R Laird and Narendra N. Khanna and Zoltan Ruzsa and Rajesh Singh and Sanjay Saxena and Mustafa Al Maini and Klaudija Viskovic and Manudeep K Kalra and Andrew Nicolaides and George Kitas and Vijay Viswanathan and Janet N A Ajuluchukwu and Luca Saba and Jasjit S. Suri},
  journal = {Cardiovascular Diagnosis and Therapy},
  year = {2026},
  month = oct,
  doi = {10.21037/cdt-2026-0347},
  url = {https://doi.org/10.21037/cdt-2026-0347},
  note = {CC BY-NC-ND 4.0},
}
RIS
TY  - JOUR
TI  - A comparative study of the state-of-the-art artificial intelligence models for automated plaque measurement in a carotid ultrasound Japanese cohort
AU  - Nikhil Singh
AU  - Pankaj K. Jain
AU  - Rubeena Vohra
AU  - Arun K. Dubey
AU  - Mainak Biswas
AU  - Ekta Tiwari
AU  - Vinod Anand
AU  - Inder M. Singh
AU  - Esma R Isenovic
AU  - Amer M. Johri
AU  - Laura E. Mantella
AU  - John R Laird
AU  - Narendra N. Khanna
AU  - Zoltan Ruzsa
AU  - Rajesh Singh
AU  - Sanjay Saxena
AU  - Mustafa Al Maini
AU  - Klaudija Viskovic
AU  - Manudeep K Kalra
AU  - Andrew Nicolaides
AU  - George Kitas
AU  - Vijay Viswanathan
AU  - Janet N A Ajuluchukwu
AU  - Luca Saba
AU  - Jasjit S. Suri
PY  - 2026
DA  - 2026/10/20
JO  - Cardiovascular Diagnosis and Therapy
DO  - 10.21037/cdt-2026-0347
UR  - https://doi.org/10.21037/cdt-2026-0347
AB  - Background: Automated carotid intima-media thickness (cIMT) measurement from B-mode ultrasound is important for cardiovascular risk stratification. Direct comparisons of deep learning architectures on the same cohort under controlled conditions remain limited. This study aimed to compare the accuracy and computational cost of three deep learning architectures with the commercial AtheroEdge™ 2.0 (AE 2.0) system for carotid ultrasound analysis.  Methods: This study evaluated three deep learning architectures: carotid non-attention double-stage convolutional neural network (cNonAtt-dsCNN), carotid attention-based single-stage UNet (cAtt-ssUNet), and carotid attention-based double-stage UNet (cAtt-dsUNet) against the commercial AE 2.0 system on 400 carotid ultrasound images from 200 Japanese patients. All deep learning models were trained using identical preprocessing, augmentation, hyperparameters (Adam optimizer, learning rate =1e−4, batch =32), and 5-fold cross-validation. Evaluation metrics included Dice similarity coefficient (DSC), Jaccard index, lumen-intima error (LIe), media-adventitia error (MAe), intima-media thickness error (IMTe), and computational cost.  Results: Under identical conditions, cAtt-dsUNet achieved the highest accuracy [DSC: 91.2%, Jaccard index: 85.5%, correlation coefficient (CC): 0.98, LIe: 0.031 mm, IMTe: 0.022 mm], reaching commercial-grade overlap and correlation but at ~2.5× longer training (20.2 vs. 8 min) and ~1.7× more graphics processing unit (GPU) memory than cNonAtt-dsCNN (DSC: 81.5%, CC: 0.92, LIe: 0.09 mm, IMTe: 0.124 mm). cAtt-ssUNet was intermediate (DSC: 88.5%, CC: 0.95, LIe: 0.07 mm). AE 2.0 retained the lowest boundary errors (LIe: 0.012 mm, IMTe: 0.011 mm) with 6 ms inference.  Conclusions: A clear trade-off between accuracy-efficiency exists across the evaluated methods. cAtt-dsUNet is optimal for high-accuracy research settings with adequate GPU resources, while cNonAtt-dsCNN and AE 2.0 are preferable for speed-critical clinical screening. The single-cohort Japanese dataset limits generalizability; multi-center, multi-ethnic validation is recommended before clinical deployment.
SN  - 2223-3652
SN  - 2223-3660
ER  -