December 2002
Volume 43, Issue 13
Free
ARVO Annual Meeting Abstract  |   December 2002
Ultrasonic Detection of PAS-Positive Patterns Associated with Metastatic Risk in Uveal Melanoma
Author Affiliations & Notes
  • RH Silverman
    Ophthalmology Weill Medical College of Cornell University New York NY
  • R Folberg
    Pathology University of Illinois Chicago IL
  • HC Boldt
    Ophthalmology University of Iowa Iowa City IA
  • FL Lizzi
    Biomedical Engineering Directorate Riverside Research Institute New York NY
  • MJ Rondeau
    Ophthalmology Weill Medical College of Cornell University New York NY
  • HO Lloyd
    Ophthalmology Weill Medical College of Cornell University New York NY
  • DJ Coleman
    Ophthalmology Weill Medical College of Cornell University New York NY
  • Footnotes
    Commercial Relationships    R.H. Silverman, Cornell Research Foundation P; R. Folberg, None; H.C. Boldt, None; F.L. Lizzi, Riverside Research Institute P; M.J. Rondeau, None; H.O. Lloyd, None; D.J. Coleman, Cornell Research Foundation P. Grant Identification: NIH Grants EY10457, EY01212, Research to Prevent Blindness and the Dyson Foundation
Investigative Ophthalmology & Visual Science December 2002, Vol.43, 1130. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      RH Silverman, R Folberg, HC Boldt, FL Lizzi, MJ Rondeau, HO Lloyd, DJ Coleman; Ultrasonic Detection of PAS-Positive Patterns Associated with Metastatic Risk in Uveal Melanoma . Invest. Ophthalmol. Vis. Sci. 2002;43(13):1130.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Abstract: : Purpose: Certain histologic patterns (networks, loops, arcs, parallels with crosslinks) best revealed by periodic acid-Schiff (PAS) staining have been shown to act as independent predictors of metastatic risk in uveal melanoma. Because these patterns are expressed in the tumor's microarchitecture, they are not detectable in vivo by procedures such as fine needle biopsy. Our aim was to determine our ability to detect the presence of high-risk patterns in vivo using ultrasound signal processing methodologies. Methods: 117 eyes with untreated uveal melanoma were scanned and digital 10 MHz radiofrequency echo data acquired prior to enucleation. Scans and tissue sections were obtained in corresponding planes. PAS-stained sections were digitized, and areas in which networks/closed loops, arcs, and parallels with crosslinks were color-coded and superimposed on the image. Independently, spectral parameter images representing effective scatterer concentration and diameter were produced from the ultrasound data. Both histologic and ultrasonic images were automatically divided into anterior, posterior and core regions. Analysis consisted of spatial correlation of ultrasound parameters with histologic patterns, and development of multivariate models based on acoustic parameters for classification of tumors based on histologic characteristics. ROC analysis was used to evaluate classifiers. Results: Stepwise linear discriminant analysis in which tumors with 5% or more of their area consisting of closed loops and networks resulted in a model consisting of six acoustic parameters measuring scatterer size, concentration, heterogeneity and tumor size. High- and low-risk tumors were correctly identified 81%, and 83% of the time, respectively. Leave-one-out analysis showed negligible degradation (79% correct classification overall). ROC analysis of this model showed area under the curve to be 0.89 (0.82-0.97). Conclusion: Our findings demonstrate that patterns associated with metastatic risk can be detected from acoustic echo characteristics. This is likely a consequence of the spatial dimension of the patterns (on the order of one-half to one wavelength). Acoustic methods for risk-assessment are advantageous in that they are non-invasive and sample large tumor cross-sections.

Keywords: 464 melanoma • 430 imaging/image analysis: clinical • 429 image processing 
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×