A Hand-held Mosaicked Multispectral Imaging Device for Early Stage Pressure Ulcer Detection

  • Authors:

  • Hairong Qi;Linghua Kong;Chao Wang;Lidan Miao

  • Affiliations:

  • Electrical Engineering and Computer Science Department, University of Tennessee, Knoxville, USA 37996;Center for Assistive Technology and Environmental Access, Georgia Institute of Technology, Atlanta, USA 30318;Center for Assistive Technology and Environmental Access, Georgia Institute of Technology, Atlanta, USA 30318;Microsoft, Inc., Seattle, USA

  • Venue:
  • Journal of Medical Systems
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

The use of a custom filter mosaic overlaying a CMOS/CCD sensor represents a novel idea to multispectral imaging. The innovation provides simple, miniaturized, low cost instrumentation that has many potential biological applications which require a hand-held detector. This makes it extremely adaptable and can serve as an integrated component to distributed diagnosis and home healthcare (D2H2). A mosaicked sensor is a monolithic array of many sensors, arranged in a geometric pattern with each sensor covered by an optical filter sensitive to a specified wavelength. In this way, only one spectral component is sensed at each pixel and the other spectral components must be estimated from neighbors. Although with great potential, one challenge faced by this device, however, is the reconstruction of the high-resolution full-spectral image from the low-resolution input. Due to the physical limitations in fabrication and the usage of the multispectral filter mosaic, two types of degradations exist, including filter misalignment and the missing spectral components, that must be corrected using intelligent algorithms to take full advantage of the hardware capability of the device. In this paper, we first describe a custom geometric correction method to restore the image from the misalignment distortion. We then present a binary tree-based generic demosaicking algorithm to efficiently estimate the missing special components and reconstruct a high-resolution full-spectral image. We choose early detection of pressure ulcer as a targeting area as early stage pressure ulcers and other subcutaneous lesions are nearly invisible in clinical settings, particularly so for dark pigmented skin. We show how the geometric correction and demosaicking algorithms successfully reconstruct a full-spectral image from which apparent contrast enhancement between damaged skin and the normal skin is observed.