Image Processing

Image processing is a crucial element of modern digital mammography. Optimizing mammogram presentation may lead to more efficient reading and improved diagnostic performance. Despite that the effects of image processing are often much larger than those of

  • PDF / 1,911,688 Bytes
  • 15 Pages / 547.087 x 765.354 pts Page_size
  • 50 Downloads / 240 Views

DOWNLOAD

REPORT


Image Processing Nico Karssemeijer and Peter R. Snoeren

CONTENTS

KEY POINTS 5.1

Introduction

5.2

Grayscale Transforms 70

5.3 5.3.1 5.3.2 5.3.3 5.3.4

Spatial Enhancement 72 Unsharp Masking 72 Adaptive Histogram Equalization 73 Multiscale Image Enhancement 74 Peripheral Enhancement 75

5.4

Matching Current and Prior Mammograms 77

5.5

Physics-Based Methods 80

5.6

Evaluation of Mammogram Processing 81

References

69

82

Image processing is a crucial element of modern digital mammography. Optimizing mammogram presentation may lead to more efficient reading and improved diagnostic performance. Despite that the effects of image processing are often much larger than those of acquisition parameter settings, little is known about how image processing can be optimized. Experts agree that comparison of features in various mammographic views is very important. This issue must be addressed by processing. Variation of image presentation across views and subsequent mammograms should be minimized. The dynamic range of electronic displays is limited. Therefore, processing techniques should be designed to limit the dynamic range of mammograms. This can effectively be done by applying peripheral enhancement in the uncompressed tissue region near the projected skin–air interface. Adaptive contrast enhancement can be applied to enhance microcalcifications and dense tissue in the interior of the mammogram. Mammogram processing should be aimed at displaying all relevant information in good contrast simultaneously, as human interaction to manipulate contrast during reading is too time-consuming to be applied on a regular basis.

5.1 Nico Karssemeijer, PhD Department of Radiology, Radboud University Nijmegen Medical Center, PO Box 9101, 6500HB Nijmegen, The Netherlands Peter R. Snoeren, PhD Department of Radiology, Radboud University Nijmegen Medical Center, PO Box 9101, 6500HB Nijmegen, The Netherlands

Introduction The goal of mammography is to detect and diagnose breast cancer, a task which is generally performed by experienced and skilled radiologists. For optimal reader performance, mammograms have to be matched to the human visual system when they are

70

N. Karssemeijer and P. R. Snoeren

presented. In other words, characteristic features of cancer have to be displayed with optimal contrast to avoid being missed or misinterpreted. Within the limited possibilities of conventional screen-film mammography, this matching problem has received much attention in the past. Many innovations have been made in mammography to enhance visibility of cancers. Most notable in this respect is the gradual increase of contrast in the interior region of the breast at the cost of contrast in the periphery, where cancer seldom occurs. This change was only possible due to development of more accurate automatic exposure control (AEC) devices. With the latest generation of mammography films, the skinline and large parts of the periphery are hardly visible on mammography film alternators. With the introduction of digital mammography, a wide range o