Performance Evaluation of CADe Algorithms in Mammography
Abstract
Mammography plays a central role in the process of detecting abnormalities in breast cancer screening. A mammogram is a x-ray projection of the 3D structures of the breast, obtained by compressing the breast between two plates. Unlike most other x-ray or computed tomography images, mammograms have an inherent "€œfuzzy"€ or diffuse appearance. This is due in part to the superimposition of densities from differing breast tissues, and the differential x-ray attenuation (absorption) characteristics associated with these various tissues. Mammograms are essentially transparencies, where the superposition of structures located in different planes produces ambiguity in the image. For example, two blood vessels in different planes may combine to produce a single masslike structure that can be mistaken for an abnormality. One of the limitations of conventional screening mammography is that the false-negative rate is inappropriately high. This reduced sensitivity is likely due to suboptimal performance in perception of lesions and analysis of perceived findings, with the result being "€œmissed" cancers. Studies suggest that 10-€“30% of cancers that could have been detected are in fact missed. Double reading of mammograms has been advocated as a means of increasing sensitivity. Studies have shown that different readers overlook different findings, and that having more than one person interpret the images from an examination can increase the cancer detection rate as much as 15%. The use of computer prompting to increase the sensitivity in screening has gained increasing attention. Computer-aided detection (CADe) is designed to provide a radiologist with visual prompts on a series of mammograms. It works by marking a mammogram with marks that indicate regions where the detection algorithm recognizes a suspicious entity that warrants further investigation, thereby complementing the radiologists' interpretation. CADe in breast cancer is the application of computational techniques to the problem of interpreting breast images, usually mammograms. CADe, in the purist sense, only really exists for mammography. While computer-aided systems such as CADstream are evolving for modalities, such as magnetic resonance (MR), they usually concentrate more on diagnosis, with limited detection performed by means of temporal analysis.
Online access to SPIE eBooks is limited to subscribing institutions.
Lens.org Logo
CITATIONS
Cited by 12 scholarly publications.
Advertisement
Advertisement
RIGHTS & PERMISSIONS
Get copyright permission  Get copyright permission on Copyright Marketplace
KEYWORDS
Mammography

Computer aided diagnosis and therapy

Breast

Cancer

Breast cancer

Computing systems

Tissues

Back to Top