Journal article
Contextual computer-aided detection: Improving bright lesion detection in retinal images and coronary calcification identification in CT scans
Medical Image Analysis, Vol.16(1), pp.50-62
01/2012
DOI: 10.1016/j.media.2011.05.004
PMID: 21689964
Abstract
FROC curves for the classification performance of hard exudates using local and contextual CAD systems. The diamond represents the result of the second human observer. The performance of a majority CAD system is also included. This system classifies the candidate in the same class as the majority of the candidates in the neighborhood present after local classification. Additionally, the performance of a contextual CAD system using only the proximity to red lesions as contextual information is included. The horizontal axis has a logarithmic scale. [Display omitted] ► Exploiting contextual information for lesion detection in medical images improve significantly the final classification. ► Contextual information helps to solve ambiguities in lesion classification. ► Exploiting contextual information makes the method more robust against segmentation errors. ► Using contextual information, the CAD system reached the performance obtained by human observers. ► Combining local and contextual information helps to emulate human cognitive process in medical image understanding. Contextual information plays an important role in medical image understanding. Medical experts make use of context to detect and differentiate pathologies in medical images, especially when interpreting difficult cases. The majority of computer-aided diagnosis (CAD) systems, however, employ only local information to classify candidates, without taking into account global image information or the relation of a candidate with neighboring structures. In this paper, we present a generic system for including contextual information in a CAD system. Context is described by means of high-level features based on the spatial relation between lesion candidates and surrounding anatomical landmarks and lesions of different classes (static contextual features) and lesions of the same type (dynamic contextual features). We demonstrate the added value of contextual CAD for two real-world CAD tasks: the identification of exudates and drusen in 2D retinal images and coronary calcifications in 3D computed tomography scans. Results show that in both applications contextual CAD is superior to a local CAD approach with a significant increase of the figure of merit of the Free Receiver Operating Characteristic curve from 0.84 to 0.92 and from 0.88 to 0.98 for exudates and drusen, respectively, and from 0.87 to 0.93 for coronary calcifications.
Details
- Title: Subtitle
- Contextual computer-aided detection: Improving bright lesion detection in retinal images and coronary calcification identification in CT scans
- Creators
- Clara I Sánchez - Image Sciences Institute, University Medical Center Utrecht, The NetherlandsMeindert Niemeijer - Image Sciences Institute, University Medical Center Utrecht, The NetherlandsIvana Išgum - Image Sciences Institute, University Medical Center Utrecht, The NetherlandsAlina Dumitrescu - Department of Ophthalmology and Visual Science, University of Iowa Hospitals and Clinics, USAMaria S.A Suttorp-Schulten - Ophthalmology Service, OLVG Hospital, Amsterdam, The NetherlandsMichael D Abràmoff - Department of Electrical and Computer Engineering, University of Iowa, USABram van Ginneken - Image Sciences Institute, University Medical Center Utrecht, The Netherlands
- Resource Type
- Journal article
- Publication Details
- Medical Image Analysis, Vol.16(1), pp.50-62
- Publisher
- Elsevier B.V
- DOI
- 10.1016/j.media.2011.05.004
- PMID
- 21689964
- ISSN
- 1361-8415
- eISSN
- 1361-8423
- Language
- English
- Date published
- 01/2012
- Academic Unit
- Roy J. Carver Department of Biomedical Engineering; Electrical and Computer Engineering; Stead Family Department of Pediatrics; Ophthalmology and Visual Sciences
- Record Identifier
- 9983806262202771
Metrics
27 Record Views