Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2017 Feb 15:7:42703.
doi: 10.1038/srep42703.

Automated Detection of Malarial Retinopathy in Digital Fundus Images for Improved Diagnosis in Malawian Children with Clinically Defined Cerebral Malaria

Vinayak Joshi  1 , Carla Agurto  1 , Simon Barriga  1 , Sheila Nemeth  1 , Peter Soliz  1 , Ian J MacCormick  2   3   4 , Susan Lewallen  5 , Terrie E Taylor  6   7 , Simon P Harding  2
Affiliations

Automated Detection of Malarial Retinopathy in Digital Fundus Images for Improved Diagnosis in Malawian Children with Clinically Defined Cerebral Malaria

Vinayak Joshi et al. Sci Rep. .

Abstract

Cerebral malaria (CM), a complication of malaria infection, is the cause of the majority of malaria-associated deaths in African children. The standard clinical case definition for CM misclassifies ~25% of patients, but when malarial retinopathy (MR) is added to the clinical case definition, the specificity improves from 61% to 95%. Ocular fundoscopy requires expensive equipment and technical expertise not often available in malaria endemic settings, so we developed an automated software system to analyze retinal color images for MR lesions: retinal whitening, vessel discoloration, and white-centered hemorrhages. The individual lesion detection algorithms were combined using a partial least square classifier to determine the presence or absence of MR. We used a retrospective retinal image dataset of 86 pediatric patients with clinically defined CM (70 with MR and 16 without) to evaluate the algorithm performance. Our goal was to reduce the false positive rate of CM diagnosis, and so the algorithms were tuned at high specificity. This yielded sensitivity/specificity of 95%/100% for the detection of MR overall, and 65%/94% for retinal whitening, 62%/100% for vessel discoloration, and 73%/96% for hemorrhages. This automated system for detecting MR using retinal color images has the potential to improve the accuracy of CM diagnosis.

PubMed Disclaimer

Conflict of interest statement

This research work was funded through an NIH grant awarded to VisionQuest Biomedical LLC. Joshi, Agurto, Nemeth, Barriga, and Soliz are employees or have financial interests in VisionQuest Biomedical LLC., and could potentially benefit from this research work. Taylor, MacCormick, Harding, and Lewallen do not have competing financial interests.

Figures

Figure 1
Figure 1
(a) Proposed CM diagnosis using MR detection, (b) CM clinical management using MR detection.
Figure 2
Figure 2
(a) Retinal image mosaic, (b) Grader’s annotation of MR lesions.
Figure 3
Figure 3
(a) Vessel discoloration, (b) ground truth annotation.
Figure 4
Figure 4
(a) Enhanced green channel, (b) Enhanced ‘a’ channel.
Figure 5
Figure 5
Image showing: (a) discolored vessels (green) annotated by grader, (b) discolored vessels detected (blue) or missed (yellow) by the algorithm.
Figure 6
Figure 6
MR hemorrhages: (a) Fundus image, (b) Grader’s annotation.
Figure 7
Figure 7
Hemorrhage detection using: (a) Supervised classification, (b) Unsupervised classification, (c) Hybrid method. Each image shows false positive (magenta) and false negative (yellow) detections.
Figure 8
Figure 8
(a) Retinal whitening and camera reflex, (b) Grader’s annotation for whitening in blue and reflex in shiny green color.
Figure 9
Figure 9
(a) Retinal image, (b) Preprocessed image with reflex minimization.
Figure 10
Figure 10
(a) Retinal image, (b) Image splats (regions) formed by watershed segmentation.
Figure 11
Figure 11
MR whitening detection: (a) Retinal image, (b) Grader’s annotation of whitening, (c) Image splats with assigned probability of whitening.
Figure 12
Figure 12
ROC curves for: (a) vessel discoloration detection, (b) hemorrhage detection, (c) whitening detection.
Figure 13
Figure 13
(a) Retinal image, (b) Automated hemorrhage detection.
Figure 14
Figure 14
(a) Retinal image, (b) Ground truth annotation, (c) Vessel discoloration detection.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. UNICEF fact sheet: Malaria, A Global Crisis. http://www.unicef.org/media/media_20475.html (Date of access: 30/11/2016) (2004).
    1. World Malaria Report: World Health Organization global malaria program. (Date of access: 30/11/2016) (2012).
    1. Trampuz A., Jereb M., Muzlovic I. & Prabhu R. Clinical review: severe malaria. Critical Care. 7, 315–323 (2003). - PMC - PubMed
    1. Essuman V., Ntim-Amponsah C. & Astrup B. Retinopathy in severe malaria in Ghanaian children: overlap between fundus changes in cerebral and non-cerebral malaria. Malar. J. 12, 9–232 (2010). - PMC - PubMed
    1. Taylor T. E. et al.. Differentiating the pathologies of cerebral malaria by postmortem parasite counts. Nat. Med. 10, 143–145 (2004). - PubMed

Publication types

MeSH terms