TY - JOUR
T1 - Advancing bag-of-visual-words representations for lesion classification in retinal images
AU - Pires, Ramon
AU - Jelinek, Herbert F.
AU - Wainer, Jacques
AU - Valle, Eduardo
AU - Rocha, Anderson
N1 - Funding Information:
The authors have the following interests: This work was supported in part by a research grant (2009–2011) from Microsoft Research ( http://research.microsoft.com ). It was an agreement between Microsoft Research and the São Paulo Research Foundation FAPESP in which they funded interesting research through the Microsoft Research–FAPESP Institute for IT Research ( http://www.fapesp.br/en/5392 ). The institute supports high-quality fundamental research in information and communication technologies that is geared towards addressing social and economic development needs of the region. This work was also supported in part by Samsung Eletrônica da Amazônia ( http://www.samsung.com ). It is a scholarship program established by Samsung and our institution (Institute of Computing – http://ic.unicamp.br ) to finance hard-working students. This does not alter the authors’ adherence to all the PLoS ONE policies on sharing data and materials.
PY - 2014/6/2
Y1 - 2014/6/2
N2 - Diabetic Retinopathy (DR) is a complication of diabetes that can lead to blindness if not readily discovered. Automated screening algorithms have the potential to improve identification of patients who need further medical attention. However, the identification of lesions must be accurate to be useful for clinical application. The bag-of-visual-words (BoVW) algorithm employs a maximum-margin classifier in a flexible framework that is able to detect the most common DR-related lesions such as microaneurysms, cotton-wool spots and hard exudates. BoVW allows to bypass the need for pre- and post-processing of the retinographic images, as well as the need of specific ad hoc techniques for identification of each type of lesion. An extensive evaluation of the BoVW model, using three large retinograph datasets (DR1, DR2 and Messidor) with different resolution and collected by different healthcare personnel, was performed. The results demonstrate that the BoVW classification approach can identify different lesions within an image without having to utilize different algorithms for each lesion reducing processing time and providing a more flexible diagnostic system. Our BoVW scheme is based on sparse low-level feature detection with a Speeded-Up Robust Features (SURF) local descriptor, and mid-level features based on semisoft coding with max pooling. The best BoVW representation for retinal image classification was an area under the receiver operating characteristic curve (AUC-ROC) of 97.8% (exudates) and 93.5% (red lesions), applying a cross-dataset validation protocol. To assess the accuracy for detecting cases that require referral within one year, the sparse extraction technique associated with semi-soft coding and max pooling obtained an AUC of 94.2±2.0%, outperforming current methods. Those results indicate that, for retinal image classification tasks in clinical practice, BoVW is equal and, in some instances, surpasses results obtained using dense detection (widely believed to be the best choice in many vision problems) for the low-level descriptors.
AB - Diabetic Retinopathy (DR) is a complication of diabetes that can lead to blindness if not readily discovered. Automated screening algorithms have the potential to improve identification of patients who need further medical attention. However, the identification of lesions must be accurate to be useful for clinical application. The bag-of-visual-words (BoVW) algorithm employs a maximum-margin classifier in a flexible framework that is able to detect the most common DR-related lesions such as microaneurysms, cotton-wool spots and hard exudates. BoVW allows to bypass the need for pre- and post-processing of the retinographic images, as well as the need of specific ad hoc techniques for identification of each type of lesion. An extensive evaluation of the BoVW model, using three large retinograph datasets (DR1, DR2 and Messidor) with different resolution and collected by different healthcare personnel, was performed. The results demonstrate that the BoVW classification approach can identify different lesions within an image without having to utilize different algorithms for each lesion reducing processing time and providing a more flexible diagnostic system. Our BoVW scheme is based on sparse low-level feature detection with a Speeded-Up Robust Features (SURF) local descriptor, and mid-level features based on semisoft coding with max pooling. The best BoVW representation for retinal image classification was an area under the receiver operating characteristic curve (AUC-ROC) of 97.8% (exudates) and 93.5% (red lesions), applying a cross-dataset validation protocol. To assess the accuracy for detecting cases that require referral within one year, the sparse extraction technique associated with semi-soft coding and max pooling obtained an AUC of 94.2±2.0%, outperforming current methods. Those results indicate that, for retinal image classification tasks in clinical practice, BoVW is equal and, in some instances, surpasses results obtained using dense detection (widely believed to be the best choice in many vision problems) for the low-level descriptors.
UR - http://www.scopus.com/inward/record.url?scp=84902336222&partnerID=8YFLogxK
U2 - 10.1371/journal.pone.0096814
DO - 10.1371/journal.pone.0096814
M3 - Article
C2 - 24886780
AN - SCOPUS:84902336222
SN - 1932-6203
VL - 9
JO - PLoS ONE
JF - PLoS ONE
IS - 6
M1 - e96814
ER -
