Detecting false vessel recognitions in retinal fundus analysis

Abstract

Automatic tracking of blood vessels in images of retinal fundus is an important and non-invasive procedure for the diagnosis of many diseases. Tracking techniques often present a high rate of false positives. This paper presents six methods to discriminate false detections from true positives, each based on a different model of the vessel. They describe a candidate vessel in terms of its average geometric and grayscale properties considered along the full trajectory of the vessel itself. The rationale is that false vessels are caused by the small scale of the tracking algorithm necessary during the tracking phase. Once tracking has been completed, we can gather information from the full vessel trajectory and solve ambiguities that cannot be fixed during tracking. We apply Fisher linear discriminant analysis to these features to get the desired discrimination. Results on 28 images show satisfactory rejection of false positives and better results when using more complex models.

Automatic refinement of vascular tracking in retinal images: False vessels detection

“…The indexes have been calculated considering the vessel units and not the pixel units. Results reported in [9] (i.e. a sensitivity of 0.8 and a TPV of 0.96 for the training dataset, and of 0.62 and 0.97 for the testing dataset) are much better than those on our images, showing a possible lack of "portability" of this algorithm.…”

“…In Table 1 the results obtained by the algorithm by Giani et al [9] as sensitivity (rate of false vessels correctly detected) and True Positive Value (rate of groundtruth false vessel among the positives), applied to our datasets have been reported for two significant values of the tuning parameter w th . The indexes have been calculated considering the vessel units and not the pixel units.…”

“…a sensitivity of 0.8 and a TPV of 0.96 for the training dataset, and of 0.62 and 0.97 for the testing dataset) are much better than those on our images, showing a possible lack of "portability" of this algorithm. These question- Table 1: Results of the method by Giani et al [9] applied to our dataset. In the last two columns we report their results as presented in [9].…”

“…These question- Table 1: Results of the method by Giani et al [9] applied to our dataset. In the last two columns we report their results as presented in [9]. The algorithm works in different ways according to the images analyzed .…”

“…Despite the importance of this step, only Giani et al [9] proposed a postprocessing algorithm to distinguish between true and false vessels. It was based on a model of the vessel considering the average geometric and grayscale properties along the full trajectory of the vessel itself.…”