A method for corneal nerves automatic segmentation and morphometric analysis

https://doi.org/10.1016/j.cmpb.2011.09.014Get rights and content

Abstract

The segmentation and morphometric analysis of corneal sub-basal nerves, from corneal confocal microscopy images, has gained recently an increased interest. This interest arises from the possibility of using changes in these nerves as the basis of a simple and non-invasive method for early detection and follow-up of peripheral diabetic neuropathy, a major cause of chronic disability in diabetic patients. Here, we propose one method for automatic segmentation and analysis of corneal nerves from images obtained in vivo through corneal confocal microscopy. The method is capable of segmenting corneal nerves, with sensitivity near 90% and a percentage of false recognitions with an average of 5.3%. The nerves tortuosity was calculated and shows statistically significant differences between healthy controls and diabetic individuals, in accordance to what is reported in the literature.

Highlights

► We developed a method for automatic segmentation of corneal nerves. ► The method uses phase-symmetry for nerve identification. ► The method sensitivity is near 90%. ► The average percentage of false recognitions is 5.3%. ► Morphology of corneal nerves can be used to diagnosis diabetic neuropathy.

Introduction

There is a growing interest on the segmentation and analysis of corneal nerves images. These images, obtained in vivo, in conscious patients, through corneal confocal microscopy (CCM), can document corneal nerve changes. There is increasing evidence that it is possible to diagnosis and assess peripheral diabetic neuropathy through the morphometric analysis of the sub-basal nerve plexus of the cornea.

The development of a simple, non-invasive method for diagnosis and follow-up of peripheral diabetic neuropathy will have great social impact. Peripheral diabetic neuropathy, is a major cause of morbidity in diabetic patients [1], [2] affecting 50% of diabetic patients 25 years after diagnosis of diabetes [3], and being implicated in 50–75% of non-traumatic amputations [4], [5]. Early diagnosis and assessment of disease progression are important to accurately define at risk patients, evaluate therapies and decrease patient morbidity [6]. Currently, this diagnosis often fails or occurs only when patients became symptomatic [7]. The follow-up, through the documentation of small-fiber damage, is only possible with invasive biopsy [8], [9], [10].

Being one of the most innervated tissues in humans and directly assessable to inspection using light, the cornea became a natural target for the diagnosis of peripheral diabetic neuropathy. This is particularly true since the development of CCM allowed non-invasive in vivo imaging of corneal sub-basal nerve plexus. CCM has been used to quantify the corneal nerve morphologic features [11], nerve density in normal and diabetic patients [12], and the extent of degeneration and regeneration of corneal nerve fibers in diabetic patients with increasing neuropathic severity [13], [14]. It was also demonstrated that the number of fibers in the cornea sub-basal nerve plexus of diabetic patients was significantly lower than in healthy humans, even for short diabetes duration [15], [16]. So, there is evidence to suggest that the quantification of sub-basal plexus morphometric parameters may constitute a basis for diabetic neuropathy diagnosis [17], [18].

At present, the analysis of sub-basal corneal nerve parameters is usually based on a manual laborious task [11], [12], [13], [19] that is subjective and prone to errors. Fully automatic methods, requiring no specific expertise from the user, were recently proposed. Holmes et al. [20] developed a method to automatically trace the nerve fibers in corneal micrographs and used it to show a relationship between the presence of peripheral neuropathy and the density and shape of corneal nerve fibers.

As far as we know, Scarpa et al. [21] were the first to propose automatic methods for the recognition and tracing of the corneal nerve structures, in CCM images [22].

In this work, we propose and evaluate the performance of a segmentation method for automatic extraction of sub-basal corneal nerves morphometric parameters. Our method starts by normalizing the image contrast and reducing the noise, using contrast equalization, a phase symmetry-based method and histogram procedure. Then a fast search is done to locate candidate regions that are expanded to identify each corneal nerve. Discarding false positives and merging all remaining nerves results in the identification of complete nerve structure [23], [24].

Section snippets

Corneal nerves segmentation algorithm

The segmentation algorithm method is schematically described by the functional block diagram of Fig. 1. The algorithm comprises several steps:

  • 1)

    Local equalization, for enhancing image details.

  • 2)

    Phase symmetry, using wavelet transform filter, to identify the nerves structures.

  • 3)

    Nerve reconstruction using morphologic operations.

These steps are detailed in the next sections.

Materials

The method described herein was implemented in Matlab and was applied on 47 images (36 healthy controls and 11 diabetic patients) from a database available online [22], containing 90 images. Two images with stromal nerves were discarded. Images whose en face section traverses obliquely several corneal layers were not considered. In these images, the sub-basal nerves appear only on a small region, with the remaining field of view containing other corneal features, such as stromal keratocytes or

Results

Corneal nerves segmentation results for the developed algorithm are illustrated in Fig. 3, where the first and second rows show the best results, while the third row shows the worst result, in terms of sensitivity.

Table 1 shows the values of automatically and manually segmented nerve lengths and densities. Densities values are obtained dividing the nerve length by the image area. The results show that the automatic algorithm underestimates the nerve length. They also show that the nerve lengths

Discussion and conclusions

In this work we presented a fully automatic segmentation algorithm for the identification of corneal nerves in CCM images. We evaluated the algorithm performance in a set of 47 images of the sub-basal nerve plexus from non-diabetic and diabetic corneas. This set was extracted from a larger set of 90 images available online, by rejecting low quality images and images whose en face section traverses obliquely several corneal layers, in order to have a set representative of the images obtained by

Acknowledgements

This work was supported by the following Grants: Portuguese Foundation for the Science and Technology (FCT), PTDC/SAU-BEB/73425/2006 and PTDC/SAU-BEB/104183/2008.

References (28)

  • D.V. Patel et al.

    Contemporary in vivo confocal microscopy of the living human cornea using white light and laser scanning techniques: a major review

    Clinical and Experimental Ophthalmology

    (2007)
  • C. Quattrini et al.

    Surrogate markers of small fiber damage in human diabetic neuropathy

    Diabetes

    (2007)
  • D.V. Patel et al.

    Mapping of the normal human corneal sub-basal nerve plexus by in vivo laser scanning confocal microscopy

    Investigative Ophthalmology and Visual Science

    (2005)
  • M. Tavakoli et al.

    Corneal confocal microscopy: a novel noninvasive means to diagnose neuropathy in patients with Fabry disease

    Muscle and Nerve

    (2009)
  • Cited by (26)

    • A Review On digital image processing techniques for in-Vivo confocal images of the cornea

      2021, Medical Image Analysis
      Citation Excerpt :

      It is based on a contrast-limited adaptive histogram, followed by a methodology that uses wavelet transform filters (i.e., log-Gabor filters) to obtain local phase frequency information. This method exploits the phase symmetry of the images containing nerve structures because the nerves have a specific phase response Ferreira et al. (2011, 2012). It must be remarked that Ferreira et al. (2011) and Ferreira et al. (2012) use a selection of images from an online data set Scarpa et al. (2008), where the images of insufficient quality are rejected.

    • A fully automatic nerve segmentation and morphometric parameter quantification system for early diagnosis of diabetic neuropathy in corneal images

      2016, Computer Methods and Programs in Biomedicine
      Citation Excerpt :

      The algorithm achieved an average sensitivity of about 0.85, a false detection rate of 0.05 and an execution time of 25 seconds per image, but again, further details about how the execution time was measured are not available. Ferreira et al. [34] developed an automatic segmentation and morphometric analysis system for sub-basal corneal nerve images obtained by in vivo corneal confocal microscopy. The system starts by enhancing the image contrast using an adaptive histogram equalization method.

    • Detection of pathologic liver using ultrasound images

      2014, Biomedical Signal Processing and Control
    View all citing articles on Scopus
    View full text