Automatic analysis of diabetic peripheral neuropathy using multi-scale quantitative morphology of nerve fibres in corneal confocal microscopy imaging
Graphical abstract
Corneal confocal microscopy is a new imaging technique with the potential to become a valuable endpoint in the assessment of peripheral neuropathy. For automated analysis of CCM images, weak fibre signals need to be detected against a noisy background. Having recently described a fibre detector based on a combined model of foreground and background at a single scale, we develop a multi-scale version that classifies pixels on the basis of response to the dual model at a range of scales and orientations, using both random forest and neural net classifiers. ROC analysis demonstrates the superior performance of this detector over the single scale version and over a number of well-known linear feature detectors. Comparison with expert manual analysis shows that automatically derived fibre measurement is strongly correlated with manual measurement and produces similar results in stratifying disease.
Highlights
► Corneal confocal microscopy is a novel imaging modality for quantifying neuropathy. ► We describe a novel multi-scale detector for low-contrast curvilinear structures. ► We compare it quantitatively with a number of well-known alternative algorithms. ► We evaluate random forest and neural network approaches to pixel classification. ► The automatic quantification of nerve fibres is equivalent to manual measurement.
Introduction
According to numerous clinical reports (DiabetesUK, 2010), diabetes is among the most challenging chronic health problems. For example, in the UK it is estimated that one in twenty people has diabetes, whether diagnosed or undiagnosed, and by 2025 four million people will have the condition. Damage to the peripheral nerves (diabetic peripheral neuropathy, DPN) is one of the commonest long-term complications of diabetes occurring in at least 50% of patients with diabetes (Boulton, 2005). As a consequence, about one in six diabetic patients have chronic painful neuropathy, compared to one in 20 non-diabetic subjects (Daousi et al., 2004). It is the main initiating factor for foot ulceration, Charcot’s neuroarthropathy and lower extremity amputation. As 80% of amputations are preceded by foot ulceration, an effective means of detecting and treating neuropathy would have a major medical, social and economic impact. The development of new treatments to slow, arrest or reverse this condition is of paramount importance but is presently limited due to difficulties with end points employed in clinical trials (Dyck et al., 2007). Therefore accurate detection and quantification of DPN are important to define at-risk patients, anticipate deterioration, and assess new therapies. Current methods are unsatisfactory, lacking sensitivity and requiring expert assessment, and focus only on large fibres (neurophysiology) or are invasive (skin/nerve biopsy). Unfortunately, diabetic neuropathy lacks a non-invasive surrogate for nerve damage (Tesfaye et al., 2010).
Recent research (Malik et al., 2003, Kallinikos et al., 2004, Hossain et al., 2005) using corneal confocal microscopy (CCM) suggests that this non-invasive, and hence reiterative, test might be an ideal surrogate endpoint for human diabetic neuropathy. The establishment of CCM as a surrogate for early diagnosis and an early biomarker for diabetic neuropathy could identify those at risk and prompt more intense intervention including improved glycaemic, blood pressure and lipid control. Furthermore a sensitive surrogate endpoint would significantly lower hurdles to the development of disease-modifying therapeutics by enhancing the capacity to test therapeutic efficacy. The major advance of CCM is the entirely non-invasive and relatively rapid (≈2 min) acquisition of images of small nerve fibres in patients. However, the major limitation preventing extension of this technique to wider clinical practice is that analysis of the images using interactive image analysis is highly labour-intensive and requires considerable expertise to quantify nerve pathology. To be clinically useful as a diagnostic tool, it is essential that the measurements be extracted automatically.
If an automatic CCM image analysis system is to be applied clinically, especially to define early degeneration or regeneration, then a key step is the automatic detection of low-contrast nerve fibres among image noise (see Fig. 1). The literature on this topic is not extensive, although the problem has a superficial similarity to other, more widely investigated, applications, such as detection of blood-vessels in retinal images. Ruggeri et al., 2006, Scarpa et al., 2008 describe a heuristic method that was adapted from retinal analysis. A number of methods have been developed to enhance the contrast of such linear structures. In a previous study (Dabbah et al., 2009), we used the 2D Gabor filter (Jain and Farrokhnia, 1991) to detect nerve fibres in CCM images. The filter is a band-pass filter that consists of a sinusoidal plane wave with a certain orientation and frequency, modulated by a Gaussian envelope. This spatial domain enhancement is based on the convolution of the image with the even-symmetric Gabor filter that is tuned to the local nerve-fibre orientation. We subsequently extended this to form a dual-model detector (Dabbah et al., 2010), see Section 4.
The automated system of analysing CCM images presented in this paper is an extension of our previous single scale dual-model fibre detector (Dabbah et al., 2010). The new detection algorithm uses the dual-model property in a multi-scale framework to generate feature vectors from localised information at every pixel. These vectors are then used to classify pixels using random forests (RF) (Breiman, 2001) and neural networks (NNT) (Moller, 1993).
In the remainder of the paper we introduce CCM imaging, the image characteristics and the metrics that have been used to quantify the nerve morphology by interactive image analysis (Sections 2 Corneal confocal microscopy, 3 Linear-structure and feature detection). We describe the single-scale dual model filter (Dabbah et al., 2010) and its extension to multiple scales with pixel classification (Sections 4 Single-scale dual-model enhancement, 5 Multi-resolution dual-model enhancement, 6 Nerve fibre classification). In Section 7 we describe a comparative evaluation showing the improved performance of the multi-scale version over not only the single-scale filter but a number of other multi-scale detectors. We also demonstrate that the automatically detected fibres result in morphometric features equivalent to those generated by expert interactive analysis.
Section snippets
Corneal confocal microscopy
The cornea is one of the body’s most innervated tissues. The sub-basal nerve plexus runs parallel to the surface of the cornea in the Bowman’s membrane, lying between the outer epithelial layer and the stroma. Bowman’s layer is about 8–12 μm thick, and the nerves may be imaged by confocal microscopy using either a white-light source or a laser source. In this study laser confocal microscopy was used.
Linear-structure and feature detection
Detection of curvilinear structures is a requirement in several applications of medical image analysis. A method of linear structure detection (Line Operator – LinOp), originally developed for detection of asbestos fibres (Dixon and Taylor, 1979) has also been shown to be effective in detecting ducts in mammograms (Zwiggelaar et al., 2004). LinOp exploits the linear nature of the structures to enhance their contrast by computing the average intensity of pixels lying on a line passing through
Single-scale dual-model enhancement
All of the methods described in Section 3 are potential means of enhancing the linear nerve structures in the face of the image corruption outlined in Section 2.3. In Dabbah et al. (2010) we reported on the performance of the single-scale dual-model detector in comparison with these methods. We showed that the detectors specifically designed for detection of linear structures performed better than more general feature detectors, such as the Monogenic filter. In particular the single-scale
Multi-resolution dual-model enhancement
The single resolution detector described in Section 4 makes use of local orientations calculated on a regional basis and operates with a single wavelength parameter for the Gabor filter, thereby assuming a single width for all fibres. In this section we extend the model to multiple resolutions using a scale pyramid as shown in Fig. 3. We also calculate responses over a range of orientations, selecting the most appropriate scale and orientation of the response by pixel classification. There are
Nerve fibre classification
We consider three possible ways of using the feature vector to assign pixels (i, j) to the foreground or background classes.
Database and experimental settings
The evaluation is conducted on a database of 521 CCM images captured using the HRT-III microscope from 68 subjects (20 controls and 48 diabetic patients). The images have a size of 384 × 384 pixels, 8-bit grey levels and are stored in BMP format. The resolution is 1.0417 μm and the field of view is 400 × 400 μm2 of the cornea. For each individual, several fields of view are selected manually from near the centre of the cornea that show recognisable nerve fibres. Other than the processing inherent in
Conclusion
The analysis of CCM images requires the identification of fibre-like structures with low contrast in noisy images. This is a requirement shared by a number of imaging applications in biology, medicine and other fields, and a number of methods have been developed and used in these various applications. In the present work we present a new multi-scale dual-model method to detect corneal nerve fibres in CCM images and we compare this with some more generic methods. In our evaluation the
Acknowledgements
This work is supported by a Juvenile Diabetes Research Foundation (JDRF) scholar Grant 17-2008-1031 and subject to patent application (UK Patent Application No. 1005905.3).
References (45)
Two-dimensional spectral analysis of cortical receptive field profiles
Vision Research
(1980)- et al.
Early detection of diabetic peripheral neuropathy with corneal confocal microscopy
The Lancet
(2005) - et al.
Unsupervised texture segmentation using Gabor filters
Pattern Recognition
(1991) A scaled conjugate gradient algorithm for fast supervised learning
Neural Networks
(1993)- et al.
Corneal confocal microscopy: a novel means to detect nerve fibre damage in idiopathic small fibre neuropathy
Experimental Neurology
(2010) - et al.
The north-west diabetes foot care study: incidence of, and risk factors for, new diabetic foot ulceration in a community-based patient cohort
Diabetic Medicine
(2002) - Ali, R., Gooding, M., Christlieb, M., Brady, M., 2008. Advanced phase-based segmentation of multiple cells from...
- et al.
Shape quantization and recognition with randomized trees
Neural Computation
(1997) Management of diabetic peripheral neuropathy
Clinical Diabetes
(2005)Bagging predictors
Machine Learning
(1996)
Random forests
Machine Learning
Dual-model automatic detection of nerve-fibres in corneal confocal microscopy images
The International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI)
Chronic painful peripheral neuropathy in an urban community: a controlled comparison of people with and without diabetes
Diabetic Medicine
Automated asbestos fibre counting
Challenges in design of multicenter trials: endpoints assessed longitudinally for change and monotonicity
Diabetes Care
The monogenic signal
IEEE Transactions on Signal Processing
Practical Methods of Optimization
The random subspace method for constructing decision forests
IEEE Transactions on Pattern Analysis and Machine Intelligence
Fingerprint image enhancement: algorithm and performance evaluation
IEEE Transactions on Pattern Analysis and Machine Intelligence
Cited by (245)
Evaluation of Corneal Nerve Regeneration After Minimally Invasive Corneal Neurotization
2023, Asia-Pacific Journal of OphthalmologyMorphological alterations in corneal nerves of patients with dry eye and associated biomarkers
2023, Experimental Eye ResearchImproving corneal nerve segmentation using tolerance Dice loss function
2024, Signal, Image and Video Processing