On the development of conjunctival hyperemia computer-assisted diagnosis tools: Influence of feature selection and class imbalance in automatic gradings.

OBJECTIVE: The sudden increase of blood flow in the bulbar conjunctiva, known as hyperemia, is associated to a red hue of variable intensity. Experts measure hyperemia using levels in a grading scale, a procedure that is subjective, non-repeatable and time consuming, thus creating a need for its automatisation. However, the task is far from straightforward due to data issues such as class imbalance or correlated features. In this paper, we study the specific features of hyperemia and propose various approaches to address these problems in the context of an automatic framework for hyperemia grading. METHODOLOGY: Oversampling, undersampling and SMOTE approaches were applied in order to tackle the problem of class imbalance. 25 features were computed for each image and regression methods were then used to transform them into a value on the grading scale. The values and relationships among features and experts' values were analysed, and five feature selection techniques were subsequently studied. RESULTS: The lowest mean square error (MSE) for the regression systems trained with individual features is below 0.1 for both scales. Multi-layer perceptron (MLP) obtains the best values, but is less consistent than the random forest (RF) method. When all features are combined, the best results for both scales are achieved with MLP. Correlation based feature selection (CFS) and M5 provide the best results, MSE=0.108 and MSE=0.061 respectively. Finally, the class imbalance problem is minimised with the SMOTE approach for both scales (MSE<0.006). CONCLUSIONS: Machine learning methods are able to perform an objective assessment of hyperemia grading, removing both intra- and inter-expert subjectivity while providing a gain in computation time. SMOTE and oversampling approaches minimise the class imbalance problem, while feature selection reduces the number of features from 25 to 3-5 without worsening the MSE. As the differences between the system and a human expert are similar to the differences between experts, we can therefore conclude that the system behaves like an expert.

Defining the Optimal Region of Interest for Hyperemia Grading in the Bulbar Conjunctiva.

Conjunctival hyperemia or conjunctival redness is a symptom that can be associated with a broad group of ocular diseases. Its levels of severity are represented by standard photographic charts that are visually compared with the patient's eye. This way, the hyperemia diagnosis becomes a nonrepeatable task that depends on the experience of the grader. To solve this problem, we have proposed a computer-aided methodology that comprises three main stages: the segmentation of the conjunctiva, the extraction of features in this region based on colour and the presence of blood vessels, and, finally, the transformation of these features into grading scale values by means of regression techniques. However, the conjunctival segmentation can be slightly inaccurate mainly due to illumination issues. In this work, we analyse the relevance of different features with respect to their location within the conjunctiva in order to delimit a reliable region of interest for the grading. The results show that the automatic procedure behaves like an expert using only a limited region of interest within the conjunctiva.

Computerized measurement of retinal blood vessel calibre: description, validation and use to determine the influence of ageing and hypertension.

OBJECTIVE: To validate a computer-based method for measuring the calibre of retinal blood vessels, and use it to determine the effects of ageing and arterial hypertension on the calibres of these vessels and on their ratio. METHODS: Digital eye fundus images covering a 50 degrees field and centred on the optic disc were obtained using a 540 nm filter. The boundaries of blood vessels crossing a series of circles concentric with the optic disc were located by an image analysis program; the calibres of vessels crossing the circles perpendicularly were determined automatically, the average arteriole and average vein calibres were calculated, and the arteriovenous ratio (AVR) was calculated as the ratio of these averages. The within-operator, between-operator and within-eye reliability of this method was investigated using images of 30 or 40 eyes; the effects of ageing on average vessel calibres and AVR using both eyes of each of 120 normotensive volunteers aged 10-69 years (60 males, 60 females); and the effects of arterial hypertension using a group of 54 hypertensive patients aged 50.9 +/- 13.9 years. RESULTS: Within-operator, between-operator and within-eye correlation coefficients for AVR were all better than 0.95, and the corresponding coefficients of variation were all better than 3%. The results of Bland Altman approach show a very good agreement. There were no significant differences between right and left eyes or between the sexes in either the normotensive or the hypertensive group. In the normotensive group, vein calibre was almost constant (111 +/- 6 microm), but arteriole calibre and AVR fell significantly from 96 +/- 6 microm and 0.870 +/- 0.046, respectively, in the second decade of life to 85 +/- 4 microm and 0.761 +/- 0.044 in the seventh decade. Arterial hypertension was not associated with changes in vein calibre, but was associated with decreases in arteriole calibre (from 91 +/- 7 microm among normotensive individuals to 84 +/- 2 microm) and AVR (from 0.816 +/- 0.056 to 0.755 +/- 0.027). Age significantly modulated the AVR-reducing influence of hypertension (P = 0.003). CONCLUSION: The proposed method of measuring retinal blood vessel calibres is reliable and precise (especially for the AVR). In this study, its results confirmed that increasing age and arterial hypertension are both associated with reductions in retinal arteriole calibre and AVR.