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.

Medición de la microcirculación retiniana luego del tratamiento antihipertensivo / Measuring retinal circulation after antihypertensive treatment

Background and objective: With the development of image processing techniques, it has become possible to measure the changes in retinal vessels of hypertensive patients by means of eye fundus photographs. Patients and method: In this paper we aim to classify retinal vessels automatically into arterioles and venules. In order to do so, we have compared three different strategies based on the colour of the pixels in images through an analysis of 78 hypertensive patients' eye fundus images. The first strategy classifies all the vessels by applying a clustering algorithm. The second divides the retinal image into four quadrants and classifies the vessels that belong to the same quadrant independently from the rest of the vessels. The third strategy classifies the vessels by dividing the retinal image into four quadrants that are rotated inside the mentioned image. Results: The third strategy was the one that obtained the best results, since it minimizes the number of unclassified vessels. In the initially analysed set of 20 images, we correctly classified 86.53% of the vessels, and this percentage remains similar in a set of 58 images examined by three medical experts. This confirms the validity of the method that automatically calculates the arteriovenous ratio (AVR).Conclusion: Our results are an improvement on those previously described in the bibliography, reducing the number of non-classified vessels. Furthermore, the method entails low computational costs.

Fundamento y objetivo: El desarrollo de técnicas de procesado de imágenes ha devuelto interés para poder medir de una forma objetiva los cambios en la estructura microvascular del hipertenso a través de las fotografías digitales del fondo de ojo. Pacientes y método: Para clasificar de forma automática los vasos de la retina en arteriolas y vénulas, con una elevada precisión, hemos comparado tres estrategias diferentes basadas en la información del color de los pixeles de la imagen del fondo de ojo, analizando 78 imágenes de fondo de ojo de hipertensos. La primera estrategia clasificaría todos los vasos aplicando un algoritmo de agrupamiento. La segunda divide la retina en cuatro cuadrantes y clasifica los vasos que pertenecen al mismo cuadrante independientemente del resto de los vasos. La tercera estrategia clasifica los vasos dividiendo la retina en cuadrantes que son rotados. Resultados: La mejor estrategia resultó la tercera porque minimiza el error y el número de vasos no clasificados. La característica vectorial más determinante está basada en la media o la mediana del componente gris del espacio de color RGB. Para las 20 imágenes inicialmente analizadas hemos clasificado correctamente el 86.53% de los vasos, y este porcentaje permanece similar en el grupo de 58 imágenes examinadas por tres expertos, lo que confirma la validez del método, para el cálculo del índice arteriovenoso de forma automática. Conclusión: Nuestros resultados son superiores a los descritos previamente, reduciendo además el número de vasos no clasificados. Por otro lado, el costo computacional del método es bajo

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.