Over the past couple of decades, Case Based Reasoning (CBR) has evolved as a popular paradigm for decision
making in real world problems. CBR hinges on the fact that similar cases have similar solutions. This methodology
has been adapted from the physicians’ approach to diagnosis and therapy planning. The knowledge acquired by
medical experts is a combination of text book knowledge and knowledge acquired from clinical experience. The
basic idea of CBR 1 is to retrieve cases from a case database and establish the relevance between candidate and
prototype cases through a similarity measure. If the first case history involves the analysis and classification of sets
of longitudinal series of multimedia image sets, automatic indexing using digital content, referred to as ContentBased Image Retrieval (CBIR) 2 is a possible solution for defining similarity measures. Thus, retrieval of images
from a database based on similarity based criteria is an important procedure in CBR 3, 4, 5. When the retrieval process
involves heterogeneous information like images and contextual information, the CBR system encounters problems
in aggregating these variables and dealing with missing information. Decision trees 3 are well suited for retrieval to
process heterogeneous as well as incomplete information. This note proposes an approach to classify abnormalities
in fundus images of the retina using decision trees. The abnormalities of the retina that have been considered for the
present study are those arising from age related abnormalities like Age Related Macular Degeneration(AMD) and
those arising from Diabetic Retinopathy (DR) like micro aneurysm(MA), hard exudates (HE), cotton wool spots
(CWS) and hemorrhages. Table 1 lists the manifestations of the abnormalities as seen in retina fundus images,
obtained in public databases. To the best of our knowledge, there is no existing work combining AMD and DR. This
is a pioneer attempt in integrating contextual information with images.