GRAFA stands for General Radiology Automated Feature Analysis. GRAFA is an offline application for data mining of some topics in the radiology. The objective of this project is to help radiologists in organising their differential diagnosis when they are facing a difficulty in making a clear diagnosis or when the case in hand has non-specific radiological signs which could imply many possibilities. It applies a dedicated mathematical algorithm to retrieve the important features in each case and then match them with the common diseases in descending order of importance. All what the user needs to do is to select the system where his case belongs, then selecting the variables that exist in each system among a list of many variables. The result will then appear as a list of diseases in a descending order of importance that reflects the strength of the match between them and the case.
It is important to stress that the degree of matching in this application does not depend on the prevalence of the diseases in certain community. It will rather highlight the similarity in the pattern based on the presented variables. For example, it would be natural in the result to rank a rare disease higher than a common disease if the variables selected match more with the former than with the latter.
Interestingly, the name “GRAFA” is a Russian word means “column” and this goes nicely with the way some aspects of the algorithm work using “columns” of data. It is worth mentioning that GRAFA is far from being complete because it still requires data base to cover more fields in the radiology, however, we hope that many doctors and health professionals will find it helpful in their work. Sometimes, it may happen to have artefact in the result like any other automated algorithm but hopefully this should not affect the overall performance.
https://www.youtube.com/watch?v=htAZPTjfKDo.