amatov / DifferentialDiagnosisCBIR

Image retrieval can facilitate medical diagnosis by identifying categories of similar to a new patient presented for diagnosis phenotypes which have already been assigned a diagnosis

Python code I wrote to implement a Reverse Image Search / Content-Based Image Retrieval (CBIR)

Based on the feature descriptors generated by the scale-invariant feature transform and the bag-of-words model (BoW), it performs reverse image search.

Content-based image retrieval query is accomplished via ElasticSearch using ranking based on a k-means clustering algorithm.

See my preliminary demo of the BoW ranking at: https://vimeo.com/365998906/e94e02a8af

For further information regarding differential diagnosis based on in vivo imaging, please see: https://github.com/amatov/FluorescentFundusNeurodegeneration