This is a project collaborate with Georgetown University medical school. Hospital has large amount of images of DNA fragmentation without being taken fully advantage of. Now, we want to make machine detect out the distorted DNA fragmentation from these images accurately and automatically. Then we extract features as many as we can. Next, we are expected to apply the deep learning techniques, especially the
convolutionary-based neural network, to train our data in order to help doctors see how cancer is related to the DNA objects. If possible, our work would expand to map all the projection of panel DNA diagnostic images into a real-3D space which would be really fancy and valuable!!
Noise deduction & smooth processing
Objects detection (selective search algorithm)
Feature extraction (length & count & distorted etc.)
Comparison with artificial work
Modification >> feature engineering