Depends on the strengths of the department and what faculty members are willing to supervise. Classification problems in medical images is an emerging area of research. The basic application is to identify abnormal regions (disease detection and localization) in images.  The cool part is it uses a lot of techniques like image processing, segmentation as well as reconstruction strategies.

As mentioned before medical image segmentation using machine learning approaches such as deep elarning

Diabetic Retinopathy Detection- Identify signs of diabetic retinopathy in eye images

this problem is on kaggle 

for more detail visit https://www.kaggle.com/c/diabetic-retinopathy-detection

depends upon the strength of guide and ur willingness .. if you can devote more time then you can go for microarray  problem..http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3467903/

Thank you all

Although its late answer, you can go through deep learning techniques in image classification such as plants species, or medical images as it is now widely adapted and an interesting research area.