1. Availability of Database is the Biggest Problem in Medical Image Processing

2. Complexity involved in the Image Texture (You need a Strong Segmentation method)

3. in Medical images, recognition is not sufficient, you need to Predict future Values..

You can find the sample Database from fallowing link

https://github.com/beamandrew/medical-data

See: Max A.Viergever et al. A survey of medical image registration – under review. Medical Image Analysis. Volume 33, October 2016, Pages 140-144

This review is rather specific to DNN, but gives some overview of the challenges:

Article A Survey on Deep Learning in Medical Image Analysis

Cell segmentation is also one of the major problems.

Dear Mazin,

In the microscope images there is no big problem for segmentation process.

The real challenge in such images is how to make the computer produce a decision to determine the type of disease.

Regards

1. Availability of Database is the Biggest Problem in Medical Image Processing

2. Complexity involved in the Image Texture (You need a Strong Segmentation method)

3. in Medical images, recognition is not sufficient, you need to Predict future Values..

You can find the sample Database from fallowing link

https://github.com/beamandrew/medical-data

What we found is that the boundaries of medical images are fuzzy in nature --and this is the major problem. As traditional methods are unable to deal with this fuzziness. Some pre-processing techniques are utmost required before going or final segmentation. Few are discussed in the following paper: Article Importance of Image Enhancement Techniques in Color Image Se...

Proper pigmentation of the cell by the expert if you are working on your own dataset. It will help you lot in finding optimal threshold value so that you can extract image as per your wish.