enquiry

[conluana88]

I have tested your segmentation tool on two different MRI datasets. One dataset involves fetuses of pregnant women, aiming to identify the liver and lungs, while the other focuses on breast imaging to detect tumor lesions.
For the first dataset, I achieved good results.
However, with the second dataset, despite having twice as many images compared to the first one, the outcomes varied. The DICE coefficient ranged from totally 0 to levels surpassing 80.
I am currently investigating the reasons behind these discrepancies. One significant difference lies in the fact that, in the first dataset, I was able to perform a 3D cascade, whereas this operation was not allowed in the breast dataset.

I would appreciate your assistance in understanding how to enhance the results. Any insights or suggestions you can provide to improve the performance, especially in the context of the breast dataset, would be highly valuable.