This is my first post and I’m new to coding with python so apologies for my ignorance. I’m physician and working on my own DL project about cardiac imaging as a PhD thesis.
I try to minimize redundat work due to scarce time I’m able to put into this project since I’m doing it alongside my daywork.
I have done small pilot project which proved that the concept works and I managed to accumulate small dataset n=90 but now it’s time to scale up and complicate the problem a bit. I have previously created a workflow to automaticly process DICOM files into ground truth and mask files and do some augumentation and it worked fine with U-Net.
The problem: I’m now adding multiple numerical parameters to the images and expanding the dataset so I figured out that building proper organized dataset would be really beneficial for the project. I have only worked with image file dataset and it was fairly simple. I tried to browse for tips to build a datasets but I only found really general advices. Are there any “good practices” for building a dataset and are there any great resources for learning to build your own multilabel dataset from scratch with different feature types?
Mayby more accurate question would be: What would be efficient/smart way to integrate my automatic DICOM processing script to a database generating script and what methods should I look into to create database generating script?
Thank you in advance!
edit: I would like to emphasise that I don’t have any experience with other programming languages other than Python and I have only worked with python for 4 months now so I’m really novice in programming but I’m quick to learn.