In machine learning related to medical imaging, there is a need for large data sets which quite often is not available. Is it acceptable to use multiple images from the same object or case to increase the sample size? Are there any references for studies that use this approach?
Dear Laith R Sultan
These links might be helpful, have a look:
1. Article “How many images do I need?” Understanding how sample size p...
2. https://www.researchgate.net/post/Which-are-the-best-methods-to-Increase-image-datasets
3. https://journalofbigdata.springeropen.com/articles/10.1186/s40537-019-0197-0
4. https://arxiv.org/ftp/arxiv/papers/2010/2010.08186.pdf
5. https://machinelearningmastery.com/impact-of-dataset-size-on-deep-learning-model-skill-and-performance-estimates/
6. https://towardsdatascience.com/how-do-you-know-you-have-enough-training-data-ad9b1fd679ee
Kind Regards
Qamar Ul Islam
Laith R Sultan ,
The question is not as straightforward as it seems. From a traditional medical view it is not the same to take an x-ray from a lateral view that from an anterior view. Location of the particular object may be obstructed due to tissue density, composition, etc. and one view may be more favourable than another. If you are looking to exclude particular anomalies from a particular subject, then it does matter.
I would take a closer look at what is the composition of the dataset, the number of individuals, the particular object of interest and which subject was it taken from to arrive at the conclusion.
Regards
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size