Suppose we want to demonstrate a new device has improved accuracy than the old device, so we have a training study and clinical trial. The training study is used to develop machine learning algorithms and set cutoffs for the clinical trials. It can also give us an initial estimate of accuracy.
My question is: what is the minimum patient sample size for the training set and the clinical trial? I know we would always prefer more patients, but it is limited by time, cost, infrastructure, etc. Thank you.
Our goal is to test if the new device is significantly better than the old device. Both devices will be tested on the same group of patients. I am interested to know if the old device has an expected accuracy of 50%, the new device has an expected accuracy of 70%, how to calculate the required sample size?
Hi Xu. I attached two papers that you may find useful.
Article Estimating the sample size for a pilot randomised trial to m...
Article Sample size of 12 per group rue of thumb for a pilot study
Hi, here is useful paper by Jurgen Popp et al.
Article Sample Size Planning for Classification Models
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis