Excessive amount of seeds being dropped in the merge step of Initialization

[Pranav Sawant]

The first plot attached represents the seed initialization (generating over complete set of seeds). Now after performing peak to noise ratio refine and merging the seeds, I find that most of the initialized seeds are flagged as red (false seeds) and are eventually dropped. However, to the best of my knowledge, most of these seeds do correspond to an actual cell and hence should not be dropped. Due to the iterative nature of the Minian pipeline, seeds keep getting dropped successively and I end up with only about 10 cells in the final csv output where I'm expecting atleast 50-60 cells. Why are so many cells getting dropped in this process? Is it something to do with the noise frequency parameter?

Note:

The first plot attached represents the seed initialization (Seeds_Initialized)
The second plot attached represents the seeds after merging (white = actual cells, red = false cells) (Seeds_Merged)
The third plot represents the peak to noise ratio refine visualization (I chose noise frequency as 0.06) (Noise_freq)