Differen number of neurons detected on different computers.

[Tuu-ticki]

Hi!

I'm very surprised by the MIN1PIPE behavior on different computers.

I have desktop with 32-Gb RAM and Matlab R2020a and laptop 16-Gb RAM and Matlab R2021a. I've installed MIN1PIPE on both computers and then run demo_min1pipe.m on the data in the "demo" folder. On desktop I got 42 neurons detected, on laptop - 46 neurons. But the supplementary graph "demo_visualization.png" shows more than 70 neurons.

I'm wondering what could be wrong with my installation? And why could I get different number of active neurons on different computers? Could the reason be in the RAM or in Matlab version?

[JinghaoLu]

I am not sure why you got 42 versus 46 on different computers - the ROI initialization process does contain some level of randomization, but it does not seem likely to be found in the demo video. Maybe I am wrong on this point. Also to my knowledge it is less likely/independent of the RAM or matlab version in the computers.

For the other question: if you are asking about the difference from the figure you see in the Readme document, then the answer is MIN1PIPE has evolved a lot since the initial implementation and I simply did not change that figure. The main difference would be that the current implementation will have a more robust seeds selection process that works well on the usual full scale FOVs directly from any miniscope, but might not keep the ROIs close to the edges in the highly cropped video (like the one in the demo folder). In addition, it tends to keep only the most confident ROIs as well.

[Tuu-ticki]

1. I got inconsistence in neuron numbers not only in demo video, but in my data too. It's funny that min1pipe produces the same neuron number if I rerun it on the same computer. And in another matlab (on different computer), the program stably gives a different value.
   I've install MATLAB R2021a on the desctop and now I get the same number of neurons on both computers. This is very strange, but it seems to depend on the matlab version.
2. Thank you for the clarification!

[JinghaoLu]

@Tuu-ticki That's new to me. Maybe some builtin matlab functions have been implemented differently.