Parameter questions

[Adam]

I have two remaining issues. First, is 'areaMin' just a suggestion? I have set it to 60 but consistently get cell outlines that are much smaller than this (~40). Second, I get a number of "ghost cells" that do not seem to be associated with anything in the phase image. I would like to know which parameter is responsible for them. I have attached an annotated screen shot.

Thank you,
Adam

[Seán Murray]

Hi Adam,

this is because of a bug in microbetracker. There is no quality control after cell splitting. So in timelapses or in independent mode with split1=1, cell splitting can result in cells with an area smaller than areaMin. In timelapses, they would get quality checked and deleted on the next frame (still leaving one frame with a bad cell and of course an incorrect mother cell).

Ghost cells are caused by a failure of the alignment routine (the contours gets 'evolved' away from the cell). It can be caused by the bug above -  a bad split is performed, the alignment goes crazy but the cell is kept due to the lack of quality control.

I have made the following change (between the comment lines) at around line 4371 and also at 4706

            mesh1 = model2mesh(cCell1,p.meshStep,p.meshTolerance,p.meshWidth);

            mesh2 = model2mesh(cCell2,p.meshStep,p.meshTolerance,p.meshWidth);

            

            

            %added by sean murray - prevent split if resultant cells are

            %too small  (mother is not reprocessed below, only daughter)

            if length(mesh1)>1

                cellarea1 = polyarea([mesh1(:,1);flipud(mesh1(:,3))],[mesh1(:,2);flipud(mesh1(:,4))]);

            else cellarea1=0;

            end

            if length(mesh2)>1

                   cellarea2 = polyarea([mesh2(:,1);flipud(mesh2(:,3))],[mesh2(:,2);flipud(mesh2(:,4))]);

            else cellarea2=0;

            end

          

            if(fitquality1>p.fitqualitymax || fitquality2>p.fitqualitymax || p.areaMin>cellarea1 || p.areaMin>cellarea2 || p.areaMax<cellarea1 || p.areaMax<cellarea2)

                gdisp(['Tried to split cell ' num2str(cell) ' but failed']);

                if(fitquality1>p.fitqualitymax || fitquality2>p.fitqualitymax) gdisp(['Fit qualities ' num2str(fitquality1) ' ' num2str(fitquality2) ' of daughter cells are too low']);end

                if( p.areaMin>cellarea1 || p.areaMin>cellarea2) gdisp(['Areas ' num2str(cellarea1) ' ' num2str(cellarea2) ' of daughter cells are too low']);end

                if( p.areaMax<cellarea1 || p.areaMax<cellarea2) gdisp(['Areas ' num2str(cellarea1) ' ' num2str(cellarea2) ' of daughter cells are too high']);end

            else

                % end of addition apart from 'end'

                cellStruct1.algorithm = p.algorithm;

                cellStruct2.algorithm = p.algorithm;

                cellStruct1.birthframe = frame;

                cellStruct2.birthframe = prevStruct.birthframe;

                if size(pcCell1,2)==1, model=pcCell1'; else model=pcCell1; end

the extra 'end' comes after the following 'continue'.

I hope that is clear. I do not have access to the Yale git repo that you posted to. It seems to be internal.

Sean

[Adam]

Hi Sean,

I do not have access to the Yale git repo that you posted to. It seems to be internal.

Sorry about that. I thought 'public' repositories on Yale's git server were public, but they are not.

I have moved the repository to my public account: https://github.com/nodice73/MicrobeTracker

One problem I'm having is that 'fitquality1' and 'fitquality2' are not defined anywhere else. What am I missing?

Thanks,
Adam

[Adam]

Hi Sean,

I think I figured it out. Had to get fitqualities from the respective align functions. Changes are in last two commits: https://github.com/nodice73/MicrobeTracker/commits/master/microbeTracker.m

[Seán Murray]

oops, sorry about that!

[Adam]

The ghost cell remained after this fix. I got rid of it by increasing the fitquality parameter to 0.3.