How to fit function with 3 variables

kx=-0.5:0.00666:0.5;
ky=kx;
w=5:0.1:20;
KXYW=[kx(:) ky(:) w(:)];

f = @(KxyW,b) b(1)-((b(2)*(KxyW(:,1)-X0)-b(3)*(KxyW(:,2)-Y0))/b(4)).^2+((b(3)*(KxyW(:,1)-X0)+b(2)*(KxyW(:,2)-Y0))/b(5)).^2+((KxyW(:,3)-W0)/b(6)).^2;
B0=[0.008 0.5 0.5 0.1 0.1 0.5];
B = lsqcurvefit(f,B0,KXYW,M);

What is important M is size 151x151x151 so when i run this code i got this statment:

Error using lsqcurvefit (line 262)

Function value and YDATA sizes are not equal.

I know f should be the same size as M but i dont know how to do this.

Generaly my task is to find parameters (b(1),b(2)...) of fitted function to area near maximum value. I did find max of M.

This is result of calling : surface(M(:,:,6)) (so you know how it's look, there is 151 of this)