Let's assume a sphere of constant density p and radius r, which grows m times larger over the time t.
A point that initially has a distance of d from the center after the growth has a distance of m*d from the center, meaning it moved m*d-d = (m-1)d far.
In that case, point of that nature have a KE per volume of 0.5*p*(m-1)d/t.
Points of that nature build a sphere with initial radius d. The sphere has a surface area of 4*pi*d^2.
So the KE per thickness of sphere is 0.5*p*(m-1)d/t * 4*pi*d^2 = 2*pi*p*(m-1)*d^3 / t.
Next we need to "sum up" all these spheres. So we integrate that over d from 0 to r.
Integral 0 to r of (2*pi*p*(m-1)*d^3 / t) dt = (pi * (m - 1) * p * r^4) / (2 * t)
And that's your KE.... I think.
@Ugarik does that make sense or should I have gone the safe derivative approach after all