If you acquire very high resolution data, and at the correct b-value, then this should be possible. I would suggest using a technique for selective k-space - GE calls this FOCUS, Siemens call it ZoomIt - to acquire a limited FOV in k-space and therefore very high resolution. I would suggest using this technique to acquire something around 0.5mm (500 micron) resolution with at least 2 averages. Your radiographer will suggest the best b-value, but I would try somewhere between 300 and 500. You shouldn't need to acquire many diffusion directions (I think 25 would be sufficient).