Normalization of of DWI to Standard space without intermediate anatomical image

Hello Neurostars community.

I want to normalize 2d diffusion weighted images (with approximately 2x2x3mm resolution; not DTI but routine b0-1000 diffusion weighted images), to MNI space. However no volumetric anatomical images, such as 3D T1, 3D T2 or 3D FLAIR images were acquired (only SWI, and 2D FLAIR (5mm slices with 2mm gaps were acquired additionally))
I am aware that routinely diffusion is registered to T1 and then normalized to MNI space, but given the available data I have access to, is there a reasonable, or acceptable method?
Also, I want to segment lesions (e.g. high signal on DWI, or tissue loss) on DWI, should I segment the lesions then normalize, or segment then normalize?

Thanks in advance.

HJ Lee.