Generative AI has the potential to significantly mitigate data privacy and security risks by enabling the sharing of synthetic images containing features of cancer diseases across multiple cancer care centres. In this work, we focused on training a conditional denoising diffusion probabilistic model (DDPM) to generates synthetic images of Maximum Intensity Projection (MIP) of Diffusion-weighted Images (DWI) in different rotational views. Our model attains 40.30 of FID score. Furthermore, the visual dissimilarity between synthetic samples and their closest raw data per Mean Square Error indicates the absence of mode collapse. An expert radiologist achieved an accuracy rate of 52% in blinded test that distinguishes real and fake MIPs from the provided images. Moreover, synthetic images scored 3.2 on average, whilst real images scored 3.6 on a subjective radiological image quality score (1 = poor quality to 5 = excellent image quality).