Image Fusion for Prostate Cancer Localization

There is a huge inter and intraobserver variability with human readers when multispectral MRI is used by radiologists for prostate cancer localization. Our goal in this project is to provide the radiologist a single image that carries most of the significant information from the whole multispectral MRI. In this way, both inter and intraobserver variability will be decreased and localization performance of the human readers will increase.

Image fusion methods we develop allows us to see significant information (e.g. tumors) from all MR image types in a single image. We observe that tumors showing in different MR images can be seen in the single fused image. When viewed by a radiologist, this single fused image has the potential to increase localization accuracy and decrease inter- and intra-observer variability.