Speaker: Maya Mayberg
M.Sc. student under the supervision of Dr. Arnaldo Mayer and Prof. Nahum Kiryati
Sunday, January 26th, 2020 at 15:00
Room 011, Kitot Bldg., Faculty of Engineering
Neural image enhancement of MRI accelerated scans
Abstract
MRI is an essential tool in diagnostic imaging, often revealing findings that are not detectable by other imaging modalities, e.g. X-ray, CT and Ultrasound. A typical brain MRI scan lasts several minutes with overall protocol duration exceeding 30 minutes. Image quality, in terms of spatial resolution and SNR, is strongly dependent on acquisition duration.
Currently, the throughput of MRI scanners is clearly limited by the long scanning times, leading to elevated scan costs, long queues for patients and poor profitability of the imaging centers. Therefore, shortening MRI scans is crucial.
In this research, we investigate the enhancement of low-resolution (LR), fast acquisition MRI scanning, to reduce scanning time without compromising the diagnostic value.
We propose a unique fully convolutional neural network, optimized for LR MRI data obtained by down-sampling the k-space phase encoding only. The network is trained to transform the LR acquisitions into corresponding high-resolution (HR) counterparts. In our approach, the LR images used for training are real acquisitions rather than smoothed, down-sampled versions of the HR images, as usually performed in previous studies.
The proposed method is validated qualitatively and quantitatively for an acceleration factor of 4. Promising results are obtained, indicating that the proposed method may become a valuable approach to the reduction of MRI scanning time.
