Towards Autonomous MR imaging: World models

MR imaging is a powerful and diverse imaging technique employed to investigate and diagnose a range of diseases. MRI scans are acquired by employing specific parameters in a “sequence” to encode in data in arbitrary space known as “k-space”. Image is reconstructed by applying mathematical transforms (mainly Fourier) to the k-space data. To obtain images of particular contrast (T1w, T2w, T2* etc) optimal sequence settings must be employed. Briefly, image contrast arises from the magnetic property of hydrogen atoms, and image contrast can be varied using setting such as echo time (Te) and Repetition time (Tr) in an MRI scanner. These settings are obtained by solving Bloch equations and are required to be predetermined before starting the acquisition process. In addition, to acquire an image, each line of k-space needs to be acquired using the sequence with these predetermined settings. This leads to slow encoding process and longer scans duration. Therefore, this post outlines the AI framework, which can be employed to automate this process to obtain optimal images in the shortest scan duration. The goal is to highlight the key AI technologies, using the fastMRI dataset. This dataset was acquired using fixed image contrast settings, therefore is not suitable for optimisation of acquisition setting such as Te, Tr, however, it does allow us to develop technology toward optimal speed up methodologies. This framework can be combined with additional datasets which are acquired with variable image acquisition settings to obtain optimal image acquisition settings, or…