Dual-Boot MRI for Multi-Site and Longitudinal Studies of Autism Under Stable Conditions

Background: MRI studies of autism typically span years, may be longitudinal, and may involve multiple institutions.  To reduce variability, a stable scanning environment across time and sites is essential.  However, changes in scanner software can be frequent.  MRI vendors provide software changes nearly annually, which can provide new scanning routines, but can also change existing routines and prevent operation of custom procedures.  Conflicts can arise between changing-versus-maintaining scanner software.  Changing software is typically desired by clinicians (diagnosis does not require stability); vendors (minimizing supported software levels); and some researchers who want to begin new studies.  Maintaining software is typically desired by researchers who have ongoing studies or use custom procedures.

Objectives: These competing needs could be met by "dual booting", where the operator chooses the software level for booting the scanner.  However, dual-booting is not commercially available on any FDA-approved MRI scanner.  While dual-booting is done in the factory during software development, it is unclear whether an entire installed scanner can be seamlessly switched between two software environments.  We aimed to develop dual-booting at a second scanning site to expand enrollment in our studies of autism and other disorders, while preserving ongoing longitudinal studies.

Methods: We devised procedures to dual-boot a Siemens 3T Allegra scanner at a second site (Pittsburgh; va30 software) to match the primary site (St.Louis; va25 software).  The scanner has a host computer (running the scanner) and an imager computer (reconstructing images).  Host and imager hardware/software differ between environments.  To preserve scanning conditions, we used separate computers for va25 booting.  The original va25 computers were recovered/tested at the second site, and rebuilt by replacing missing/unreliable components.  We reinstalled all original va25 firmware/software, and moved the single-user license dongle from va30 to va25 host computer.  After the stand-alone va25 host/imager were running, we plugged them into a network switch and set internal/external IP addresses.  All communication errors resolved upon rebooting.  We then connected the va25 computers to the scanner network by switching all cabling between va30/va25 computers, and then powering up the scanner.  The va25 computers and all electronics booted up in the va25 environment, without errors.

Results: After booting under va25, we tuned the scanner using vendor procedures, storing tune settings on the va25 host.  All performance specifications were met.  Autism scan protocols ran seamlessly without modification, and image quality was excellent.  The full switching procedure (scanner power-off, computer switching, scanner rebooting) takes only 13-15 minutes, during which the participant can be taken out of the scanner and the next participant prepared.  For ongoing longitudinal studies, the scanner is run under va30 (using va30 tuning settings) to maintain those study conditions.

Conclusions: A dual-boot MRI environment can be used with longitudinal/multi-site autism studies to stabilize scanning conditions.  Such studies are crucial to understanding brain developmental trajectory and biological heterogeneity in autism, which require long-term stability and large sample sizes.  Setting up dual-booting requires a site willing to take risks/provide access; motivated/experienced investigators; available prior computers/software/manuals; and minimal scanner hardware changes.

Funding: MH090494; NS052470; AS03799; BC073839.