MRI system upgrade achieves real-time brain scanning

MRI system upgrade achieves real-time brain scanning

Magnetic-resonance-imaging (MRI) technology is used to map the human anatomy based on the oscillation of hydrogen nuclei contained in soft body parts, such as in muscle, blood, and brain tissue. Until recently, however, real-time generation of MR images has not been possible due to electronic technology limitations in scanner speed and central-processing-unit capabilities.

But at the Medical College of Wisconsin (Milwaukee, WI), Dr. Andre Jesmanowicz, assistant professor of biophysics, and his colleagues are about to overcome those limitations. By modifying a 3-Tesla Medspec magnetic- resonance-scanning system from Bruker Instruments Inc. (Billerica, MA), Jesmanowicz has reconfigured the scanner to acquire as many as 26 images per second. This acquisition rate is accomplished by using gradient-recalled echo-planar imaging, a technique that allows one image to be recorded every 40 ms rather than the usual time of one minute.

Developing this technique required redesigning the scanner to include a phased-array capability, which markedly improves the sensitivity of the received signal. This improvement was achieved by adding four RF receive coils and four 4261 VME-based analog-to-digital (A/D) converters from Pentek (Upper Saddle River, NJ) to the scanner system.

In operation, an amplitude- and phase-modulated, 3-ms, 1-kW signal is applied to the main RF head coil in the scanner. The resulting oscillating magnetic field excites hydrogen nuclei in the patient`s brain at different frequencies, which are captured by the four RF receive coils.

Because the main head-coil pulse is applied through a quad hybrid subassembly, the coil can act as both a receiver and a transmitter, allowing the scanner to display unprocessed images for preliminary analysis. After the imaging signals are captured by the RF head coil, they are mixed with a reference 123.75-MHz signal in four downconverters to produce difference signals in the 1--1.5-MHz range. These difference signals are first bandpass-filtered to reduce signal aliasing effects and then applied to the A/D converters, where they are sampled at 1 Msample/s. Because these signals are downconverted to lower frequencies, signal processing can be performed more economically.

Depending on the desired nuclear-magnetic-resonance scanning sequence, in some cases the bandwidth can be further reduced digitally to 125 kHz. Both in-phase and quadrature components are then computed by multiplying each digital output with digital sine and cosine functions. The resultant complex signals are used to compute backward propagation functions and to reconstruct images of the patient`s brain on a 748i workstation from Hewlett-Packard Co. (Palo Alto, CA). The net result, according to Jesmanowicz, permits real-time scanning of the human brain.