Abstract: This dissertation describes magnetic resonance imaging (MRI) of protons performed in a precession field of 132 μT. In order to increase the signal-to-noise ratio (SNR), a pulsed 40-300 mT magnetic field prepolarizes the sample spins and an untuned second-order superconducting gradiometer coupled to a low transition temperature superconducting quantum interference device (SQUID) detects the subsequent 5.6-kHz spin precession. Imaging sequences including multiple echoes and partial Fourier reconstruction are developed. Calculating the SNR of prepolarized SQUID-detected MRI shows that three-dimensional Fourier imaging yields higher SNR than slice-selection imaging. An experimentally demonstrated field-cycling pulse sequence and post-processing algorithm mitigate image artifacts caused by concomitant gradients in low-field MRI. The magnetic field noise of SQUID untuned detection is compared to the noise of SQUID tuned detection, conventional Faraday detection, and the Nyquist noise generated by conducting biological samples. A second-generation microtesla MRI system employing a low-noise SQUID is constructed to increase SNR. A 2.4-m cubic, eddy-current shield with 6-mm thick aluminum walls encloses the experiment to attenuate external noise. The measured noise is 0.75 fT Hz 1/2 referred to the bottom gradiometer loop. Solenoids wound from 30-strand braided wire to decrease Nyquist noise and cooled by either liquid nitrogen or water polarize the spins. Copper wire coils wound on wooden supports produce the imaging magnetic fields and field gradients. Water phantom images with 0.8 x 0.8 x 10 mm3 resolution have a SNR of 6. Three-dimensional 1.6 x 1.9 x 14 mm3 images of bell peppers and 3 x 3 x 26 mm3 in vivo images of the human arm are presented. Since contrast based on the transverse spin relaxation rate (T₁ ) is enhanced at low magnetic fields, microtesla MRI could potentially be used for tumor imaging. The measured T₁ of ex vivo normal and cancerous prostate tissue differ significantly at 132 μT. A single-sided MRI system designed for prostate imaging could achieve 3 x 3 x 5 mm3 resolution in 8 minutes. Existing SQUID-based magnetoencephalography (MEG) systems could be used as microtesla MRI detectors. A commercial 275-channel MEG system could acquire 6-minute brain images with (4 mm)3 resolution and SNR 16.