We have compared the use of two (93 and 185 keV) and three (93, 185, and 300 keV) photopeaks for Ga-67 tumor imaging and optimized the placement of each energy window. The figures of merit were a Bayesian signal-to-noise ratio (SNR) for detection of spheres embedded in a realistic anthropomorphic digital torso phantom and an ideal SNR for estimation of their size. Seven spheres of radii ranging from 1 to 3 cm, located at several sites in the torso, were simulated using a realistic Monte Carlo program. We also calculated the detection metric from simple phantom acquisitions. For detection and estimation tasks, the optimum windows were similar for all sphere sizes and locations. For the 93 keV photopeak, the optimal window width for detection and estimation tasks was 87-102 keV. This window is slightly narrower than the usual 20% window used in the clinic (83-101 keV). For the 185 keV photopeak, the optimal window was 170-215 keV for detection and estimation tasks; this is substantially different than the 15% window used in our clinic (171-199 keV). For the 300 keV photopeak, the optimal window was 280-330 keV for detection and estimation tasks, substantially wider than the 15% window (278-323 keV) used in our clinic. Using the three optimized, rather than the two lower-energy, windows yielded a 9% increase in SNR for detection of the 3 cm-radius sphere and a 7% increase in SNR for estimation of its size. Detection SNR calculated from the acquired phantom data increased by 9% when three optimized, rather than two, energy windows were used. Using three, rather than two, Ga-67 photopeaks should yield better performance in both detection and quantitation tasks.