We investigate the ability of the Laser Interferometer Space Antenna (LISA) to measure the center of mass acceleration of stellar-origin black hole binaries emitting gravitational waves. Our analysis is based on the idea that the acceleration of the center of mass induces a time variation in the redshift of the gravitational wave, which in turn modifies its waveform. We confirm that while the cosmological acceleration is too small to leave a detectable imprint on the gravitational waveforms observable by LISA, larger peculiar accelerations may be measurable for sufficiently long lived sources. We focus on stellar-mass black hole binaries, which will be detectable at low frequencies by LISA and near coalescence by ground based detectors. These sources may have large peculiar accelerations, for instance, if they form in nuclear star clusters or in active galactic nucleus (AGN) accretion disks. If that is the case, we find that in an astrophysical population calibrated to the LIGO-Virgo observed merger rate, LISA will be able to measure the peculiar acceleration of a small but significant fraction of the events if the mission lifetime is extended beyond the nominal duration of 4 years. In this scenario LISA will be able to assess whether black hole binaries form close to galactic centers, particularly in AGN disks, and will thus help discriminate between different formation mechanisms. Although for a nominal 4-year LISA mission the peculiar acceleration effect cannot be measured, a consistent fraction of events may be biased by strong peculiar accelerations which, if present, may imprint large systematic errors on some waveform parameters. In particular, estimates of the luminosity distance could be strongly biased and consequently induce large systematic errors on LISA measurements of the Hubble constant with stellar-mass black hole binaries.