Tuberculous meningitis (TBM) is a devastating form of tuberculosis (TB), and key TB antimicrobials, including rifampin, have restricted brain penetration. A lack of reliable data on intralesional drug biodistribution in infected tissues has limited pharmacokinetic (PK) modeling efforts to optimize TBM treatments. Current methods to measure intralesional drug distribution rely on tissue resection, which is difficult in humans and generally limited to a single time point even in animals. In this study, we developed a multidrug treatment model in rabbits with experimentally induced TBM and performed serial noninvasive dynamic ¹¹C-rifampin positron emission tomography (PET) over 6 weeks. Area under the curve brain/plasma ratios were calculated using PET and correlated with postmortem mass spectrometry. We demonstrate that rifampin penetration into infected brain lesions is limited, spatially heterogeneous, and decreases rapidly as early as 2 weeks into treatment. Moreover, rifampin concentrations in the cerebrospinal fluid did not correlate well with those in the brain lesions. First-in-human ¹¹C-rifampin PET performed in a patient with TBM confirmed these findings. PK modeling predicted that rifampin doses (≥30 mg/kg) were required to achieve adequate intralesional concentrations in young children with TBM. These data demonstrate the proof of concept of PET as a clinically translatable tool to noninvasively measure intralesional antimicrobial distribution in infected tissues.