It was gradually recognised that HIV-associated TB is more likely to be sputum smear negative [3]. This discovery led to a 2007 WHO recommendation for prompt CXR in PLHIV with smear-negative presumptive TB [4]. In addition, the development of molecular tests that are more sensitive, more specific and substantially more expensive than smear microscopy created the need for a screening test to identify patients who should be evaluated with the new molecular tools [5]. Symptom-based screening is a limited tool for this use case, suffering from a high number needed to screen in most populations [6–8]. By contrast, with its high sensitivity and relatively low cost, CXR is an excellent tool for this use case [9], including in PLHIV, although its sensitivity is lower than among HIV-negative people, especially in children [10]. Finally, prevalence surveys show that a substantial proportion of TB cases are asymptomatic but detectable by CXR followed by microbiological testing, a state that has been reclassified as subclinical TB [11] and which is likely to play an important role in TB transmission at the population level [12]. Given the 2015 End TB Strategy’s focus on the early detection and treatment of all patients with TB [13, 14], imaging has a vital role to play as a screening tool and to support clinical diagnosis in people who are unable to produce sputum [15, 16]. In recent years, the pace of technological development in the imaging of TB has been rapid, with substantial advances made in artificial intelligence (AI) and computer-aided detection (CAD) of CXRs, and with the arrival on the global market of new portable and ultra-portable CXR and point-of-care ultrasound (POCUS) devices. In this chapter, we review recent developments in CXR and POCUS hardware and interpretation, as well as the evidence and new global policy surrounding their use. CXR In 2007, the WHO noted that “the limitations that exist on the wider use of CXR, such as nonavailability at peripheral health facilities and the difficulty of interpreting results, even by trained physicians, need to be addressed, including through training” [4]. Sixteen years later, these limitations largely remain in place. Although CXR is a relatively inexpensive technology per patient when compared with molecular tests for TB, the upfront investment costs in terms of both hardware and the requirement for skilled technicians and radiology staff to create and interpret the images limit its availability in many low-resource, high-TB-burden settings, especially at the primary-care level [17, 18]. A 2015 survey of national TB programmes in 22 countries found that cost was a major barrier to the rollout of CXR in 68% of countries, with 73% lacking equipment such as vans to enable transportation of CXR machines and 59% lacking qualified readers such as radiologists [19]. A 2021 study of diagnostic availability in 10 low–middle-income countries found that only 61.5% of surveyed hospitals had access to any X-ray equipment [20], with access at the community level, where many patients first seek care, likely to be substantially worse [19, 21, 22]. These access limitations have important consequences, as CXR is currently recommended by the WHO in several different use cases for the prevention and diagnosis of TB (table 1). Given these recommendations by the WHO, there has been substantial interest in recent developments in both CXR hardware and CXR interpretation software that may help to bridge these access gaps. CXR hardware There are two stages to producing a CXR image. First, beams of X-rays produced by a generator are fired through a patient’s chest into an image receptor, creating a latent image. https://doi.org/10.1183/2312508X.10024322 79 IMAGING AND DIAGNOSIS |J. BIGIO ET AL.