HIV/TB/MCH Screening

AI-powered diagnostic imaging with X-ray and POCUS enable a cost-effective, integrated approach to HIV/TB and Maternal healthcare. By consolidating screening into a single point of care, providers can more effectively manage HIV/TB co-infections and flag high-risk pregnancies, ensuring timely referrals for better patient outcomes.

Integrated AI-Enabled screening at primary care is a scalable intervention that accelerates progress toward national HIV, TB, and Maternal & Child health targets by 2030 at reduced cost. These innovations allow for a better triaging of patients in the high risk intersection of TB/HIV and Pregnancy.

AI for HIV/TB

TB remains a leading cause of death among PLWH, with the WHO reporting that individuals with HIV are 15-22 times more likely to develop active TB than those without HIV. The integration of digital X-ray and CAD into HIV programs for systematic screening for active TB can improve health outcomes for PLWH through early TB case detection and ruling out TB for Preventive Treatment (TPT).

Early detection of TB in PLWH is essential for reducing mortality and interrupting transmission, and digital X-ray with CAD offers a scalable and cost-effective solution to HIV programs to address these challenges. For HIV programmes, systematic TB screening represents an opportunity to improve patient outcomes while strengthening the delivery of comprehensive, integrated HIV services.

Digital X-ray provides a rapid, non-invasive, and highly sensitive means of detecting pulmonary abnormalities suggestive of TB. When integrated with CAD software, which uses artificial intelligence (AI) to identify and quantify radiographic abnormalities suggestive of TB, these tools can significantly enhance accuracy and efficiency of collaborative TB/HIV activities.

Digital X-ray/CAD enables the detection of early-stage symptomatic and asymptomatic TB and also atypical presentations common among PLWH. This rapid triage is particularly valuable in high-burden settings where access to large scale molecular testing is limited.

• Scientific studies confirm the impact of CAD-based screening that achieved comparable accuracy to expert radiologists in identifying presumptive TB on chest X-rays, significantly reduce the reliance on scarce human resources [1].
• In a high TB-HIV prevalence setting of Kenya, CAD4TBv6 met the optimum TPP; with the threshold set to achieve a mean sensitivity of 95% (optimum TPP), specificity was 83.3%, at CAD4TBv6 threshold: 55. [2]
• Various studies confirm that CAD is effective in identifying TB also among HIV-positive individuals with atypical presentations, outperforming symptom screening.
• As TB may present with atypical or minimal radiographic abnormalities in PLWH, the diagnostic accuracy of all currently available CAD algorithms is lower than in HIV-negative populations. X-ray/CAD should therefore be used as part of a comprehensive WHO-recommended screening pathway, including symptom screening and confirmatory diagnostic testing where indicated, rather than as a standalone diagnostic test.

Key benefits X-ray/CAD screening for collaborative HIV/TB programs

• The investment in digital X-ray/CAD screening among PLWH is justified not only by the clinical benefits but also by the potential cost savings. Early detection and treatment of TB prevent severe disease, reducing hospitalization costs. Additionally, CAD minimizes the need for highly trained staff, making it a sustainable option for resource-constrained settings also for TB Preventive Treatment (TPT).
• Collaborative HIV/TB activities, such as systematic TB screening of PLWH aligns with the WHO End TB Strategy and the UNAIDS’s 95-95-95 targets for HIV care. By integrating digital X-ray and CAD into HIV programs, health systems can bridge the gap between HIV and TB care, contributing to the dual goals of reducing HIV/TB mortality and morbidity for PLWH.

AI for HIV/TB & Maternal Health

Key Challenge

Approximately 95% of maternal deaths occur in low-resource settings, where timely diagnosis and effective management of obstetric complications remain major challenges. The high prevalence of TB and HIV co-infection in these settings poses additional risks to pregnant women, further increasing the burden of maternal morbidity and mortality.

Active TB disease in HIV-infected pregnant women increases the risk of maternal mortality by nearly 300%. [3] Maternal TB increases the risk of mother-to-child transmission (MTCT) of HIV 2.5-fold [4]. Meta-analysis indicates that HIV-infected women had a risk of a pregnancy-related death that was 8 times higher compared to HIV-negative women [5].

Pregnant women with advanced or untreated HIV are much more susceptible to serious, life-threatening opportunistic infections, including TB, severe pneumonia, and malaria. Therefore, integrated screening for HIV/TB and obstetric risks can be accomplished during pregnancy if these activities are combined.

HIV can also exacerbate direct pregnancy complications due to a higher risk for maternal sepsis, anemia and postpartum hemorrhage. The risk level depends on whether a woman has access to Antiretroviral Therapy (ART).

• Without Treatment: The risk of maternal death is at its highest. Advanced HIV (low CD4 cell counts) coupled with the physical stress of pregnancy drastically increases the likelihood of fatal complications.
• With Treatment (ART): lower risk as viral load becomes suppressed, the immune system stays strong, and maternal mortality risk drops close to that of an HIV-negative woman. Furthermore, the initiation of ART commonly reduces the risk of MTCT to less than 1%.

Screening Access Gap

Ultrasound can support obstetric risk screening by providing information on gestational age, fetal number and growth, fetal anatomy, placental location, and other features associated with pregnancy complications. Early identification of pregnancies requiring further assessment or closer monitoring supported by ultrasound scans facilitate timely referral and management.

However, prenatal ultrasound screening is routinely used in wealthier countries and is recommended by the WHO, but the equipment cost and the need for trained sonographers make conventional ultrasound largely inaccessible in many regions of the world. To effectively address this access and equity gap in resource-constrained setting, innovative ultra-portable Point-of-Care Ultrasound (POCUS) powered by proven Artificial Intelligence (AI) offers a potential solution.

Solution

AI-Enabled Point-of-Care Ultrasound

Low-cost, smartphone-based POCUS devices with AI-assisted image acquisition and interpretation, enable primary care providers to perform the entire screening workflow offline and at a fraction of the cost of conventional ultrasound after a 2-hour training. This approach could significantly expand access to ultrasound screening for earlier detection and management of several life-threatening pregnancy complications in resource-limited settings leading to better Maternal and Child Health (MCH) outcomes.

The deployment of AI-Enabled POCUS at health center level is expected to deliver transformative improvements in maternal and newborn health outcomes. By enabling non-specialist health workers to perform basic obstetric scans, POCUS addresses a critical gap in the early detection of pregnancy-related risks, especially in rural and underserved regions.

Data-Driven POCUS Screening

• Improving maternal health SDG 3.1: to reduce maternal mortality ratio to < 70 per 100,000 live births by 2030; 3.2: to reduce neonatal mortality < 12 deaths per 1,000 live births.
• WHO emphasizes that all pregnant women should receive quality ANC by skilled providers to detect complications early and contribute to better health for both mothers and newborns.
• AI-Enabled POCUS like the Delft Imaging “Baby Checker” allows primary-care lworkers to conduct routine screenings, so potential complications are identified early for timely referrals.
• Access to ultrasound tends to stimulate demand for ANC in LMIC.
• Women attending ANC are more likely to receive an HIV test than those not attending, so every ANC visit is an opportunity for voluntary HIV testing, the detection of TB and PMTC.
• The POCUS “BabyChecker” AI analyses the 6-sweeps scan in real time and generates a risk assessment with a rapid output on: gestational age, fetal presentation, placenta localisation, estimated fetal weight, and detection of multiple gestations.




POCUS/AI use in Sierra Leone

• Each POCUS scan produces standardized data, which can be aggregated into a central dashboard to monitor the number of women accessing care, the timeliness of contact, the frequency and type of risks detected, risks per age group, risks among TB/HIV+, risks per geographic location. Programs obtain an overall real-time insight in trends essential for targeted interventions and continuous improvement.


• AI from the specialist EPCON facilitates targeted POCUS interventions by analyzing and visualizing routine data for decision-making in line with WHO's MNCAH guidance. From “collecting information about mothers and babies” → “turning every data point into an actionable next step.”

Detect → Control → Act → Refer → Follow up → Measure outcome → Predict



• AI-Enabled POCUS interventions gives MCH providers a dashboard showing:
  • Where high-risk pregnancies are concentrated
  • Where pregnant women are at a higher risk of TB/HIV (co)-infections
  • Which facilities have the highest referral rates
  • Which communities have the most missed ANC visits
  • Where in the country pregnancy danger signs are increasing
  • Where referral completion is lacking
  • Which social and contextual factors are predictive for high-risk pregnancies.
  • MCH performance per geography, facility or socioeconomic group

For more cross-disease impact, the same POCUS device can in the near future also be used for bedside radiation free lung health screening by first line health workers at health centers using the LUS4TB AI algorithm. The integrated AI model analyzes lung images directly on the device without requiring internet connectivity.

Key benefits POCUS for HIV/TB programs

Global and domestic funding expanded HIV testing and access to anti-retroviral treatment to prevent MTCT, but many pregnant women in LMIC still do not know their HIV status. Missed ANC visits lead to missed voluntary HIV testing opportunities. These gaps undermine the effectiveness of PMTCT programs, as unidentified women cannot access timely antiretroviral therapy. ANC is the main entry point to reach pregnant women for HIV testing and PMTCT [6]. However, there are two main gaps:

• Not all pregnant women attend ANC, particularly in rural and low-resource settings, due to limited awareness, long distances, travel costs, low perceived value, and cultural barriers.
• HIV grants fund infrastructural capacity and commodities, yet a portion of the target group never reaches or fully uses these services [7]. Studies show that women attending ANC are significantly more likely to receive an HIV test than those not attending.

• When POCUS scans become available at primary care level, experience shows that it is an additional reason to attend ANC, hence Ultrasound stimulates demand.
• Integrated primary care screening facilitates earlier detection of TB, HIV, maternal risks, and initiation of PMTCT supporting the elimination of mother-to-child transmission (eMTCT)
• If HIV Voluntary Counselling and Testing (VCT-HIV) is routinely offered with a POCUS scan, then every additional woman attending ANC is an opportunity for HIV testing and PMTCT.
• More ANC visits create more opportunities to systematically screen pregnant women for TB, TB/HIV co-infection facilitating earlier detection, diagnosis, and timely treatment initiation.
• POCUS supports the timely detection and management of high-risk pregnancies with the right interventions and timely referrals to mitigate this high-risk intersection TB/HIV/Pregnancy.
• AI-enabled screening for TB and other lung diseases with X-ray or POCUS and obstetric risk screening with POCUS (e.g. BabyChecker) do not require specialists to interpret the images.
• POCUS with LUS4TB supports frontline health workers in detecting tuberculosis earlier by guiding them through simple lung scans with AI analyzing the images directly on a smartphone, even without internet access.
• POCUS LUS4TB achieved an AUROC of 0.97 for detecting lung abnormalities, with 90 percent sensitivity at 70 percent specificity for TB detection on unseen test data.
• Integration and data improve provider efficiency and can lower screening and care cost with enhanced prevention, risk prediction and better treatment outcomes.
• AI reduces cost through data-driven interventions to identify the right mother or newborn, at the right time, and trigger the right intervention.

Conclusions

Integrated AI-Enabled screening at primary care can improve health outcomes:

• detection of TB/HIV co-infection
• increased ANC attendance
• improved access to patient-centered integrated screening
• enhanced referral pathways especially in the high risk TB/HIV/MCH intersection
• earlier interventions to reduce preventable mortality and MTCT.

Routine data, combined with AI-driven analytics:

• measure outcomes, identify trends, and predict maternal risks
• support better-informed decisions and improved outcomes over time

Early users of POCUS/AI

Lion Heart Foundation, UNFPA Honduras, PharmAccess/Kisumu County, CUAMM, Cordaid Ethiopia and Uganda PSI are among the early users.

• In Kisumu County, Kenya, a routine antenatal visit can mean a 30-kilometre journey. One many expectant mothers never make. Without regular monitoring, complications go undetected, contributing to high maternal and neonatal mortality in the region. BabyChecker is how digital health contributes to close the gap, by bringing lifesaving care to the doorstep of women who need it most.
  https://www.pharmaccess.org/

• Cordaid was a partner in the Dutch supported R&D consortium with FMOH Ethiopia, ORHB, EngenderHealth, Radboud University, CUAMM and Delft Imaging to develop the POCUS BabyChecker.
• Lion Heart Hospital in Sierra Leone was the first to include POCUS BabyChecker in routine ANC use, providing valuable operational feedback

---