Access to WHO-recommended obstetric ultrasound before 24 weeks of gestation remains limited in Sierra Leone due to infrastructure, training, and equipment constraints. Artificial Intelligence (AI) enabled smartphone-based ultrasound offers a potential solution to improve access and quality of maternal care. This study assessed the usability of this technology during a pilot implementation in Tonkolili District since October 2020.

Methods

A convergent parallel mixed-methods design was employed. Quantitative data from 2,315 obstetric sweep protocols conducted by 81 healthcare providers across seven facilities (October 2020–March 2023) were analysed using mixed-effects logistic regression in Stata 17 to identify factors influencing scan quality. Scans were initially reviewed through remote expert interpretation, with AI-enabled analysis introduced later; thus, the evaluation reflects both remote and AI-supported workflows. Qualitative data from interviews, focus group discussions, and observations with providers and clients were analysed thematically in NVivo, guided by the Non-Adoption, Abandonment, Scale-up, Spread, and Sustainability (NASSS) framework.

Results

Overall, 83.8% of obstetric sweeps were of sufficient quality for remote or AI analysis, and 59.3% of providers achieved the 80% threshold of sufficient quality images. Tutorial videos (OR: 14.1, 95% CI: 6.4–31.1, p < 0.001) and provider skill level significantly improved scan quality. The device was well accepted by providers and clients. However, infrastructure constraints, supply chain limitations, absence of comprehensive guidelines, and funding challenges hinder broader impact.

Conclusion

AI-enabled ultrasound is a feasible and usable tool in resource-limited primary care settings. Addressing system-level barriers will be essential for sustainable scale-up.