Artificial Intelligence in Radiation Treatment Planning: A Survey-Based Observational Study From A Cancer Centre in Nigeria

Introduction Low- and middle-income countries (LMICs) like Nigeria face rising cancer incidence and mortality, with late-stage presentation and limited resources. Only eight government-funded radiotherapy centres serve a population of 223.8 million—far below the estimated 280 radiotherapy machines required. To increase patient throughput we evaluated integration of AI auto-contouring tools to expedite treatment planning, specifically target and organ-at-risk delineation. Materials and Methods We performed an observational, survey-based study of radiation oncology staff at our Cancer Centre. Participants were consultant and resident oncologists and medical physicists. The survey compared time spent using AI auto-contouring versus manual contouring and collected perceptions of impact, benefits, and limitations. Results Thirty-one staff responded: 20 (64.5%) oncologists and 11 (35.5%) medical physicists. Experience with AI varied (33% ≤ 6 months; 13% ≈2 years). Respondents reported increased confidence in planning: 11 (35%) moderate, 12 (39%) moderate–high, and 8 (26%) high. Common limitations were licence availability (20, 64.5%) and technical expertise (19, 61.3%). Most respondents (20, 65%) would recommend the tool. The principal benefit was improved workflow efficiency (25, 81%). AI-assisted planning significantly reduced planning time for most tumour sites; sites with complex anatomy showed no time benefit, reflecting the need for intensive manual correction. Conclusion Deployment of AI auto-contouring at a Nigerian cancer centre reduced planning time for most sites and improved clinician confidence, but complex anatomical regions still require detailed manual oversight and additional AI training. AI tools can increase throughput in LMIC radiotherapy services, though licensing, infrastructure, and training barriers exist and must be addressed to ensure safe implementation. Future work should include multi-centre validation, formal inter-rater reliability assessment, and prospective patient-level outcome evaluation and cost-effectiveness analyses.