314 B. Amaoui et al.: Radioprotection 2025, 60( 4), 310 – 317
report limited proficiency in QA( 30 % of G1 and only 7.7 % of G2). This indicates that quality assurance is an underaddressed aspect in training and practice, which could have implications for treatment safety and efficacy. Enhanced training in this area seems necessary, especially for residents.
The results also show that there is no significant difference between the two groups G1 onco-radiotherapists and G2 residents in terms of knowledge of AI( p = 0.075). The results reveal that 60 % of G1s report moderate knowledge of AI, compared to 72 % of G2s who report having little knowledge. This indicates that residents, although younger and potentially more exposed to emerging technologies, still lack structured training in AI. In addition, the majority of AI knowledge is acquired through self-study, which highlights a absence of structured continuing education. In the same context, Ryan et al., 2021 showed that only 19.6 % of diagnostic radiologists and 28.36 % of radiotherapists had followed formal training in AI( Ryan et al., 2021). Existing educational systems in medicine, in general, do not include AI-focused courses, and there are not enough teachers or practitioners qualified in AI technologies also limits the use of AI( Semghouli et al., 2025). These results argue in favour of a more systematic integration of AI in training courses, particularly in radiotherapy. All participants, both G1 and G2, demonstrated high interest in learning more about AI, especially its clinical applications, theories, and ethical implications. This enthusiasm is a encouraging for AI adoption of AI in radiotherapy. However, the lack of current knowledge and adequate training could hinder adoption. Preferences for face-to-face workshops( 60 % of G1 and 53.8 % of G2) suggest that interactive and practical methods are preferred for learning AI, although a significant proportion of participants are receptive to self-directed online training. In fact, formal training would be useful with more practical applications and more teaching on how to integrate AI into clinical practice( Hindocha et al., 2023).
The results in Table 2 show that several functions in radiotherapy, such as image acquisition and reconstruction, image fusion, and plan optimisation, are partially automated. However, critical tasks such as contouring target volumes and organs at risk remain predominantly manual( 50 % of G1 and 77 % of G2 for organs at risk contouring). This reflects an opportunity for AI to play a greater role in automating these tasks, which could improve both productivity and accuracy. Indeed, in their study, Hindocha et al., 2023 reported that 45 % of respondents indicated clinical use of AI contouring clinically in their department. 16 % reported that although it was not currently used, their department planned to introduce it in the next year. They added that while this was mainly for OAR contouring, respondents reported using AI for prostate, thoracic and bladder tumour contouring( Hindocha et al., 2023). earlier studies have noted that the concordance of positive results for the clinical use of AI segmentation and OARs is reassuring( Warren et al., 2023). As an increasing number of organs at risk are considered in treatment planning( due to improved imaging quality and treatment accuracy), automation will help to avoid workflow congestion, freeing up valuable time for other tasks requiring human interaction( Korreman et al., 2020).
Tables 3 and 4 reveal that the majority of participants, both G1 andG2, are in favour of increasing productivity and quality in almost all radiotherapy functions through AI. For example, 90 % of G1 and 77 % of G2 want improved quality in image acquisition and reconstruction. These results indicate that professionals see AI as a potential tool to streamline their workflow and improve patient outcomes. The introduction of machine learning methods in image reconstruction to replace conventional reconstruction techniques has reduced artefacts and potentially increased reconstruction quality and consistency( Kida et al., 2018). In addition, deep learning methods are being developed that can automatically and rapidly perform rigid and deformable image registration( De Vos et al., 2019). Furthermore, the use of automated methods for treatment planning that have been introduced over the last five years has shown strong potential for improving the efficiency and quality of treatment plans( Hansen et al., 2016). In terms of quality assurance, Automation and data mining can be used to optimize quality assurance schedules and to auto-detect and identify errors / deviations.( El Naqa et al., 2019). However, uncertainty persists about the impact of AI on their professional role, particularly among G1s( 40 % believe AI will affect their role). Participants are generally optimistic about the impact of AI in radiotherapy. A majority of G1s( 70 %) and G2s( 84.6 %) believe AI will increase job satisfaction and have a positive impact on the patient care pathway. However, a significant proportion of participants( 40 % of G1s and 31 % of G2s) take a neutral position, waiting for the“ black box” aspect of AI to become more transparent before fully adopting it. This highlights the importance of developing explainable and transparent AI systems to gain clinicians’ trust.
The survey revealed several barriers to artificial intelligence( AI) adoption in Moroccan healthcare, especially in radiation oncology. Like other African countries, Morocco lacks essential digital infrastructure such as electronic medical records, integrated databases, and reliable connections between hospital departments( Eric Naab Manson et al., 2023). Staff must manually gather data from paper records, which hampers historical research and AI development( Hassan Abdelilah Tafenzi et al., 2025). The country has no publicly available AI models that use biomarkers or common imaging data( MRI, CT) required for AI training and testing. Financial and technical constraints, including limited cloud storage and insufficient computing resources, also restrict broader AI implementation. Healthcare professionals have few training opportunities, with many radiation oncologists and residents lacking necessary AI knowledge and experience to use these tools effectively. Additional concerns include questions about AI system reliability and accuracy. Clinicians are also concerned that AI might undermine their professional autonomy and expertise, while creating potential ethical and legal issues.
The introduction of AI methods into medical practice in Morocco is remains at an early stage. The absence of structured patient data and limited training in AI continue to be major barriers. Therefore, ongoing research is essential on the subject, respond to concerns and expectations of healthcare professionals, and advocate for the benefits of AI in the healthcare sector( Sami et al., 2023). In addition, it is necessary to identify practical solutions to the constraints linked to workforce training, data rights frameworks, local equipment, infrastructure, and national regulatory frameworks( Edzie et al., 2023). Staff and students be adequately trained to prepare them to work with these new technologies( Doherty et al., 2024).
Several studies conducted in Morocco have raised the issue of the variability of radiation protection practices( Amaoui et al., 2023; El fahssi et al., 2024a) and the substantial dose