Background
Healthcare is quickly evolving, and current technological developments are centred largely on the increasing integration of complex computerised algorithms into equipment modalities [
1‐
3]. Artificial intelligence (AI) is a key component of these complex algorithms and is currently applied innovatively in healthcare because of its reported advantages and the potential to improve patient care [
2,
4]. AI has been widely used in many clinical circumstances to diagnose, treat and predict the outcomes [
5]. Specifically, healthcare and research applications include prediction of disease prognosis and response to treatment, drug development, remote patient observations, medical data management, digital patient consultation and in some cases administrative hospital management [
6‐
8].
In medical imaging practice, AI has shown impressive accuracy and sensitivity in the identification and characterisation of abnormalities leading to enhanced service delivery and quality of patient care [
9]. Increasing developments of the technology has propelled its application beyond image interpretation in medical imaging practice. Currently, AI systems are employed across equipment modalities to improve image acquisition and quality of care [
1,
10,
11]. The impact of AI in medical imaging has been summarised in a publication by Lewis and colleagues [
4]. They indicated that AI would lead to workflow improvement, production of high-quality images, improvement in image interpretation, image segmentation, automated image registration and radiomics analysis and dose optimisation, automate scheduling and protocolling and reduce scanning times. These would improve the work of radiographers, radiologists and other medical imaging staff and help deal with workloads which have led to staff shortage. It is indeed thought to revolutionise the entire imaging and healthcare industry in the near future. Studies are exploring the use of AI in the provision of human trait-related solutions like empathy and companionship to enhance quality of life through systems
like “Vits” [
12]. This has instigated arguments about the roles and responsibilities of healthcare professionals including radiographers.
Of note, referring physicians will have responsibilities as well, since they decide on therapeutic management based on the imaging report (and often multidisciplinary team discussions), including evaluation on early response on therapy. In addition, many (minimally invasive) current interventional procedures are image-guided, based on AI-processed images and fusion of different modalities. As a result of the broad areas that AI could be useful, it is anticipated AI would greatly bear on the radiologist’s regular routine and improve practice, but it would not entirely change their core practice [
9].
Notwithstanding, the medical community must anticipate the potential unknowns and professional requirements of this technology to ensure effective, continuous and safe incorporation into diagnostic imaging practice [
1,
10,
13]. This has initiated arguments about the roles and responsibilities of imaging professionals who will use this technology, particularly, radiographers and radiologists [
13,
14]. Lewis and colleagues [
4] reported that some radiographers may have a scary or exciting perception about AI, and these could be heightened by the thought of having an ‘AI work colleague’ in the radiology department. This clearly suggests that AI systems may need to fulfil certain conditions to be fully embraced by health professionals and the society [
15].
To understand the views and readiness of radiographers on the use of AI in medical imaging practice, some studies have been conducted [
16‐
18]. These studies explored the perspective of radiographers on the integration of AI into medical imaging practice to identify factors that could help improve the implementation process. However, these studies mostly employed quantitative approaches, with associated methodological limitations relating to the depth of perspectives provided. To the best of our knowledge, there is rarity of studies employing qualitative designs to explore perspectives on AI in medical imaging practice. Radiographers serve as the interface between technologies and their patients [
14]; thus, this study qualitatively explored the perceptions of this workforce on the use of AI in practice in Africa—a low-resource setting. Qualitative opinions about new developments could be predictable; thus, it is expected that radiographers as end users would have some concerns about AI. However, radiographers in Africa would have unique concerns due to the resource challenges they deal with at work.
Methods
The study is a sequel to a quantitative study [
18] which concluded with an open-ended question for participants to share their views about AI. Purposive sampling was used as it is the most appropriate approach for content analysis inquiry [
19]. The sample used consisted of radiographers across the five geographical regions of Africa with different cultural and linguistic backgrounds. To ensure anonymity of the participants, the radiographers were assigned pseudonames such as “Rad 01”. In the COVID-19 pandemic, semi-structured face-to-face interview was not possible and to maximise response across multiple countries in Africa, open-ended online surveys were employed for this study.
This exploratory cross-sectional online survey was hosted on Google Forms (Google, Mountain View, CA). The instrument broadly included questions relating to demographics, general attitudes, and perspectives on AI and how it should be implemented in Africa, job security, the future of medical imaging including workforce development and ethics in relation to the integration of AI (Additional file
1: Appendix 1- Questionnaire). The free text comments provided in response to the open-ended question “any other comments” are of interest because information provided by such open-ended questions is used to substantiate answers to a structured questionnaire [
20]. The use of open-ended questions provided study participants, the chance to express their judgements about an issue [
20]. Qualitative survey studies are essential for determining the diversity of opinions, perceptions and experiences on a topic of interest within a given population [
21]. The non-anglophone responses (Arabic and French) were translated into English by bilingual academics who are radiographers and members of the research team. No transcription was done because participants provided textual response. To ensure a robust analysis and interpretation, the obtained data were collaboratively analysed by three researchers (W.K.A., B.O.B., T.N.A.) using qualitative content analysis—a recommended approach for the analyses of textual data [
22]. Trustworthiness for the content analysis was ensured using established criteria [
23]. They argue that the purpose of using trustworthiness in a qualitative study is to back the argument that the research outcome is “worth paying attention to”.
Ethical consideration
The Ethics and Protocols Review Committee of the School of Biomedical and Allied Health Sciences of the University of Ghana granted approval for the study (SBAHS/AA/RAD/29245/2019–2020).
Discussion
The current study is a follow-up to a previous quantitative inquiry [
18], and qualitatively explored in-depth, the perception of radiographers in Africa about AI implementation in practice. Although the quantitative aspect [
18] had 1020 respondents, only approximately half (
n = 475) provided valid responses to the open-ended survey questionnaire. This was an opportunity for participants to provide unrestrictive views about what they perceived about AI implementation and usage in Africa.
Participants claimed that they were aware about AI (although their actual levels of awareness could not be ascertained) which is yet to gain grounds in Africa and had very positive expectation for the technology to change the traditional mode of medical imaging practice. This is consistent to the views reported in previous study from Ghana where 86.1% of the study participants expressed an awareness of AI in medical imaging practice [
17]. Participants in the current study also shared that considering the numerous advantages, AI was certain to be the right technology for use in Africa. Besides they were confident that AI would bring about several innovations and growth in medical imaging practice and beyond (as it is already doing in many disciplines in medicine such as nuclear medicine, pathology, laboratory, genetics). Despite this positive stance, some participants accepted AI with mixed feelings, with concerns that the workforce may be affected negatively. Interrogations about AI application seem to cloud the minds of various healthcare institutions [
24]. Many African countries are still using conventional radiographs with wet-image processing techniques and no wonder one participant expressed an attachment to the usage of conventional radiographs and would not probably like to see any change. A seeming misunderstanding centred on the possibility of reduced professional–patient interaction which to some might affect the psychological aspects of care. This claim is somehow doubtful as AI will not take away the radiographer–patient interaction opportunities that aid in understanding a patient’s psychological state (e.g. level of pain) and tailor care to offset any unusual emotions. On the contrary, the radiographer’s work involves direct patient interaction [
25]. Patients would still need to be positioned by the radiographer for a procedure which provides the opportunity for the radiographer to communicate with the patient before the examination. The radiographer is obliged to pay attention to the psychological state of the patient to understand the diverse emotional states to inform patient-centred care for better outcomes [
26].
As previously reported [
17], participants in the current study also indicated that AI was not appropriate for Africa because these tools will lead to diagnostic medical errors, considering the associated margins of error with all mechanical systems. This appears to be an erroneous perception. Evidence suggests that the AI tools would rather improve the practice of the profession including high-quality diagnosis and minimal errors [
2,
4,
5,
27]. A study [
27] also indicated that AI tools provide better diagnostic decisions (thus, it takes into consideration the results from every medical technical result, e.g. radiology, nuclear medicine, pathology and others to produce a single and personalised diagnostic report). They also improve treatment outcomes and reduce medical errors, thus making it easier for medical professionals to care for a larger number of patients. For example, several AI-assisted diagnostic equipment currently has advanced cancer identification and detection functionalities, particularly in areas such as screening mammography, lung cancer screening and histopathological assessment of breast images [
4]. This example leads to the argument that AI has certainly improved medical imaging, by demonstrating remarkable precision and sensitivity in the identification and classification of anomalies leading to improved care [
9]
. In medical imaging practice, AI could immensely improve the contribution of radiographers to processing bulky images, as for instance, chest radiographs. Even a simple AI-based tool to differentiate 'normal' from 'abnormal' could be helpful to reduce the radiologist's reporting workload and improve significantly radiological reporting (e.g. in depth and broadening even beyond the clinical context or radiological question) or particularly assist radiographers in rural and other low-resource settings where radiologists are not available. It must be emphasised also that the role of the radiologist will not change totally in the era of AI. The radiologist will be the responsible 'validator' of the radiological report, whatever additional AI-based tools are used.
Participants raised concerns about poor equipment maintenance culture in Africa which to some might not make AI a sustainable technology for Africa. Studies on medical equipment infrastructure in the West African sub-region attest to among other infrastructural challenges, poor equipment preservation, poor-quality management systems and obsolete medical equipment [
28‐
30]. Besides, they expressed their uncertainties in areas such as cyber security, safety of patient data and software corruption of AI tools. This could be addressed if the healthcare industry antedates any potential surprises and meets professional requirements for AI technology to enable uninterrupted and effectual use of the technology in diagnostic imaging [
1,
10]. Participants were doubtful about the cost of the AI tools and its implementation in Africa considering the economic challenges that some African countries face. Economic challenges mostly lead to broken technologies as a result of poor maintenance, and this is likely to be the same in the case of AI as the cost of implementation and maintenance may be too expensive for underdeveloped countries, pushing them further behind in improving healthcare [
27]. In contrast, one participant perceived AI as an important tool in medical imaging, as it is in other fields. AI was seen as the way forward to aid accurate diagnosis and reduce errors as well as cost of expensive manpower and studies output as indicated in some publications [
1,
10,
31].
Despite AI acceptance among most of the respondents, they were, however, anxious about the probable negative impact AI would have on their workforce. They argued the potential of knowledge gap that the acceptance of AI in Africa would lead to job losses and also make radiography education programmes irrelevant. This bone of contention by the participants is not surprising as there has been significant consideration of this issue that AI will lead to mechanisation of professions and extensive replacement of the human workforce [
18,
24]. Despite these fears, evidence suggests that in healthcare, to date, no one has forfeited job as a result of AI implementation; rather, AI is serving an advancement role in almost all professions. This career insecurity perception has been reported in other studies [
16‐
18,
24,
25]. However, AI will transform radiology, but not remove radiologists [
24] and for that matter radiographers whose work includes direct patient interaction [
25]. A few of the participants were of the view that the whole idea of AI technology should be re-examined and that there is the urgent need to raise awareness among radiographers through education, training and continuous professional development. Radiographers had previously described their willingness to interact/communicate with the patient. This indeed can be very rewarding in routine practice. Nevertheless, this seems to be somewhat contradictory with their willingness on specific training requirements.
The authors are in agreement with the findings of a publication [
27], which argued that though there are several questions to answer on AI, the technology has not come to eliminate health professionals but would rather replace those who fail to use the technology, and thus, all should be prepared for its total implementation. Although previous studies have raised several issues about AI, none suggested any uncertainties about the capability of AI to advance medical imaging practice. Herein, the impression is given that AI will only be in radiology; AI will be everywhere in medicine, admittedly probably earlier in so-called medical technical disciplines, i.e. radiology, nuclear medicine, pathology, laboratory and genetics. Even the entire electronic patient file will be almost continuously explored. Therefore, there is a need for African radiographers, imaging professional bodies, AI manufacturers/vendors, policy makers and all stakeholders including governments of various African countries to work at addressing end users’ concerns about AI in medical imaging for a successful implementation of AI in Africa.
Conclusions
The participants in this study were receptive of the idea to implement AI in medical imaging practice in Africa and agreed it is an indispensable tool which needs to be embraced. However, they did not hesitate to amplify some concerns including cost of the technology, impact on the workforce, ethical and legal regulatory concerns that might arise from data insecurity. The study found that radiographers in Africa would need more education on the technology and assurance of their job security. Although the qualitative responses could be predictable as with any new technology, people have various misgivings about it. In a resource poor setting, where uptake of new technologies could be a problem due to professionals being comfortable with traditional way of doing things, there is the need to assess concerns of the radiographers in particular, to address these anxieties before these implementations are in place.
The concerns of participants in the current study cannot be ignored as there are still arguments about AI algorithm performance. For this reason, this paper argues that AI would require a lot of attention from its technical performance to address the concerns of all clients. If end user concerns like issues of probable job losses, equipment maintenance culture, training and education of personnel are not addressed particularly in the African contest, AI could do more harm than good and have no positive impact on patients and the professionals who would use it. Therefore, there is a need for African radiographers, imaging professional bodies, AI manufacturers/vendors, policy makers and all stakeholders including governments of various African countries to work at addressing end users’ concerns about AI in medical imaging for a successful implementation of AI in Africa.
Acknowledgements
We would like to appreciate all the radiographers who took time to participate in this study. We also like to thank the following for helping in promoting the survey: Mr. Prince Rockson (Department of Medical Imaging, University of Health and Allied Sciences, Ghana); Mr. Stephen Samson Mkoloma (Ocean Road Cancer Institute, Tanzania); Mrs. Elizabeth Olasunkanmi Balogun (National Orthopaedic Hospital, Igbobi, Lagos, Nigeria); Dr. Wiam Elshami (Department of Medical Diagnostic Imaging, College of Health Sciences, University of Sharjah, United Arab Emirates); Mr. Joe Bwambale (Society of Radiography of Uganda); Mr. Caesar Barare (Kenyatta National Hospital, Kenya); Dr. Sibusiso Mdletshe (University of Auckland, Faculty of Medical and Health Sciences) and Samuel Arkoh (University of Ghana).
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.