Artificial intelligence (AI) is rapidly transforming healthcare with far-reaching potential to reshape disease diagnosis, laboratory testing, treatment selection and patient education and management 1, 2. Within AI, machine learning (ML), deep learning (DL) and large language models (LLMs) have distinct potential applications in transfusion medicine (TM), spanning the entire transfusion continuum from donor recruitment and health, through blood collection and blood bank operations, to transfusion practices and outcomes in recipients, and affecting healthcare professionals, educators, learners and researchers 2. These include marketing for blood donation, developing educational material, donor communication via chatbots, scheduling donation appointments with pre-donation eligibility screening and predicting donor return for targeted outreach initiatives 3, 4. AI can support blood bank processes, such as machine- and deep-learning-based image analysis for component quality control and the evaluation of red cell storage lesions through morphological changes 2, 4, 5. Additionally, AI can be used for inventory management, demand forecasting and optimization, transfusion threshold modelling to support transfusion decisions and for predicting transfusion reactions 2, 5. AI integration in blood bank operations can streamline their operations by automating repetitive tasks and routine processes. AI can also support transfusion education by assisting in the development of curricula, creating teaching aids and educational materials and designing assessment questions 2, 6. It is also used in medical research for the analysis of complex, large-scale data and for predictive modelling 5-7. As elsewhere in healthcare, AI can facilitate communication and dissemination of novel findings and best practices in TM, particularly for investigators and professionals less at ease with scientific writing or English as a foreign language, supporting more equitable representation and input across a diverse global community. The introduction of publicly available open-access LLMs in 2022 has generated significant interest in AI and is initiating transformative changes within the field of TM. However, concerns have arisen regarding the accuracy, safety and reliability of LLM-generated health information 2, 6. Furthermore, patient access to LLM-generated information may enhance patients' understanding while also influencing how physicians counsel patients and shaping the overall patient–doctor interaction. Additional issues include those potential consequences of over-reliance on AI tools threatening clinical and interpretative skills, potential bias from LLMs model training leading to underrepresentation of certain populations and adaptability to the local regulatory and medical ethical frameworks 2, 6. Other challenges include addressing ethical, governance, privacy, confidentiality and medico-legal concerns among healthcare professionals 2, 6, as well as content plagiarism and copyright issues that remain significant in the research field 2. This themed issue of Vox Sanguinis features a wide range of papers covering different AI-related topics, including an overview on AI and its current applications in healthcare, its future directions and its transformative potential in TM. It addresses the digital transformation of TM and leveraging of big ‘vein-to-vein’ data as roadmaps for AI integration in clinical application and medical research, while also examining existing challenges such as variability in blood bank organizational structures, data biases, costs and the need for substantial investments in technology and training 5, 7. This issue also contains studies assessing the current status of adoption of AI technologies among TM professionals as well as the facilitators and barriers to its use 2, 6, 8. Studies evaluating different applications of AI in donor management 3, 4, blood component quality assessment 5, 9, 10, or to guide decisions on blood product manufacturing optimization and personalized allocation of blood products are included as well. Finally, studies evaluating LLMs and AI chatbots in TM education are included 11-13, alongside discussions of ethical 14, legal and governance considerations, as well as educational and training needs and resources that need to be addressed for broader adoption of AI 2, 6, 8. Wider adoption of AI requires the development of clinical and technical expertise, formal structured AI-specific education and training 6, 8 and addressing barriers and concerns regarding the utilization of AI tools 8. LLMs and chatbots should be seen as assistive tools, not authoritative sources, with expert review needed before use in patient care or teaching. Accurate information from AI-powered LLMs requires expert evaluation and validation, with non-biased, inclusive and representative training datasets, verification and continuous performance monitoring to ensure that they generate accurate results while protecting personal healthcare data. Wider adoption also requires access to AI professionals and tools 6, 8 and digitalized infrastructure adjusted to different settings 6. Building big data infrastructure is important considering that ML models rely heavily on the availability of high-quality data 5, 7. Collaboration with technology vendors and providers can help, not only to deploy AI technologies to further advance the TM field but also to overcome financial constraints and restricted resources that can limit adoption. Cybersecurity risks must be addressed as well. Integrating AI and bioinformatics-focused education in undergraduate and postgraduate medical education is required to better prepare future professionals to safely use and oversee AI application in TM 2, 8. Moreover, it requires technical literacy and big data infrastructure. Establishing standards, policies, regulatory guidance and ethical frameworks is essential to guide healthcare organizations and professionals for safe, responsible, equitable and ethical use of AI tools in TM 2, 8. Moreover, ongoing research is recommended to evaluate the quality and accuracy of data generated by continuously evolving LLMs for medical education in different languages 11-13, as well as to develop region-specific legal and ethical frameworks for model training 11. To conclude, this themed issue of Vox Sanguinis reviews current and potential applications of AI in transfusion education, research and practice. It urges scientific committees, policy makers and industry partners to overcome existing challenges to leverage wider adoption of AI in the field of TM. It also emphasizes the need to translate AI applications into robust medical tools that can be adopted in daily transfusion practice. Given the rapid progress in this area, we urge scientists to prioritize research on AI applications in TM to foster future advancement in the field for professionals, educators and researchers. We are grateful to all contributors for sharing their expertise, insights and commitment to the development of AI in transfusion medicine. The authors declare no conflicts of interest.
Al‐Riyami et al. (Mon,) studied this question.