George Kourounis

Advances in AI hold considerable promise for organ transplantation. While every transformation brings change, not all change is transformative. Despite the rapid growth of AI in medicine, most applications remain in developmental or experimental stages, with relatively few having been successfully integrated into routine clinical practice. As a professional society, ESOT recognises that achieving meaningful impact will require more than technical progress. This position paper outlines five critical domains for successful implementation. (1) High-quality development: Coordinated collaboration and methodological rigour are prerequisites for trust; AI is only as robust as the data used to train it. (2) Ethical considerations: We must address risks to equity and access to care, and move from generic ethical principles to transplantation-specific ethical guidance. (3) Regulatory landscape: AI in transplantation is regulated under both EU medical device and AI legislation; compliance is central to stakeholder trust. (4) Responsible adoption: AI should augment, not replace, human expertise. Strengthening AI literacy is essential for meaningful adoption. (5) Participatory design: Active involvement of transplant professionals and patients is essential to address real clinical needs. These statements serve as a strategic framework to guide clinicians, researchers, and policymakers in making AI a genuine force multiplier for the transplant community.

Artificial intelligence (AI) has the potential to expand access to high-quality surgical care in low-income and middle-income countries (LMICs), yet the extent and maturity of AI research in these settings remain unclear. We conducted a prospectively registered scoping review (osf.io/9PV6A) to synthesize primary evidence on the use of AI in LMIC surgical care. PubMed, Scopus, and Web of Science were searched for studies evaluating AI in surgical contexts within LMICs up to July 14, 2025. From 2602 records, 475 studies met inclusion criteria. Most were conducted in upper-middle-income countries (n = 376, 79·1%), with the overwhelming majority from China (n = 305, 64·2%). Only 46 studies (9·7%) were conducted in lower-middle-income countries and 5 (1·1%) in low-income countries. Research was predominantly retrospective (68%), and only nine randomised controlled trials were identified (2%). Most studies focused on model development (67%), with few reporting external validation (30%) or clinical deployment (3%), mostly as pilot trial-based integrations. Barriers to AI implementation included fragmented data systems, limited infrastructure, and workforce constraints. Facilitators included widespread smartphone access and growing international collaborations. Despite rapid growth, AI research remains in the early stages of development. Focus on model accuracy alone is insufficient if health systems lack the capacity for adoption and integration.

Introduction: Accurate evaluation of liver steatosis and overall organ quality is critical for optimizing safe organ utilization in liver transplantation. Recent advances in computer vision offer promising tools to standardize and enhance this process. This review maps the current evidence on AI-enabled photographic evaluation of liver steatosis and identifies areas for future development. Methods: A scoping review of the literature, including searches of PubMed, SCOPUS, and Web of Science, was conducted to identify studies published from inception to 27/03/2025 reporting on the development of AI-enabled tools for assessing liver organ quality from photographs taken during the donation process. A qualitative synthesis and critical review of the literature was conducted in accordance with PRISMA-ScR guidelines. The review protocol was registered with the Open Science Framework (osf.io/zfcuk). Results: After screening 219 citations, six studies from three independent research groups met the inclusion criteria. Sample sizes ranged from 40 to 192 donors. Five studies employed binary classification models using a 30% steatosis threshold, while one study reported a graded approach. Reported accuracies ranged from 0.81 to 0.92. Common challenges included small and imbalanced datasets with a dependence on supplementary donor data, such as blood tests and radiological findings. None of the studies conducted external validation. Discussion: Current evidence is drawn from a small and methodologically heterogeneous literature. Publications from several independent groups nevertheless highlight growing interest in developing these tools. Future work should prioritize larger studies with robust external validation to strengthen their credibility and build trust in their clinical use.

Regression models provide information on complex relationships between patient factors, investigations, diagnoses, treatments, and outcomes. These inferences underpin evidence-based medicine. However, by default regression models assume straight-line relationships, and the common approach of splitting continuous variables into groups has several disadvantages. We discuss pitfalls with these common approaches, and provide an interactive regression model playground, which acts as a point-and-click showcase of these concepts. More flexible regression modelling techniques are available, which allow non-linear relationships between predictors and outcome to be captured. However, they have been shown to be underused in medical research. We feel a major contributor to this is that more flexible non-linear models are typically explained for a statistical audience, creating a barrier for medical professionals. In this article, we introduce non-linear regression for medical researchers. Specifically, we focus on restricted cubic splines (RCS), which allow curved relationships to be fit, within the context of regression models. This has the benefit that the overall structure of the regression model and its outputs, which are familiar to medical researchers, stays the same, with the simple addition of non-linear modelling of specific variables. We implement RCS in a case study, with accompanying example R scripts (available on GitHub). We also launch an R package ("rmsMD") which aims to make this technique approachable to medical researchers, as well as creating publication-ready tables and plots. Overall, this article equips medical researchers with an intuitive understanding of non-linear modelling, which can then be applied with the easy-to-use tools provided.

Objectives: Medical photography is ubiquitous and plays an increasingly important role in the fields of medicine and surgery. Any assessment of these photographs by computer vision algorithms requires first that the area of interest can accurately be delineated from the background. We aimed to develop deep learning segmentation models for kidney and liver organ donation photographs where accurate automated segmentation has not yet been described. Methods: Two novel deep learning models (Detectron2 and YoloV8) were developed using transfer learning and compared against existing tools for background removal (macBGRemoval, remBGisnet, remBGu2net). Anonymised photograph datasets comprised training/internal validation sets (821 kidney and 400 liver images) and external validation sets (203 kidney and 208 liver images). Each image had two segmentation labels: whole organ and clear view (parenchyma only). Intersection over Union (IoU) was the primary outcome, as the recommended metric for assessing segmentation performance. Results: In whole kidney segmentation, Detectron2 and YoloV8 outperformed other models with internal validation IoU of 0.93 and 0.94, and external validation IoU of 0.92 and 0.94, respectively. Other methods – macBGRemoval, remBGisnet and remBGu2net – scored lower, with highest internal validation IoU at 0.54 and external validation at 0.59. Similar results were observed in liver segmentation, where Detectron2 and YoloV8 both showed internal validation IoU of 0.97 and external validation of 0.92 and 0.91, respectively. The other models showed a maximum internal validation and external validation IoU of 0.89 and 0.59 respectively. All image segmentation tasks with Detectron2 and YoloV8 completed within 0.13–1.5 s per image. Conclusions: Accurate, rapid and automated image segmentation in the context of surgical photography is possible with open-source deep-learning software. These outperform existing methods and could impact the field of surgery, enabling similar advancements seen in other areas of medical computer vision.

Artificial intelligence tools, particularly convolutional neural networks (CNNs), are transforming healthcare by enhancing predictive, diagnostic, and decision-making capabilities. This review provides an accessible and practical explanation of CNNs for clinicians and highlights their relevance in medical image analysis. CNNs have shown themselves to be exceptionally useful in computer vision, a field that enables machines to ‘see’ and interpret visual data. Understanding how these models work can help clinicians leverage their full potential, especially as artificial intelligence continues to evolve and integrate into healthcare. CNNs have already demonstrated their efficacy in diverse medical fields, including radiology, histopathology, and medical photography. In radiology, CNNs have been used to automate the assessment of conditions such as pneumonia, pulmonary embolism, and rectal cancer. In histopathology, CNNs have been used to assess and classify colorectal polyps, gastric epithelial tumours, as well as assist in the assessment of multiple malignancies. In medical photography, CNNs have been used to assess retinal diseases and skin conditions, and to detect gastric and colorectal polyps during endoscopic procedures. In surgical laparoscopy, they may provide intraoperative assistance to surgeons, helping interpret surgical anatomy and demonstrate safe dissection zones. The integration of CNNs into medical image analysis promises to enhance diagnostic accuracy, streamline workflow efficiency, and expand access to expert-level image analysis, contributing to the ultimate goal of delivering further improvements in patient and healthcare outcomes.