Introducing the <i>Veterinary Clinical Pathology</i> Special Issue on Clinical Bioinformatics Technology and Artificial Intelligence

Implementation of rapidly developing advanced technology has always been at the forefront of laboratory diagnostics and is integral to providing fast, reliable, cutting-edge data to our stakeholders. This issue on ‘Clinical Bioinformatics Technology and Artificial Intelligence’ demonstrates the potential advantages and challenges of implementing some of these advanced technologies in the arena of veterinary laboratory diagnostics. Harris et al. highlights the potential of next-generation sequencing (NGS) for genomic analysis in veterinary oncology using a case of acute myeloid leukemia (AML) as an example [1]. The authors point out similarities between human and canine mutations, suggesting dogs may act as models for drug testing and may benefit from better classification to improve outcomes. While NGS is regularly used in human medicine to type AML, the implementation of similar technology in veterinary medicine has been sporadic. The authors call attention to the challenges of cost, turnaround, lack of robust bioinformatics in veterinary species, and the need for quality control. MicroRNAs (miRNAs) have been suggested as biomarkers in a number of clinical conditions and in a variety of species. Chu and Nabity review the biogenesis and use of miRNAs as biomarkers of chronic kidney disease in dogs and cats [2]. In particular, urine is intriguing as a readily available, non-invasive source of miRNAs, but comes with both pre-analytical and analytical challenges. The authors discuss in detail the technological considerations and necessary quality control measures for the measurement of miRNA in kidney tissue and urine. Discussion of artificial intelligence (AI) models has been widespread in the scientific and lay literature, and the technology is rapidly evolving. In order for AI-driven diagnostic tests to be properly evaluated and implemented, understanding the terminology and proper study design, testing, and validation are key. To this end, three review papers appear in this issue. A good place to start is the review by Neal et al., which provides an introduction to key concepts in AI, including fundamental principles and definitions [3]. The reviews by Pacholec et al. are focused on the requirements for quality assurance in model development and validation [4, 5]. A further discussion on model optimization is provided by a letter to the editor from Castillo and a Response by Pacholec et al. [6] Several original studies evaluating AI models are included and emphasize some of the key points made by the review articles. The potential for improved laboratory efficiency and workflow is demonstrated in the study by Hayes et al., in which a combination of expert-derived criteria and machine learning could be used to optimize the selection of hematology smears for manual review in a busy laboratory [7]. The pilot study by Pacholec et al. evaluates a convolutional neural network (CNN) image analysis model for differentiating canine large cell lymphoma from non-lymphoma cases [8]. This manuscript points out the need to evaluate the size and quality of the training sets. Le Boedec uses CNN to visually categorize histograms as Gaussian, lognormal, or left-skewed, which may be useful for selecting the most appropriate statistical evaluation for generating reference intervals for small populations, a common issue in veterinary medicine [9]. Overall, the papers in this issue highlight the exciting potential that these rapidly advancing technologies can bring to veterinary laboratory diagnostic testing. As a cautionary note, they all emphasize the need for proper test design, evaluation, and quality control to ensure accurate and repeatable results. AI, in particular, diverges the most from our routine laboratory tests, requiring knowledge of new vocabulary and methodology. But like any other laboratory methodologies, rigorous validation accompanied by a human pathologist/expert oversight is essential prior to the clinical use of AI-based testing. Incorporating AI literacy into our resident training programs will be necessary [10]. Other organizations that also provide diagnostic testing, such as the American College of Veterinary Radiology and European College of Veterinary Diagnostic Imaging, have begun to generate guidelines for validation, ethical use, and transparency of AI as a diagnostic tool [11]. The guidelines at Veterinary Clinical Pathology have also been updated to include guidance for authors using AI (https://onlinelibrary.wiley.com/page/journal/1939165x/homepage/forauthors.html#ai). These guidelines should be followed when preparing all articles for publication. The authors declare no conflicts of interest.