Vetnews | Oktober 2024 16 « BACK TO CONTENTS Discussion Following the current review, the data on applied AI in the veterinary sphere were assigned to seven categories, and the uncategorized citations were grouped under a diverse category. The minimum threshold criterion noted above aided in minimizing quality assessment errors. The diagnostic process is greatly influenced by the latest AI advancements, which lead to a shorter time to diagnosis and more confidence that an appropriate decision was made. Several practical settings benefit from AI, such as atrial fibrillation detection, seizures, hypoglycemia, and the diagnosis of several diseases [12]. Clinicians have acquired many AI applications to aid in diagnosing, supervising, and monitoring diseases successfully in daily practice. However, clinical precision is still questionable, and the appropriateness of a generating error was biased (Table 2). This is due to outcomes being derived from generating algorithms of pre-existing data. The possible occurrence of marginal error and limitations is linked to the design model of AI, which is also called the overfitting phenomenon [13, 14]. The reliability of AI applications has been discussed in different domains, and the behaviour of many professionals toward its outcomes is due to evidence of questionable efficiency compared to clinicians. Artificial intelligence can produce unreliable outcomes due to the lack of primary data replication and built algorithms overlapping from one case to another [12, 15]. Future studies should focus on an AI model that has flexible input data. This model could be expanded by clinicians to maximize their interaction without altering or outperforming algorithms and alleviating generated errors. Another key area is AI-derived applications that focus on revealing and confirming clinical opinions already initiated by a veterinarian. In this context, the clinicians’reluctance to embrace advances in AI may be problematic unless the technology supports the clinician’s opinion. Educational use represents 11.4% of total usage for administrative and linguistic support, data mining, learning processes, communication, and research. Epidemiological purposes (10.56%) can aid with risk factor assessment, forecasting and prevention, surveillance programs, and other assessments for envisioning potential future strategies. The challenges of AI have been addressed intensively in veterinary medicine, with 92% of the retrieved studies published in the past 3 years. However, AI outcomes should not be relied upon exclusively as they might be incomplete, heterogeneous, erroneous, or inaccurate. Bologna and Hayashi noted that the best-performing methods are often the least transparent, and those providing Artificial intelligence feasibility in veterinary medicine... <<<15 Other components of the standardized table (study length, population, limitations, and statistical approach) show zero to very low evidence of bias, meaning that the outcomes of this study are trustworthy (3rd interval, acceptable to high). Only three items from the quality assessment were rated as biased: The definition of AI essence, the data source, and the appropriateness of error generation. A score of three indicates a likelihood that data on applied AI was biased, and the ultimate outcome appears to have a low to moderate risk of bias (second interval). The quality assessment of the included studies varied in their relative quality and risk of bias. The findings of the six items that scored six out of a total of nine points on the quality scale are more robust regarding the accuracy of AI outperformance and the extent of AI in the veterinary profession. Table-3: Categorization of obtained studies (using AI in veterinary medicine) from databases PubMed, Embase, Google Scholar, and Scopus up to March 22, 2023. Study type Number of articles Peer-reviews studies 579 Conferences reports 89 Journals/books papers 215 Figure-2: Artificial intelligence allotment percentage in the veterinary sector as per databases (PubMed, Embase, Google Scholar, and Scopus) screening up to March 22, 2023.
RkJQdWJsaXNoZXIy OTc5MDU=