Page 80 - Petelin, Ana. 2024. Ed. Zdravje delovno aktivnih in starejših odraslih | Health of the Working-Age and Older Adults. Zbornik prispevkov z recenzijo | Proceedings. Koper: University of Primorska Press

Integrating Artificial Intelligence Into Healthcare Delivery
– A Vision for Service Transformation

How AI Works
Artificial Intelligence (AI) in healthcare does not work by itself, but on the ba-
sis of an ever-growing volume of data. They can only be addressed by AI-based
processing (Laukka et al., 2022). Given this data, AI-enabled devices are mainly
divided into two categories (Jiang et al., 2017). The first includes machine learn-
ing (ML) techniques that analyse structured data: imaging, genetic and electro-
physiological data. In healthcare applications, ML procedures attempt to clus-
ter patient characteristics or infer the likelihood of disease outcomes (Darcy et
al., 2016; Cato et al., 2020). The second category includes natural language pro-
cessing methods that extract information from unstructured data, i.e., clinical
records, peer-reviewed and scientific medical journals, to complement struc-
80 tured medical data. Natural language processing techniques attempt to convert
text into machine-readable structured data that can then be analysed using AI
zdravje delovno aktivnih in starejših odraslih | health of working-age and older adults
machine learning techniques (Jiang et al., 2017).
There are three types of AI techniques that are used in medical applica-
tions. The classical machine learning (ML) technique builds algorithms to an-
alyse data that include patient “characteristics”, e.g. age, gender, clinical symp-
toms, etc., sometimes including health outcomes as disease indicators, patient
survival times and quantitative disease levels, e.g. tumour size. More recent
deep learning techniques address multi-layer neural networks; they are used
for speech recognition, image recognition, text understanding, etc. Natural
language processing (NLP) methods help to convert unstructured narrative
text into structured, machine-readable text to enable information extraction
(Jiang et al., 2017; He et al., 2019). Depending on the desire to incorporate re-
sults, ML algorithms can be divided into two categories, namely supervised
and unsupervised learning (Jiang et al., 2017). AI also raises concern, mainly
due to its ability to modify its behaviour and act autonomously with minimal
human input; however, it is currently still unable to operate without human su-
pervision (Zlatanova and Veljković, 2023), and most importantly, it lacks a sub-
jective essence (Ferrarelli, 2023).

Promoting and Maintaining Health, Prevention
The ageing population, the need for specialised care and more care pro-
grammes, the emergence of chronic diseases and lifestyle-related health prob-
lems, the demands for active involvement of patients in their treatment and
more convenient care options, and value-added services dictate the need for
new strategies in healthcare: prevention, diagnosis and optimal care (Gopal
et al., 2019). AI could be used to identify individuals or groups predisposed to
certain diseases and provide them with timely and effective treatment. For in-
stance, in the elderly, AI could be used to identify cognitive decline due to neu-