Page 56 - AIH-2-3

Artificial Intelligence in Health AI in medical diagnostics: A multi-disease approach

the potential for AI in healthcare. In the following decades, functioning correctly by analyzing discrepancies between
advances in computing power, genomics, and EHRs enabled expected and observed behavior. Diagnosability, a key
the expansion of AI’s role in healthcare. Breakthroughs in concept, refers to the ability of the system to provide an
NLP, computer vision, and ML have allowed machines unambiguous diagnosis. This is particularly crucial during
to replicate human-like decision-making and perceptual system design, where a balance must be struck between
processes. AI has contributed to innovations such as robot- reducing sensor costs and increasing the ability to detect
assisted surgery, rare disease prediction through DL, and faulty behaviors. Algorithms have been developed to
more precise health prediction. Despite these advances, ensure diagnosability by either confirming whether a
the ethical challenges surrounding data, automation, and system is diagnosable or identifying the necessary set
bias remain central to discussions about AI’s future in of sensors to make a system diagnosable. Diagnosis in
healthcare. AI deals with detecting malfunctions in systems and
identifying their causes through expert systems or model-
4. Diagnosis in AI based approaches. These techniques rely on observations
Diagnosis, as a subfield of AI, is focused on creating and simulations to provide accurate diagnoses, but they
algorithms that can assess whether a system is functioning come with challenges such as expertise acquisition, system
properly. If a malfunction is detected, these algorithms complexity, and diagnosability. Figure 1 offers further
are responsible for accurately identifying the faulty insights into this issue.
component and the nature of the fault. This process is
based on observations, which provide insights into the 5. Improving medical diagnosis through AI
system’s current state. The term “diagnosis” originates from AI is reshaping medical diagnostics by delivering remarkable
the medical field, where it refers to identifying diseases advancements in accuracy, speed, and the personalization
based on symptoms, but in AI, it broadly encompasses of patient care. Through sophisticated ML and DL
both the detection of faults and the process of determining models, AI enables the processing of extensive datasets,
if a system is malfunctioning. the analysis of complex medical images, the prediction

An everyday example of diagnosis can be illustrated of disease progression, and significant enhancements in
with a car mechanic troubleshooting a vehicle. The diagnostic precision. This transformation is especially
mechanic begins by observing the car’s behavior and prominent in specialized fields such as radiology, wound
applying their knowledge of the vehicle type. If a problem and burn management, and diabetic care, where AI-driven
is detected, further tests and observations are conducted to innovations have made a substantial impact on improving
refine the diagnosis until the faulty part is discovered. In patient outcomes. 18-22
AI, expert diagnosis systems operate similarly by mapping 5.1. Vital contributions of AI in medical diagnostics
observations to diagnoses based on prior experience.
Enhanced diagnostic accuracy. AI systems, particularly
This expertise may be derived from human operators,
who encode their knowledge into a computer-readable within radiology, have shown superior performance in
pattern recognition, often identifying early signs of disease
format, or from examples of system behavior classified as that can be overlooked by the human eye. For instance,
either correct or faulty. ML techniques can then generalize to AI has demonstrated higher accuracy in detecting breast
metadata in terms of DL from these examples. Multimodal cancer from mammograms, offering critical insights that
models can be used for further exploration for finding can lead to earlier intervention and better survival rates. 18
new features and functionality. However, expert diagnosis
faces challenges, such as difficulty in acquiring sufficient Facilitating early detection. By enabling the early
expertise, especially in critical systems, the complexity of diagnosis of life-threatening conditions such as cancer,
the learning process, and potential limitations in storage cardiovascular diseases, and neurological disorders,
and robustness. AI-driven tools facilitate prompt treatment planning. Early
A more structured approach to diagnosis is model- detection through AI tools has been linked to improved
based diagnosis, which employs a model of the system to patient outcomes and reduced mortality rates, particularly
19
simulate its behavior. By comparing actual observations in cases where time-sensitive interventions are essential.
with predicted outcomes from the model, faults can be Advancements in personalized medicine. AI algorithms
identified. In this form of abductive reasoning, the model aid in tailoring treatments to individual patient profiles
may describe normal system behavior but often lacks a by analyzing personal health records, genetics, and
detailed representation of faulty behavior. The diagnostic lifestyle factors. This personalized approach supports
system uses this model to determine if the system is the development of more effective treatment plans,

Volume 2 Issue 3 (2025) 50 doi: 10.36922/aih.5173