Page 55 - AIH-2-3

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

AI is also being deployed in developing nations, where were hesitant about data sharing for AI development. The
healthcare resources are limited. With the increasing scarcity of real, accessible patient data limits the potential
availability of computers and internet access, AI-driven of AI in healthcare, with fears exacerbated by the lack of
diagnostic tools are providing life-saving services to people regulations governing AI usage in countries like the U.S.
in areas with limited access to healthcare professionals. Concerns about data being misused for financial gain,
This helps to reduce outsourcing and improve the quality as exemplified by Roche’s purchase of healthcare data for 2
of care. AI systems in these regions are tailored to offer million cancer patients, question whether patient data can
individualized treatments, adjusting based on real-time or should have a monetary value, raising broader ethical
data, thus improving patient outcomes in resource- debates around fairness and patient consent. Automation
constrained settings. in healthcare also stirs controversy. While AI has yet to
However, the widespread adoption of AI in healthcare replace healthcare jobs, research shows that automation
brings with it several regulatory and ethical concerns. might affect roles that handle digital information, such as
The risks associated with AI, such as algorithmic bias, radiology and pathology. A 2019 study in the UK estimated
patient data privacy, and the implications of machine that AI could replace up to 35% of jobs within two decades,
morality, necessitate stringent regulations. There are though doctor-patient interactions are less likely to be
53
established reporting guidelines, such as TRIPOD-AI, impacted. On the positive side, AI offers the potential to
DECIDE-AI, and CONSORT-AI, that ensure AI studies enhance healthcare by alleviating burnout and cognitive
are transparently reported for regulatory approval. In the overload for medical professionals, allowing them to
U.S., HIPAA safeguards patient data, while GDPR protects focus more on patient care. AI is even being applied in
patient data in the European Union. elder care, where robots assist with entertainment and
companionship, allowing caregivers to provide more one-
The U.S. FDA has developed a plan to regulate on-one care. Despite these benefits, skepticism remains
AI-based medical devices, focusing on areas such as good about whether AI can offer the empathy provided by
ML practices and algorithm bias. Ethical concerns in AI human healthcare professionals, as found in a 2023
include the balance between patient autonomy and the thematic review. 55,61,68 Bias in AI systems is another critical
use of AI in making critical healthcare decisions. The U.S. concern. AI’s reliance on input data means that if the data
Department of Health and Human Services has issued are biased or unrepresentative, it can lead to discriminatory
guidelines emphasizing the ethical principles of autonomy, outcomes.
beneficence, non-maleficence, and justice. Similarly, the
GDPR in Europe protects citizens’ data, ensuring fairness, Medical AI systems can unintentionally perpetuate
transparency, and respect for human dignity in AI-driven social and healthcare inequities by making more accurate
processes. As AI continues to shape healthcare, balancing predictions for majority populations, such as White males,
innovation with patient rights and ethical standards will who are overrepresented in medical datasets. This can
remain a key challenge for regulators globally. result in worse outcomes for minorities. Collecting data
from minority communities, while essential for balanced
Finally, international efforts to standardize AI algorithms, can also lead to medical discrimination,
use in healthcare are underway. The joint World such as the potential misuse of data related to conditions
Health Organization (WHO) and the International like acquired immunodeficiency syndrome. Beyond
Telecommunication Union (ITU) (ITU-WHO) Focus demographic biases, differences in clinical systems and
Group on AI for Health (FG-AI H) has been benchmarking work practices also introduce variability in AI functionality.
4
AI applications in medical settings, including cancer However, many of these biases can be mitigated through
risk assessment and diagnosis from medical imagery. careful data collection and algorithm design. A specific
These initiatives reflect the growing importance of form of bias, “label choice bias,” arises when proxy measures
ensuring that AI in healthcare is safe, effective, and are used in algorithms, such as using healthcare costs to
ethical, while promoting innovations that improve global predict patient needs, which can skew results against
health outcomes. 7-75 The use of AI in healthcare presents certain groups, like Black patients. 68,76 Addressing these
significant ethical concerns, especially regarding data biases requires closer alignment between the target of the
collection, automation, and bias. One of the key issues is the AI prediction and the actual healthcare needs of patients.
massive amount of data required to train AI systems, often Historically, AI in healthcare has evolved since the 1960s
sourced from patients. This raises privacy concerns, as and 1970s, starting with expert systems such as Dendral
many individuals are uncomfortable with sharing personal and MYCIN, which laid the groundwork for future AI
health information for technological advancements. applications in medicine. 15-75 Although these early systems
A survey in the UK revealed that 63% of respondents did not achieve widespread clinical use, they highlighted

Volume 2 Issue 3 (2025) 49 doi: 10.36922/aih.5173