Page 14 - AIH-1-4

Artificial Intelligence in Health Radiomics in early-stage lung cancer

in investigating immunotherapy response or predicting machines, ethical rules such as deontology and virtue found
relapse in oncological patients. 36-38 in humans should also be considered for using machines.
Delta radiomics has been used and found to be useful Our current moral systems are derived fundamentally
in evaluating the response to chemotherapies in colorectal from our responsibilities to other people. Therefore, a
cancer, liver metastases, and metastatic renal cell cancer. 39,40 non-human system, that is, AI, cannot be expected to
Delta radiomics has also been used to predict the risk of understand existing moral systems. These issues must be
42
developing radiation pneumonitis during treatment in resolved before using AI in human-related decisions.
patients with esophageal cancer. Delta radiomics features
41
obtained from PET-CT images have been used to estimate AI systems should be data stewards. Only the required
prognosis in patients with NSCLC. 32 data may be used and then deleted, which is also defined
as “data minimization.” Data should be encrypted and used
Changes in the tumor during treatment can be detected only by authorized individuals. Data must be collected,
using delta radiomics, which can thus help modify used, and shared according to privacy and personal data
treatment strategies. For instance, in a patient planned laws. Before using patient data, patients’ consent must be
for neoadjuvant radiotherapy, radioresistant tumors can obtained, and Ethics Committee permission should also be
be detected using delta radiomics, and the patient can be 43
referred to surgery earlier. When unresponsive patients for obtained from the concerned authorities.
lung SBRT are detected during this process, they can be 7. Conclusion
protected from unnecessary treatment toxicities. Hence,
standardizing the stages of obtaining delta radiomics can SBRT is considered the first treatment option with similar
both contribute to personalized treatments and protect oncological outcomes in patients with early-stage NSCLC
patients from unnecessary treatment toxicities. who cannot undergo surgery or refuse surgery for medical
reasons.
6. AI and ethical issues in cancer treatment
It is important to determine the patient’s risk of recurrence
Ethical issues surrounding AI in healthcare concern privacy, during the treatment planning stage to determine the most
bias, and discrimination, as well as whether it can replace ideal personalized treatment. If patients with a high risk
human judgment. Where there is technology, there is of recurrence can be selected in advance, the treatment
always the risk of inaccuracy and data breach. Moreover, intensity can be increased by changing the radiotherapy
wrong decisions can result in undesirable and devastating dose or schedule.
consequences in the treatment of patients with cancer.
There is no clear regulation on legal and ethical issues Acknowledgments
regarding the role played by AI in healthcare; therefore,
this issue needs careful consideration. None.
Innovations and new developments in technology Funding
accelerate scientific progress. It is important to explore
strategies to eliminate the potentially disastrous problems None.
of AI technologies. Conflict of interest
Machine learning algorithms are effective in identifying
and analyzing or classifying large amounts of data, referred The author declares having no competing interests.
to as “Big Data.” Big Data are used to train algorithms. Author contributions
Using more data to train an algorithm generally increases
the accuracy rate of the algorithm. The machine requires This is a single-authored article.
a set of rules and instructions, generally written in the
form of algorithms, to perform tasks. Nevertheless, when Ethics approval and consent to participate
newly acquired data are used, the machine gradually gains Not applicable.
the ability to become more flexible and operate in different
situations accordingly. This situation may increase the Consent for publication
demand for data and sometimes cause sharing of personal
or public information without considering user privacy. Not applicable.
Ethics and moral values vary across countries and even Availability of data
regions within countries. Ethnic groups and nations have
different norms. When defining values from humans to Not applicable.

Volume 1 Issue 4 (2024) 8 doi: 10.36922/aih.3541