Page 40 - Read Online

Afyouni et al. Hepatoma Res 2023;9:28 https://dx.doi.org/10.20517/2394-5079.2023.29 Page 9 of 14

exhibit persistent enhancement early after treatment, which can last up to and beyond 1 year and should not
[94]
be confused with a viable tumor . Other imaging characteristics may also provide information in the
assessment of tumor response, particularly T2 signal intensity and ADC at MRI. Typically, tumors
demonstrate a decrease in T2 signal intensity and DWI, and an increase in ADC after treatment, suggestive
of tumor necrosis .
[95]

FUTURE DIRECTIONS AND CHALLENGES
Integrating AI with radiomics and genomics for personalized medicine
Integrating artificial intelligence (AI) with radiomics and genomics has the potential to revolutionize the
field of personalized medicine in ICC management. By combining quantitative imaging features (radiomics)
with molecular and genetic information (genomics), AI algorithms can potentially identify unique tumor
signatures, predict treatment response, and estimate patient prognosis more accurately. This integration will
facilitate the development of tailored therapeutic strategies based on individual tumor characteristics and
genetic makeup, ultimately improving patient outcomes .
[96]

Development of AI-based prognostic models
AI-based prognostic models can help clinicians estimate patient survival and disease progression more
accurately, enabling better-informed decisions regarding treatment and follow-up. By analyzing a wide
range of data, including imaging, clinical, and molecular information, AI algorithms can identify subtle
patterns and interactions that may be overlooked by traditional prognostic models .
[97]
Developing and validating AI-based prognostic models for ICC will require extensive research and access to
large, diverse, and well-annotated datasets, requiring collaboration between institutions and researchers to
share and standardize data. Additionally, the integration of these models into clinical practice will
necessitate a rigorous evaluation of their performance, interpretability, and generalizability, as well as the
development of user-friendly interfaces for clinical use. Moreover, an interdisciplinary collaboration among
radiologists, oncologists, geneticists, and computer scientists will be crucial to advance the field and
[98]
translate AI-based tools into clinical practice .
Ethical considerations and data security in AI applications
The implementation of AI in ICC diagnosis and treatment raises several ethical considerations and data
security concerns. Ensuring patient privacy and the secure handling of sensitive medical data is
[98]
paramount .
Another ethical concern is the potential for biases in AI algorithms, which may be introduced by the
training data or the model design. Biased algorithms can lead to unfair or discriminatory treatment
decisions, negatively impacting patient care. It is essential to continuously evaluate and refine AI models to
minimize potential biases and ensure equitable and accurate predictions for all patient populations.

Finally, the widespread adoption of AI in clinical practice requires addressing issues related to
accountability and liability in the case of AI-generated errors or misdiagnoses. Developing guidelines and
legal frameworks that clarify the responsibilities of various stakeholders, including AI developers, healthcare
providers, and patients, will be critical to address these concerns.

CONCLUSION
Radiology has a pivotal role in the management of intrahepatic cholangiocarcinoma (ICC), providing
valuable information about tumor characteristics, staging, and treatment response. The integration of