Page 159 - Read Online

Page 226 Grewal et al. Art Int Surg 2023;3:217-32 https://dx.doi.org/10.20517/ais.2023.28

For diagnostics, current diagnostic criteria consist of varying combinations of pathology- and radiology-
based input to reach a definitive diagnosis. Both these methods, however, are limited by their current
accuracy in differentiating lesions and human error in the workflow. Radiomics has already demonstrated
the potential to accurately diagnose malignancies through their unique radiomic footprints. This is of
particular importance in pancreatic lesions where there is a significant overlap in radiological features
between these lesions. Additionally, radiomics has demonstrated the ability to noninvasively predict tumor
characteristics with high accuracy, including tumor grade and the presence of nodal disease. In particular,
for tumors that demonstrated intertumoral heterogeneity, such as pancreatic neuroendocrine tumors,
pathological assessment is limited to the tissue obtained on biopsies and might fail to provide information
about the whole lesion [134,135] . Radiomics allows for “virtual biopsies” that can overcome this by capturing the
spatial heterogeneity of the entire tumor, enabling optimal characterization. Furthermore, the noninvasive
nature of radiomics can result in repeat assessment at varying intervals for patients who are on surveillance.
Another interesting application in tumor characterization is the emerging field of radiogenomics. Albeit
limited, the literature available on this has suggested that radiomics can be used to genetically characterize
these tumors. As this field develops, radiomics has the potential to allow for accurate diagnostics and tumor
characterization that is superior to the current diagnostic modalities.

Radiomics has also demonstrated potential in predicting response to systemic therapies noninvasively.
Current biomarkers fail to accurately predict response to administered therapies, resulting in delays in
tailoring of therapies in a timely manner. If radiomics-based assessment of treatment response were
possible, it would allow for timely modification in systemic therapy, thus maximizing the chance for
complete or near complete treatment response and improved survival. In terms of surgical planning,
radiomics has demonstrated the ability to predict the presence of vascular invasion more accurately than
experienced radiologists. Lastly, radiomics has shown promise in prognostication via accurate prediction of
postoperative complications, and recurrence-free and overall survival. Accurate prognostication is essential
not only in guiding management, as previously discussed, but also in managing patient expectations
through the course of their care. Radiomics-based prognostication has outperformed a variety of existing
prognostic factors. In studies where radiomics has not independently outperformed existing factors, the
addition of radiomics to established factors enhanced the accuracy of the existing models. While it is
unlikely that radiomics models will ever independently be used to prognosticate patients, it is very likely
that models combining radiomic features and clinical factors will be used across all metrics of
prognostication in the future.

CONCLUSIONS
In conclusion, the field of radiomics is advancing rapidly and has shown promise as a tool for early
detection, tumor characterization, therapeutic selection, and prognostication in hepatobiliary and
pancreatic malignancies. Despite its shortcomings, we believe that with improvements and automation of
segmentation techniques, optimization of radiomic analyses, and introduction of standardized guidelines
for research on radiomics, the tools developed using this technology will become more robust. Clinical
application of these tools will provide precision in the management of these patients, resulting in
improvement in patient outcomes.

DECLARATIONS
Authors’ contributions
Made substantial contributions to the conception and design of the study and performed data analysis and
interpretation: Grewal M, Ahmed T, Javed AA