When the humble stethoscope was invented in the early 19th century, it represented a massive step forward in medicine. For centuries, physicians could only listen to a patient’s lungs by placing their ear close to the patient’s chest. Needless to say, this method was not very effective. The stethoscope opened up a new world of sound and medical possibilities, leaving an indelible mark on respiratory disease diagnostics.
Today, a new innovation in medicine holds great promise for revolutionizing medical diagnostics once again. It is a technology that enables the extraction of a large number of quantitative and qualitative features from medical images: radiomics.
Assessing Tumor Heterogeneity
In Infectious Agents and Cancer, Granata and colleagues wrote, “Radiomics is an emerging field that extracts and analyses data from medical images, comprising quantitative and qualitative features not detected by [the] human eye.” It consists of several components:
- Image acquisition
- Segmentation of the volume of interest by manual, automatic, or semiautomatic tools
- Features creation
- Database improvement
- Database analysis via the construction of predictive models
- Validation of the models created.
Radiomics allows us to capture subtle tissues and lesion changes over time. This holds great value in cancer studies.
“In oncology, assessment of tissue heterogeneity is of particular interest: genomic analyses have demonstrated that the degree of tumor heterogeneity is a prognostic determinant of survival and an obstacle to cancer control,” Mayerhoefer and colleagues wrote in The Journal of Nuclear Medicine.
Radiomics thus complements biopsies in that it allows us to capture heterogeneity across the entire tumor volume, not just at a single anatomical site. Radiomics can therefore be extrapolated to predict clinical endpoints, such as treatment response and survival.
Another important feature of radiomics is that it produces mineable data. This means that as data is accumulated over time, it creates a picture that tells a story about disease evolution, progression, and response to treatment. If the data mined from radiomics can be fed into artificial intelligence systems and adapted for clinical practice, we may be at the precipice of an explosion of clinical information triggering further medical innovations and treatments.
Detecting and Evaluating Liver Cancer
Given that the body of research using radiomics to diagnose and track cancer has increased over the years, let us look at the role of radiomics in assessing liver cancers, particularly cholangiocarcinoma.
Granata and colleagues wrote a narrative review on how radiomics techniques are being used to detect primary liver cancers. According to them, radiomics data obtained via magnetic resonance imaging (MRI) are a rich source of information that combines the advantages of high-contrast resolution and functional data. This helps physicians better characterize liver lesions and evaluate various treatment options.
Read more about cholangiocarcinoma etiology
In intrahepatic cholangiocarcinoma (iCCA), radiomics is mainly used to assess the risk of recurrence after surgical resection. The possibility of cancer recurrence after surgical resection is a notable source of anxiety for many patients, as it might signal the need for more aggressive therapeutic measures.
As described in Liver International, Qin and colleagues built a multilevel model and quantified the risk of early recurrence of perihilar cholangiocarcinoma after curative resection. They made use of 7 independent factors to build their model: tumor differentiation, lymph node metastasis, preoperative CA 19-9 level, enhancement pattern, A-Shrink score, V-Shrink score, and P-Shrink score.
They discovered that their model, which was based on radiomics principles, performed better than rival models and conventional staging systems. The accuracy of the model in predicting early recurrence in patients who had undergone curative resection was 0.826, while the accuracy of other models ranged from 0.581 to 0.641.
Read more about cholangiocarcinoma patient education
Studies by other researchers have validated the role that genomics can play in medical diagnostics. For example, another team of researchers managed to develop a noninvasive computed tomography (CT)-based radiomics model with the intent of predicting early recurrence in 177 iCCA patients. The researchers were able to do so by utilizing 6 established radiomics models considered stable and robust. Many cancer research institutions are now experimenting with predictive models created based on radiomics principles.
Limitations and Future Actions
It is clear that radiomics has the potential to greatly increase our ability to care for cancer patients. However, it does suffer from a few limitations, chief of which is the lack of standardization of findings. If we are to move beyond impressive models and figures on a screen, radiomics needs to be adapted strictly with clinical practice and clinical outcomes in mind.
In Insights into Imaging, van Timmeren and colleagues summarized the main limitations of radiomics this way: “The interpretability of the features, mistakes in the interpretation of the results, or the lack of comparison with well-established prognostic and predictive factors, results in reservation towards its use in clinical decision support systems.”
As physicians, we understand the ruthless truth that anything too detached from day-to-day clinical practice is likely to be impractical and expensive. Hence, the challenge of radiomics in the next decade will be to prove itself as truly useful clinically, in a way that is easy to access, easy to use, and relatively affordable.
As van Timmeren and colleagues wrote, “Using studies with large enough datasets and external validation and its combination with established methods will help move the field [of radiomics] towards clinical implementation.”
Van Timmeren JE, Cester D, Tanadini-Lang S, Alkadhi H, Baessler B. Radiomics in medical imaging-“how-to” guide and critical reflection. Insights Imaging. 2020;11(1):91. doi:10.1186/s13244-020-00887-2
Qin H, Hu X, Zhang J, et al. Machine-learning radiomics to predict early recurrence in perihilar cholangiocarcinoma after curative resection. Liver Int. 2021;41(4):837-850. doi:10.1111/liv.14763
Mayerhoefer ME, Materka A, Langs G, et al. Introduction to radiomics. J Nucl Med. 2020;61(4):488-495. doi:10.2967/jnumed.118.222893
Granata V, Fusco R, Setola SV, et al. An update on radiomics techniques in primary liver cancers. Infect Agent Cancer. 2022;17(1):6. doi:10.1186/s13027-022-00422-6