Potential Roles for Radiomics in Cancer

The main benefit that radiomics could provide in the preoperative assessment of endometrial cancer is to reduce variability among clinicians’ interpretation of images, which is important for staging the cancer, planning surgery, and determining whether there is a role for neoadjuvant or adjuvant treatment, Dr Gallix said.

“If you ask 10 radiologists or 10 pathologists to review [images], there’s going to be a lot of discrepancy in the analysis of basic things like size, vascularization, [and] invasiveness,” he explained.


Continue Reading

Dr Gallix has seen these inconsistencies first-hand while sitting on tumor boards at different university hospitals in France and Canada and has been working with colleagues over the past decade to improve techniques to make image analysis more objective and quantitative.

The application of radiomics analyzed in the review “is certainly an area of unmet need and an area that is very challenging,” said Michael V. Knopp, MD, PhD, director of the Wright Center of Innovation in Biomedical Imaging at The Ohio State University in Columbus, who was not involved in the review.

“The data show that this is not yet ready for prime time,” Dr Knopp said. “However, [that] does not mean that it is not worthwhile to utilize, to explore, to continue to develop.”

Moreover, Dr Knopp added, there are some settings in which radiomics already provides important additional information to clinicians.

In the past year, the US Food and Drug Administration approved software products that use radiomics to analyze CT images and mammograms to predict lung cancer risk and detect early-stage breast cancer, respectively.6,7

Other radiomics applications are being actively investigated; for example, the ability to characterize tumors and predict treatment responses to immunotherapy and chemoradiotherapy for various cancer types.8,9

Many research groups are exploring the role of radiomics, particularly in breast, lung, and prostate cancers because these are common cancer types and also complex in terms of interpreting the images, Dr Knopp said. The study of radiomics for brain cancer assessment is also advanced because clinicians have had many years of experience imaging the brain for other diseases.

In terms of when we can expect to see more of these radiomics applications, Dr Knopp thinks that “we are closer to 5 years than 10 years.”

However, the medical field should not be worried about AI replacing human intelligence. The idea is not that “computer replaces physician,” but that “computer makes clinician a better clinician,” Dr Knopp said.

What Does Radiomics Need?

The computer science side of radiomics has become easier than ever, according to Dr Gallix. Deep learning, which is an advanced form of machine learning, is robust and reproducible as long as there is a big enough data set to train the algorithm.

Radiomics is lagging behind more on the imaging side, Dr Gallix explained. The protocols for acquiring and processing images can vary widely across centers and the equipment and software they use, particularly for MRI.

“What is most important as it relates to the topic of radiomics is that we have consistent data quality” that is amenable to AI-based technologies, said Dr Knopp, who was part of a group that recently developed recommendations for the ongoing study of radiomics for the National Cancer Institute’s (NCI) National Clinical Trials Network (NCTN).8

As Dr Knopp explained, parameters may have to be adjusted during imaging procedures based on patient characteristics, such as height, to achieve adequate spatial or contrast resolution, rather than striving for a single cookie-cutter protocol for all imaging.

The NCI and Radiological Society of North America (RSNA) are leading efforts to define standard protocols for acquiring images to ensure high quality.

Dr Knopp and colleagues recommended to the NCTN that research groups first evaluate their AI technologies by using real-world data. Dr Knopp noted that a lot of data are already available to investigators from sources such as The Cancer Imaging Archive.10

Yet another area to be explored in radiomics is determining which clinical outcomes are appropriate endpoints. For example, Dr Gallix questioned whether a radiomics approach would be approved by regulatory groups if it could reduce the time to determine treatment efficacy from 6 months to 2 months.

“If you can anticipate the response to treatment, you can save a lot of secondary effects and a lot of money, and on top of that, you can adjust the treatment,” he noted.

References

  1. Lecointre L, Dana J, Lodi M, Akladios C, Gallix B. Artificial intelligence-based radiomics models in endometrial cancer: A systematic review. Eur J Surg Oncol. Published online June 22, 2021. doi:10.1016/j.ejso.2021.06.023
  2. Ueno Y, Forghani B, Forghani R, et al. Endometrial carcinoma: MR imaging-based texture model for preoperative risk stratification—A preliminary analysis. Radiology. 2017;284(3):748-757. doi:10.1148/radiol.2017161950
  3. Chen X, Wang Y, Shen M, et al. Deep learning for the determination of myometrial invasion depth and automatic lesion identification in endometrial cancer MR imaging: A preliminary study in a single institution. Eur Radiol. 2020;30(9):4985-4994. doi: 10.1007/s00330-020-06870-1
  4. Luo Y, Mei D, Gong J, Zuo M, Guo X. Multiparametric MRI-based radiomics nomogram for predicting lymphovascular space invasion in endometrial carcinoma. J Magn Reson Imaging. 2020;52(4):1257-1262. doi: 10.1002/jmri.27142
  5. Yan BC, Li Y, Ma FH, et al. Preoperative assessment for high-risk endometrial cancer by developing an MRI- and clinical-based radiomics nomogram: A multicenter study.
    J Magn Reson Imaging. 2020;52(6):1872-1882. doi:10.1002/jmri.27289
  6. Optellum receives FDA clearance for the world’s first AI-powered clinical decision support software for early lung cancer diagnosis. News Release. Business Wire. March 23, 2021. https://www.businesswire.com/news/home/20210323005236/en
  7. Whiterabbit, creator of AI to improve and streamline breast cancer diagnoses, emerges from stealth to announce new CEO and FDA clearance. News Release. Cision PR Newswire. May 11, 2021. https://www.prnewswire.com/news-releases/whiterabbit-creator-of-ai-to-improve-and-streamline-breast-cancer-diagnoses-emerges-from-stealth-to-announce-new-ceo-and-fda-clearance-301288111.html
  8. Nie K, Al-Hallaq H, Li XA, et al. NCTN assessment on current applications of radiomics in oncology. Int J Radiat Oncol Biol Phys. 2019;104(2):302-315. doi:10.1016/j.ijrobp.2019.01
  9. Bi WL, Hosny A, Schabath MB, et al. Artificial intelligence in cancer imaging: Clinical challenges and applications. CA Cancer J Clin. 2019;69(2):127-157. doi:10.3322/caac.21552
  10. The Cancer Imaging Archive. https://www.cancerimagingarchive.net/