Virtual Biopsy: A Safer Future for Cancer Management

Virtual biopsy may provide a robust, non-invasive alternative in the surveillance and management of several types of cancer. This new imaging-based technique can extract encoded data from routine medical imaging scans and formulate likelihoods to help characterize tissue for the purposes of screening for cancer, diagnosing and grading cancer, planning oncologic treatment, and predicting prognosis of disease--all without any surgical or invasive intervention.

Although its use has been primarily studied in the oncologic setting, it can potentially be used for any disease process that requires analysis of its pathogenesis. Here, we discuss the technique of virtual biopsy, the variety of ways physicians can utilize it, and its impact on the outcomes of patients and healthcare.

What is Virtual Biopsy?

Virtual biopsy is a non-invasive technique to characterize tissue found in imaging studies. A great wealth of information can be interpreted by radiologists on imaging scans. However, additional encrypted data regarding the tissue’s intrinsic features, gene expression patterns, gene mutations, and molecular pathophysiology are still invisible to the human eye.

Access to this additional information can be granted through the use of radiomics. Radiomics is the applied field of radiogenomics which focuses on extracting objective, quantitative data found in images⁽¹⁾. Generally, the workflow includes obtaining the image, processing the image, creating a volume of interest, and then extracting mathematical data in that volume⁽²⁾. This can be possible through several imaging modalities, such as CT, MRI, PET, radiographs, and ultrasound imaging⁽²⁾. Examples of data collected include volume, shape, density, and location in addition to complex mathematical metrics assessing textures and dimensional analysis⁽²⁾.

Through radiogenomics, these phenotypic data can then be correlated with previous genetic data and can reveal and provide insight to the molecular pathophysiology of disease by mapping out the gene expression of the tissue with the assistance of artificial intelligence and characterization algorithms⁽²⁾.

Virtual biopsy offers an alternative to traditional surgical biopsies which are used to obtain histological data for the same purpose of diagnosis and treatment planning. Compared to traditional biopsy, the benefits are numerous. With virtual biopsy being a non-invasive procedure, there is a lesser risk of complications to the patient⁽¹⁾. Additionally, there is a lower cost associated with the service compared to traditional biopsies due to eliminating the operating room⁽¹⁾. Further, the patient would benefit from a quicker report on their tumor pathology, affecting their treatment timetable⁽¹⁾. Lastly, virtual biopsies can assist in making surgical biopsies even more conclusive, as discussed next.

Use of Virtual Biopsy in Screening

One area where virtual biopsy can enhance clinical outcomes is in accurate, early cancer detection. When patients undergo low-specificity screening, it generates increased risk for patients in the form of unnecessary further imaging studies and invasive biopsies of suspected lesions which could be benign. This further workup creates increased radiation and surgical risk exposure, in addition to the increased costs of these services⁽³⁾.

One example of this narrative is in the setting of lung nodules identified on low-dose CT screening. In a study conducted by Ma. et al., researchers obtained 127 indeterminate pulmonary nodules that posed the question of malignancy⁽⁴⁾. They were able to identify 583 texture-based features, such as nodule intensity and tumor heterogeneity. After formulating a model, the researchers were able to predict malignancy of the nodules correctly 82.7 percent of the time⁽⁴⁾.

Even more remarkable, in another study conducted by Yang et al., virtual biopsy was able to correctly identify malignancies in the setting of complex ground-glass opacities, as well. Their work found that the volume and diameter of ground glass opacity tumors showed correlation with the EGFR status of the tumors, also shedding light on the potential treatment options for the patient⁽⁵⁾.

Another area where virtual biopsy can improve cancer screening is in the setting of pancreatic neoplasms. Although there is no routine screening for pancreatic cancer, certain patient populations that are high-risk may undergo routine ultrasound or MR imaging screening⁽⁶⁾. High-risk patients typically have a family history, symptomatic pancreatitis, new-onset diabetes, or asymptomatic mucinous cystic lesions found incidentally⁽⁶⁾. These patients can be subject to risky surgical intervention if their pancreatic lesion seems suspicious based on qualitative features on imaging.

However, in one study conducted by Permuth et al., researchers were able to identify 14 radiomic features (11 textural and three non-textural) that differentiated malignant pancreatic neoplasms from benign intraductal papillary mucinous neoplasms. The prediction model that incorporated the 14 radiomic features collectively proved to be fairly accurate in their diagnosis (AUC=0.77 and P<0.05). In fact, the study also reported that three of the cases that had worrisome features on pre-operative scanning were correctly identified as true negatives by radiomic analysis, displaying higher accuracy than conventional feature analysis⁽⁷⁾. This, in turn, could save patients from particularly invasive biopsies in the future if further studies can reproduce and validate the models used in this study.

Use of Virtual Biopsy in Diagnosing Tumors

There are several applications where virtual biopsies can help diagnose malignant processes in the body. One example of this is in the setting of head and neck squamous cell carcinoma. In a study conducted by Ren et al., using MR imaging, researchers were able to formulate a predictive model that stratified whether the head and neck cancer was high-grade or low-grade in pathology.

Using a total of 970 radiomic features extracted from T2-weighted and contrast enhanced T1-weighted MR images, the study was able to discriminate WHO class grade I-II tumors from grades III-IV head and neck squamous carcinoma within the images⁽⁸⁾. The final testing cohort yielded an AUC of 0.853 and 0.849 for high grade and low grade respectively, proving it to be fairly accurate⁽⁸⁾.

Similarly, MR imaging can also be used for grading neurological tumors. Another study conducted by Bai et al., utilized radiomic data, such as the water molecular diffusion heterogeneity index (WMDHI) and the mean kurtosis values for predicting the grade of gliomas. The study concluded that both of these metrics were higher in high-grade gliomas when compared to low-grade gliomas (P<.05). With the same metrics, the researchers determined specific cut-off values for differentiating high-grade gliomas from low-grade gliomas. After cross-referencing with subsequent surgical biopsies, the sensitivity of both the metrics was 92.9 percent, with specificity being 100 percent, demonstrating itself as a precise model⁽⁹⁾.

Another paradigm where virtual biopsy can aid in the diagnosis of malignancy is in the setting of invasive ductal carcinoma. In a study conducted by Guo et al., researchers were able to identify 25 computerized texture analysis features (such as neighborhood gray-tone difference matrix-based features and their wavelet decomposition) when describing echogenicity patterns of breast lesions.

The texture analysis of these features yielded a “busyness” score, where the average value for estrogen receptor positive (ER+), progesterone receptor positive (PR+) and human epidermal growth factor receptor 2 (HER2-) tumors was 1.402 and in ER- PR- and HER2- tumors was 1.912 (P<0.001). This study shows there is potential for a reliable model to accurately identify the molecular subtype of breast cancer, not only aiding in diagnosis but also influencing treatment options⁽¹⁰⁾.

Use of Virtual Biopsy in Treatment Planning

Virtual biopsy can not only aid in diagnosis of tumors, but it can also use radiogenomic data to go a step further to help plan the treatments, as well. For example, one study examined its application in treatment planning for patients with glioblastoma. Previous research established that the isocitrate dehydrogenase mutation in glioblastoma is associated with a slower progression rate.

Patients affected with glioblastoma tumors that carry this mutation have completely different treatment options in terms of urgency of surgical intervention and adjuvant therapy. In a study conducted by Arnout et al., through MR imaging and radiogenomics, his team was able to correctly identify IDH status in three different cohorts at an 87 percent-to-89 percent accuracy when confirmed by subsequent brain biopsies⁽¹¹⁾.

Virtual biopsies can further provide insight on how to perform accurate surgical biopsies for treatment planning. Tumors that are “high grade” tend to exhibit the highest amount of tumor heterogeneity, which alludes to the differential genetic expression across the same tumor mass. This becomes important clinically because the sample obtained during biopsy may or may not paint the whole picture regarding the genetic makeup of the tumor--ultimately impacting how effective the treatment of choice will be.

However, with virtual biopsy, three-dimensional images of the patient’s tumor can be obtained via CT scanning. Using radiomics combined with radiogenomic data, this will allow the physician to map out the areas with the most genetic heterogeneity of the tumor, revealing the most ideal location to biopsy the mass and allowing for a conclusive and accurate biopsy⁽¹²⁾.

Virtual Biopsy in Predicting Prognosis

Virtual biopsy further proves to deliver comprehensive cancer care for patients by providing data on the prognosis for patients, as studied in several of the following tumor settings. One context where virtual biopsies provide prognostic data is bladder cancer. A study conducted by Lin et al., looked to find a predictive model for the prognosis of transitional cell carcinoma of the bladder. The study was able to identify seven radiomic features (wavelets and gradient orientations), which were then used to build a radiomics predictive model. Using the model, the researchers were able to stratify the patients into high-risk or low-risk groups in regard to progression-free tumor survival⁽¹³⁾.

Further work conducted by Iwatate et al., looked at the prevalence of p53 gene mutations in the setting of pancreatic ductal adenocarcinoma. The study iterated that the expression of the p53 gene mutation in the pancreatic malignancy was a significant factor for a poorer prognosis. With this, the researchers set out to develop a model that was able to predict if the pancreatic lesion exhibited the p53 mutation based on quantitative imaging features and radiogenomics⁽¹⁴⁾.

Lastly, a study conducted by Tokuda et al., sought to find a predictive model for the prognosis of breast cancer patients regarding recurrence of their disease. This study utilized a previously validated risk-stratifying tool to characterize the breast tumors as high-risk or low-risk for tumor recurrence based on their genetic mutations. Then, using MR imaging and radiomics, researchers concluded that factors, such as medium kinetic volume ratio and delayed phase kurtosis were associated with higher tumor recurrence. This allowed for the stratification of patients into high and low risk of reoccurrence groups⁽¹⁵⁾.

Final Thoughts and Future Directions

Virtual biopsy has shown to be effective in a variety of manners, including increasing the accuracy of cancer screening, helping diagnose and grade various tumors, and providing input in the treatment and prognosis of patients afflicted by oncologic disease. By virtue of being a non-invasive technique, elimination of unnecessary surgical risk and associated complications will contribute to decreased morbidity and mortality. Additionally, through its increased precision in cancer screening, patients will benefit from earlier diagnosis, directly improving their prognosis in some cases.

Further, patients will also receive tailored treatments consistent with their cancer’s mutational status and prognosis. Lastly, the healthcare system can expect to see a decreased need for invasive procedures, such as costly surgical biopsies, in turn reducing the economic burden cancer creates.

However, being a newer technology, there are some limitations, as well. Due to large amounts of data, there are increased computing power and storage requirements that not all institutions have the capacity to accommodate⁽¹⁶⁾. Next, the statistical approaches involved to analyze the data need to be fine-tuned: validating these models across several imaging platforms and patient populations with varying genetic makeups is a barrier⁽¹⁶⁾. Lastly, the lack of standardization of acquisition parameters used to derive predictive models and inconsistent radiomic methods further hinder the reproducibility of virtual biopsy⁽¹⁶⁾.

This being said, the potential is vast for virtual biopsies, radiomics, and radiogenomics. Medicine is moving in the direction of non-invasive procedures and virtual biopsy fits the trend. Patients can benefit from procedures that can be performed in the outpatient setting, while still benefiting from robust data about their disease that will help their physician treat them with arguably similar accuracy compared to their surgical counterparts.

References:

1. Kang, J. U. (2010). Virtual Biopsy [Point of View]. Proceedings of the IEEE, 98(4), 503-505.

2. Shui, L., Ren, H., Yang, X., Li, J., Chen, Z., Yi, C., ... & Shui, P. (2020). Era of radiogenomics in precision medicine: an emerging approach for prediction of the diagnosis, treatment and prognosis of tumors. Frontiers in Oncology, 10, 3195.

3. Thawani R, McLane M, Beig N, Ghose S, Prasanna P, Velcheti V, Madabhushi A. Radiomics and radiogenomics in lung cancer: A review for the clinician. Lung Cancer. 2018 Jan;115:34-41. doi: 10.1016/j.lungcan.2017.10.015. Epub 2017 Nov 8. PMID: 29290259.

4. Ma J, Wang Q, Ren Y, et al. Automatic lung nodule classification with radiomics approach. Proc. Proc SPIE 2016;9879. doi:1117/12.2220768.

5. Yang, Y., Yang, Y., Zhou, X., Song, X., Liu, M., He, W., ... & Jiang, G. (2015). EGFR L858R mutation is associated with lung adenocarcinoma patients with dominant ground-glass opacity. Lung cancer, 87(3), 272-277.

6. L. Gillies, R. J., & Schabath, M. B. (2020). Radiomics improves cancer screening and early detection. Cancer Epidemiology and Prevention Biomarkers, 29(12), 2556-2567.

7. Permuth, J. B., Choi, J., Balarunathan, Y., Kim, J., Chen, D. T., Chen, L., ... & Malafa, M. (2016). Combining radiomic features with a miRNA classifier may improve prediction of malignant pathology for pancreatic intraductal papillary mucinous neoplasms. Oncotarget, 7(52), 85785.

8. M. Ren J, Tian J, Yuan Y, Dong D, Li X, Shi Y, Tao X. Magnetic resonance imaging based radiomics signature for the preoperative discrimination of stage I-II and III-IV head and neck squamous cell carcinoma. Eur J Radiol. 2018 Sep;106:1-6. doi: 10.1016/j.ejrad.2018.07.002. Epub 2018 Jul 4. PMID: 30150029.

9. O. Bai, Y., Lin, Y., Tian, J., Shi, D., Cheng, J., Haacke, E. M., ... & Wang, M. (2016). Grading of gliomas by using monoexponential, biexponential, and stretched exponential diffusion-weighted MR imaging and diffusion kurtosis MR imaging. Radiology, 278(2), 496-504.

10. N. Guo, Y., Hu, Y., Qiao, M., Wang, Y., Yu, J., Li, J., & Chang, C. (2018). Radiomics analysis on ultrasound for prediction of biologic behavior in breast invasive ductal carcinoma. Clinical breast cancer, 18(3), e335-e344.

11. Chang, K., Bai, H. X., Zhou, H., Su, C., Bi, W. L., Agbodza, E., ... & Kalpathy-Cramer, J. (2018). Residual convolutional neural network for the determination of IDH status in low-and high-grade gliomas from MR imaging. Clinical Cancer Research, 24(5), 1073-1081.

8. Iwatate, Y., Hoshino, I., Yokota, H. et al. Radiogenomics for predicting p53 status, PD-L1 expression, and prognosis with machine learning in pancreatic cancer. Br J Cancer 123, 1253–1261 (2020). https://doi.org/10.1038/s41416-020-0997-1

12. Martin-Gonzalez, P., Crispin-Ortuzar, M., Rundo, L., Delgado-Ortet, M., Reinius, M., Beer, L., ... & Sala, E. (2020). Integrative radiogenomics for virtual biopsy and treatment monitoring in ovarian cancer. Insights into Imaging, 11(1), 1-10.

13. Lin P, Wen DY, Chen L, Li X, Li SH, Yan HB, He RQ, Chen G, He Y, Yang H. A radiogenomics signature for predicting the clinical outcome of bladder urothelial carcinoma. Eur Radiol. 2020 Jan;30(1):547-557. doi: 10.1007/s00330-019-06371-w. Epub 2019 Aug 8. PMID: 31396730.

14. Iwatate, Y., Hoshino, I., Yokota, H. et al. Radiogenomics for predicting p53 status, PD-L1 expression, and prognosis with machine learning in pancreatic cancer. Br J Cancer 123, 1253 1261 (2020). https://doi.org/10.1038/s41416-020-0997-1

15 (10. Tokuda, Y., Yanagawa, M., Minamitani, K., Naoi, Y., Noguchi, S., & Tomiyama, N. (2020). Radiogenomics of magnetic resonance imaging and a new multi-gene classifier for predicting recurrence prognosis in estrogen receptor-positive breast cancer: A preliminary study. Medicine, 99(16).

16. Lohmann, P., Bousabarah, K., Hoevels, M. et al. Radiomics in radiation oncology—basics, methods, and limitations. Strahlenther Onkol 196, 848–855 (2020).