By Deborah Borfitz
February 2, 2021 | As more targeted therapies become available to cancer patients based on the phenotype of their tumor, so too will understanding of the tumor habitats to reduce—and perhaps eventually eliminate—the need for biopsies before, during, and after treatment. That is the promise of computational approach being developed by researchers at the University of Cambridge (U.K.) combining routine computerized tomography (CT) and ultrasound (US) images to create a visual guide for doctors as they obtain tissue samples.
The initial goal is fewer, more targeted biopsies that adequately capture the heterogeneity of a tumor, says Evis Sala, professor of oncological imaging at University of Cambridge and co-lead of the CRUK Cambridge Centre Advanced Cancer Imaging and Integrated Cancer Medicine Programmes. Sala led a study newly published in European Radiology (DOI: 10.1007/s00330-020-07560-8) where the precision sampling technique was used on six patients with advanced ovarian cancer who were due to have ultrasound-guided biopsies prior to starting chemotherapy.
In all cases, researchers succeeded in combining CT-, radiomics–based tumor habitats with US images, says Sala. That is, they applied data characterization algorithms to analyze and extract additional information from the data-rich images created by the CT scanner, which were then superimposed on the US image, to identify and map distinct areas and features of the tumor.
The CT/US fusion accuracy was high for larger pelvic tumors and lower for smaller clusters of cancer cells that had metastasized to the omentum (the apron of fat that protects the abdomen). Study findings suggest the approach, once validated biologically and in larger clinical studies, could be seamlessly integrated into routine clinical practice for patients with ovarian cancer, Sala says.
An ovarian cancer diagnosis is most typically high-grade serous carcinoma, for which outcomes have remained unchanged for the last 20 years, Sala says. The new computing technique is an attempt to extend to the imaging level the tumor heterogeneity known to exist at the tissue level.
Precision tissue sampling is useful in helping clinicians predict whether patients are more or less likely to respond to available treatments and make any needed course corrections based on new, incoming information, says study co-author Mireia Crispin-Ortuzar, Borysiewicz Biomedical Research Fellow at the CRUK Cambridge Institute. This includes decision-making beyond the specific drug to be given, such as the utility of giving a patient chemotherapy prior to surgery.
When ovarian cancer is suspected, patients undergo a diagnostic biopsy to confirm the suspicion and tumor type, says Sala. This can be done by surgically removing the tumor, although doctors may opt for a minimally invasive procedure, such as image-guided biopsy, to extract three or four tissue samples—following the same entrance route, or several separate passes, depending on the technique used.
Clinical Workflow
Sala began working with Crispin-Ortuzar on the new computational method about three years ago when she returned to the University of Cambridge after spending more than five years at Memorial Sloan Kettering Cancer Center, she says, where their paths first crossed. The high-level work Sala did at Memorial to capture tumor heterogeneity on images found the right combination of clinical expertise and infrastructure at Cambridge to bring closer to real-world application.
That meant developing an ultrasound-guided biopsy technique to examine tumors before and during treatment, and not just after surgery to remove the cancer, explains Crispin-Ortuzar. And it required an interdisciplinary team inclusive of gynecological radiologists, medical oncologists, computational biologists, and informaticians.
To correlate images to the underlying biology of tumors, and ensure they reflected the spatial heterogeneity of ovarian cancer, the researchers established a collaboration with Canon Medical Systems, Crispin-Ortuzar says. Their US-guided biopsy technique uses Canon’s Aplio i800 US system, PVI-475BX convex transducer, and commercially available Smart Fusion software to combine US and CT data.
While taking a biopsy on patients, radiologists can see different areas in the tumor on the final combined US image to guide tissue sampling of the various distinct habitats, Crispin-Ortuzar says. Their spatial arrangement is indicated by contrasting areas on the CT scan, and these habitats are identified, colored, and then directly mapped onto the live ultrasound image.
The software codes used are all open source, she adds, and a user-friendly package will soon be available to download allowing anyone to reproduce the approach. The only potential barrier would be access to a CT fusion machine for coupling images.
The targeted biopsy technique was added to the clinical workflow at Addenbrookes Hospital in Cambridge a few months ago, says Sala, noting that it increases standard biopsy procedure time by only about 10 minutes. Coordination among team members is essential for the pre-biopsy steps—CT image acquisition, tumor segmentation, and creation of imaging habitats that get co-registered with US images.
Virtual Biopsies
In the next phase of development, the researchers will try to relate the genetic makeup of patients to the pathology of their tissue samples, Sala says. After that, they plan to conduct a clinical study using tumor habitats identified on magnetic resonance imaging of ovarian cancer patients prior to chemotherapy.
Clinical studies will be needed to ensure a “robust correlation” between biological and imaging findings can be replicated across institutions and geographies, says Crispin-Ortuzar. Trials will also be needed to see how well expectations regarding phenotype and response to treatment can be met when a biopsy is taken from only one of the radiomic habitats.
If all goes well, says Sala, use of routine imaging modalities to sample regions with distinct radiomic habitats could be considered “more standard of care” and lay the foundation for future risk stratification. Change will occur in steps over the next two to five years, she predicts, starting with an across-the-board reduction in the number of tissue extractions and ending with their replacement by virtual, image-based biopsies.