Circulating tumor cells (CTCs) are cells that have broken away from a primary tumor and entered the circulatory system. They can initiate new tumors (metastases) that are the driving factor for the vast majority of cancer-related deaths. Biopharma researchers, cancer researchers and oncologists have focused on the detection and analyses of CTCs to monitor drug responses and determine appropriate treatments.
Dr. Peter Kuhn brainstorming with researchers at the University of Southern California and Epic Sciences. He is the principal architect of the Epic Sciences platform and has invested his career into the research, development and the clinical utility of the liquid biopsy.
Epic Sciences (USA) describe themselves as cancer fighting pioneers. Their analyses from a single blood draw produce comprehensive cancer profiles, including information on CTCs.
We spoke with Dr. Peter Kuhn, Founder and Chief Scientific Adviser, and Dr. Jiyun Byun, Director of Computer Vision Technologies, about the ground-breaking work being done at Epic Sciences and how they are using automated digital slide scanning to improve their cancer profiling techniques.
What is your mission at Epic Sciences?
Epic Sciences’ mission is to extend patient lives by delivering the clarity to guide treatment decisions based on circulating tumor cells (CTCs) in a blood sample. Each cancer patient faces a lack of certainty and fear about cancer treatment. No treatment works the same for every person, even those with the same type of cancer. Often patients do not respond to treatment or continue to receive treatment long after it stops working.
The Epic Sciences platform was invented to transform care for cancer patients by identifying and analyzing circulating tumor cells in the peripheral blood of patients to diagnose the disease, prognose the long-range outcome and predict the response to therapeutic interventions.
Instead of making any assumptions about rare cancer cells, we innovated our imaging platform to capture images of all the cells in a blood sample without bias and to identify rare cancer cells by machine learning models combining protein expression and morphological features derived from patient data collected over a decade.
Dr. Jiyun Byun leads development of the Epic Sciences imaging technology, leveraging computer vision and machine learning to support current and future products. She developed multiple imaging algorithms that are a part Epic’s clinical tests.
Our solution to this problem has its very foundation in high quality, high resolution, high speed and highly reproducible fluorescent imaging. Each cancer cell is typically hiding in a sea of millions of normal cells as they travel through the body. We developed an approach to find these heterogeneous, rare cancer cells in the background of millions of normal immune cells in a blood sample. We innovated our image acquisition and data analytics to solve the most important ‘Where’s Waldo’ puzzle.
How do you use microscopy?
When patient samples are received, the red blood cells are lysed, and the sample is centrifuged to isolate the white blood cell pellet. Then the white blood cell fraction is deposited onto glass microscope slides. Following this is staining of the cells using antibodies tagged with fluorescent dyes; this helps us visualize the morphology and biomarker expression on each type of cell.
After the immunofluorescent antibody staining for each batch of slides has been completed, the slides are prepared for automated slide scanning using ZEISS Axioscan, which can automatically scan 100 slides in a single run. Each slide is completely scanned, top to bottom, in all colors, in approximately 20 minutes and automatically uploaded to our cloud for processing for analysis.
A blood sample with staining for DNA (blue), CK (red), CD45 (green), and AR-V7 (white). This example shows two CTCs indicated by CK (red) and a majority of white blood cells showing CD45 membrane marker (green). Acquisition was done with a ZEISS Axioscan digital slide scanner with a 10x objective.
Unlike other CTC technologies, we directly image all cells in the sample to avoid any potential cell loss from enrichment steps, which allows us to fully understand clinically useful information from the patient’s cancer cells that today can only be determined by tissue samples. We are able to find, literally, a “one-in-a million” type of cell, such as rare cells or CTCs by using state-of-the-art machine learning algorithms. From a whole slide scan image, 3 million cells are automatically detected, and a unique id and slide coordinate (x,y) are assigned to each cell. The same CTC is then re-located and higher resolution images are acquired for deeper analysis.
Identified CTCs are further characterized based on protein expression or morphological characteristics associated with clinical information, providing new tools and capabilities to improve diagnosis and treatment. For example, we have characterized AR-V7+ CTCs from metastatic prostate cancer patient samples based on AR-V7 nuclear localization and currently provide Oncotype Dx AR-V7 Nucleus Detect tests to identify patients who are likely to be resistant to androgen-directed therapies.
After scanning the entire slide, identified CTCs are imaged again at higher resolution for in depth analysis using automated, digital slide scanning.
Why is digital slide scanning critical to your process?
A single patient sample will create multiple slides – up to 40 slides depending on a test – so efficient digitization of slides is critical.
The integration of the ZEISS digital slide scanner into the Epic Sciences imaging platform has enabled us to improve our throughput and efficiency.
What big challenge do you hope to overcome next?
Epic’s multi-technology approach is to provide comprehensive information from a single blood draw. Our imaging platform is being developed to maximize relevant cell information from a single draw by increasing imaging modalities (integrating brightfield), channels, and resolution. Tumors evolve and CTCs are heterogenous, so no single solution will work for all patients. Phenotypic and genotypic details of every blood draw can help patients to choose the most effective therapy and to monitor therapy response in a timely manner. In order to carry this mission, we are working to increase throughput and efficiency of imaging and continue to advance our AI-powered platform.
ZEISS Axioscan is a component of Lab Developed Tests (LDTs) in the Epic Sciences CLIA laboratory.