Pathologists, We Need to Talk About Tech

“Your biopsy results show you have cancer,” says your oncologist. Your heart sinks. “The next step is to wait several weeks for your genomic test results so we can see if you are eligible for any targeted therapies or clinical trials.” It’s hard to focus as the questions start to fill your head, but a few whys immediately spring to mind. Why me? Why is this happening? Why now? Once this new reality sets in, there’s one “why” that drowns out all the others. “Why is the wait so long?”

I hope that you and your loved ones will never hear the words written at the start of this article; in fact, my hope is that no one will need to endure the anxiety of waiting or the frustration of not getting the right treatment for their distinct cancer mutation. But, in 2022 alone, there will be an estimated 1.9 million new cancer cases diagnosed in the US (1) – and every one of those people deserves the best possible diagnostic care.

In recent years, it has become clear that genomic information is crucial for accurate cancer diagnosis and optimal treatment selection – but genomic screening is a long, expensive, and often inaccessible process. We have a growing need for genomic screening – and that calls for change.

My personal cancer journey began when my mother received a diagnosis of aggressive stage IV ovarian cancer. A data-oriented person, I dove in to learn everything I could about her disease and was surprised to realize how many uncertainties and probabilities her diagnosis and treatment process had. As an engineer, I was most comfortable when dealing with precision – highly controlled systems with repeatable outcomes. The uncertainty in my mother’s treatment showed me that there was a tech-shaped hole in oncology, so I made it my goal to empower physicians with precise, crucial, timely information to change the way cancer is diagnosed and treated.

Today, that goal has become my life’s work. My colleagues and I now know that deep learning and artificial intelligence (AI) technologies can allow physicians to receive genomic screening reports in real time, enabling immediate decision-making for additional testing, faster screening for clinical trials, and a well-timed treatment approach. By harnessing the power of AI, we can democratize genomic screening and totally transform patient care. Patients deserve a shift in the healthcare industry to take this change in precision medicine from concept to reality.

Of course, the title of this piece is intentionally confrontational. Pathologists don’t need to “talk” with anyone about technology – they use a plethora of different machines, computers, and other gizmos to do their work every day! But the sentiment of my slightly standoffish headline still rings true; why, exactly, aren’t pathologists using more tech?

Using only an H&E-stained biopsy image, advanced deep learning algorithms need just a few minutes to provide a report for distinct cancer mutations and novel biomarkers and even to predict whether a patient will respond to specific therapies. Technology can detect patterns in the biopsy image that cannot be seen by the human eye, which can help prioritize the use of biopsy tissue, accelerate clinical trial enrollment, and help physicians optimize treatment protocols.

In addition to reducing the lengthy delays inherent in present-day molecular testing and the high costs of building a next-generation sequencing (NGS) lab, AI-enabled biopsy image analysis can help address other issues that frequently arise in the pathology lab – such as not having enough tissue available, technological challenges with DNA extraction, and difficulty with result interpretation. In lung cancer, for example, less invasive core needle biopsy retrievals mean that up to 20 percent of patients cannot receive the benefit of NGS due to a lack of tissue availability. What’s more, technological and interpretational challenges often lead to “inconclusive results.”

The use of technology can significantly reduce inconclusive results because it is integrated directly into a lab’s workflow and does not consume any additional tissue. Providing pathologists with a genomic screening report as soon as a slide is scanned allows them to better prioritize confirmatory molecular testing to conserve tissue. These AI models have been tested using different types of scanners and slide staining methods and my team’s results have been reproduced in multiple medical centers. AI can also offer insight into which treatments an individual patient will most likely respond to, evaluating genomic data, phenotypic data, proteomics, and spatial mapping to reduce the likelihood of unsuccessful treatment or unnecessary side effects.

Today, precision oncology is a reality for only a small percentage of patients, but it remains our best option for achieving real breakthroughs and realizing individualized approaches to treatment. If we are going to achieve the goals of precision medicine, we must democratize genomic screening so that all physicians have the best information available to choose the most suitable treatment plan. Furthermore, leveraging AI to reduce testing costs and make genomic screening available to everyone, everywhere, will make a real impact on the quality of patient care – not only by helping to select the most appropriate treatments, but also by speeding up the selection process to reduce wait times and anxiety for patients and families. I know the anxiety of treatment uncertainty all too well – and I know that now is the moment to do something about it.