A Helping Hand from AI
Using AI to enhance personalized healthcare for patients with prostate cancer
sponsored by AIRA Matrix
An interview with Chaith Kondragunta
Could you give me an overview of AIRA MATRIX as a company?
AIRA MATRIX started its operations about seven years ago. Although our core operations involve developing artificial intelligence (AI)-based solutions for digital pathology, our beginnings were in preclinical research and development with pharmaceutical companies. Since then, we’ve expanded our scope of operations to include clinical healthcare diagnostics. In both areas, our primary focus is on AI solutions for cancer care. However, in the drug discovery and development (preclinical) stage, we focus on translational oncology solutions to help bring new drugs to market faster and more affordably. We do that by enhancing the efficiency of preclinical pathology and toxicology workflows. On the healthcare diagnostic side, we develop solutions for the diagnosis, prognosis, and prediction of a few different cancers – ultimately aiming to enable precision diagnostics and personalized healthcare.
What challenges do lab medicine professionals face in diagnosing and grading prostate cancer?
Prostate cancer is one of the more prevalent cancers, so there is a large volume of cases. Longevity is also increasing, meaning that healthcare systems across the globe are seeing an even larger rise in demand for prostate cancer diagnostics. If you dig a little deeper, you’ll see that it’s not just the volume of cases – underdiagnosis and overdiagnosis are also prevalent. When the cancer is not properly diagnosed at the right time, it can metastasize, which has a real impact on patient mortality. On the other hand, overdiagnosis can lead to morbidity, in which more radical intervention methods or stronger treatments are delivered than are strictly required.
These problems arise for two reasons: i) pathologists are overworked and ii) there is a lot of variability in the diagnosis and stratification of prostate cancer. Two different pathologists might interpret biopsy or patient samples in different ways – a discrepancy that could lead to mortality, morbidity, or even misdiagnosis.
How does AIRA MATRIX assist pathologists and clinicians in stratifying disease?
Stratification plays a large role in understanding where a patient falls on the Gleason grading scale and what intervention is required – but misdiagnosis during stratification may cause the patient to slip through the net or yield suboptimal treatment. For example, assigning a Gleason grade of three versus four can be a difficult decision; three is a “watch and wait” situation, whereas four requires intervention and treatment. As you can imagine, a patient's grade group and ensuing treatment has significant quality-of-life ramifications, so we are looking at how to more finely grade cancers – for example, whether we can say that a cancer is “a three that behaves more like a four” or vice versa. This would allow us to provide more nuanced information to physicians so they can make the right decisions for their patients.
We are also trying to augment MRI-based diagnosis by using knowledge gained from pathology reporting. MRI images are routinely used to assess images of prostate cancer and, though they are much less painful and invasive than a biopsy, they have a lot of accuracy and interpretation errors. Our project takes the gold standard of pathology reporting and translates it to an MRI diagnosis so we can leverage the strengths of each – helping patients by providing less invasive testing methods without compromising accuracy.
On the prognosis and prediction side, we’re working on extremely rich datasets with multiple patient cohorts that have been followed for years. This way, we can investigate not just the pathology or radiology modalities, but also other factors that we can consider in our conclusions on disease course or management. This will help us make prognosis predictions much more accurate.
What achievements are you most proud of so far?
What I’m most proud of is our ability to tackle complex problems. We have always viewed AI-based solutions in pathology as more than just a tool to help pathologists quantify their work or provide simple quantification or segmentation – there are plenty of other companies and AI-based techniques doing this. Instead, we have focused on the complex problems pathologists experience because of the sheer amount of information the human eye must process, which makes tasks extremely difficult or time-consuming. For example, one of our products in the toxicology space takes a few minutes to accomplish a task that would take a human pathologist several weeks. As you can imagine, drastically reducing time constraints is a massive achievement for us because we are driving the field forward with AI solutions and helping pathologists achieve better outcomes faster.
Chaith Kondragunta is CEO of AIRA Matrix, Mumbai, India.
