Predicting Cancer Outcomes

Predicting disease outcome through digital pathology-based imaging

 

Growing up in India in the early 1990s, one had two clear career pathways – either become a doctor or an engineer, stated Anant Madabhushi, professor of biomedical engineering and director of the Center for Computational Imaging and Personalized Diagnostics. With several doctors in his family, Madabhushi felt pushed toward a medical profession, a highly competitive endeavor in his country.

“Long story short, I didn’t get into medical school. An uncle working for GE sent me a couple of books on biomedical engineering and instrumentation. That piqued my interest and it seemed like a good convergence of the medical and engineering fields,” said Madabhushi, who has received several patents for his research since joining the faculty of the biomedical engineering department at the Case School of Engineering two years ago. “Just looking around at the unmet clinical need and how biomedical engineers have an opportunity to make an impact and advance medical technologies on so many levels – that’s what excites me about this field.”

Madabhushi’s work is focused on using computational image analysis, artificial intelligence and pattern recognition tools to find cues or features from MRI, CAT scan and digital tissue pathology images for predicting disease outcome, primarily in breast, prostate and lung cancer patients. While science is fairly good at diagnosing cancer early enough for treatment, the technology around characterizing the outcome is still not ideal. Which patient will benefit from more aggressive chemotherapy, or as in about 20 percent of breast cancer patients, which might require little or no therapies and be spared the toxic side effects?

What Madabhushi’s team has done is take digital images of a routinely acquired biopsy or tissue specimen and look for distinct patterns and appearances that correlate to a more or less aggressive disease. “You can take your tissue slide, put it into a scanner and create a high-resolution digital image. We then run our software programs on it and come up with a prediction of the disease outcome.”

That is a game changer, Madabhushi continued, not only because it pushes the needle in terms of precision medicine, but because it is fundamentally non-clinically destructive technology. The technology can apply to a woman diagnosed with breast cancer in Dubai or Brisbane – there are no barriers and the protocol is the same. Once the slide is digitalized and uploaded to the cloud, programs can render a diagnosis from anywhere. Think of the cost savings, especially to lower- and middle-income countries, he said.

“For me, success is the translation of these technologies to the patient,” Madabhushi proclaimed. “We’ve achieved success in terms of publications, in terms of grants and patents so far. But in my mind, as a biomedical engineer and somebody who got into this area with a clear vision of developing methods that would be used clinically, I have to say we’re not quite there yet. We’re very, very close though.”

 

  

Anant Madabhushi, professor of biomedical engineering and director of the Center for Computational Imaging and Personalized Diagnostics