Precision Medicine and AI in Oncology Patient Care

Intolerance to uncertainty drives innovations. As recently as 10 years ago, a map was necessary to find your way around a new town. Now, an app on a cellphone can not only plan out your quickest route, but will also tell you exactly how long it will take you to reach your destination. The only medium in which to receive weather reports used to be on the local news stations. Now, an app provides hourly temperature and condition updates.

What does that evolution look like in cancer care? This was the question posed by Tushar Pandey, CEO, SimBioSys, Inc, to kick off Sunday’s “Session 6: Technology in Oncology Clinical Pathways” during the 2022 Oncology Clinical Pathways Congress.

“Our answer is ‘precision medicine.’ And what I would argue is that our view of precision medicine is a little myopic,” Pandey said in his presentation, “AI in Cancer Care: Role in Precision Medicine.”

Cancer care is broader than a dictionary definition of precision medicine, Pandey said. Cancer is an intimate journey that involves precision risk stratification, screening, diagnosis, tumor profiling and testing, treatment selection, dosing, monitoring and surveillance, and survivorship.

“We have to bring all of these individual components together if we’re going to truly realize the impact that a Google Maps has had on our lives,” he said.

So, what does precision medicine mean for the future, and how do we get to that reality? First, since the patient journey includes individualized and personalized factors, such as risk stratification, precision dosing, and monitoring plans, that means every patient’s journey and needs are different.

Second, we can't get to that reality with human beings alone, Pandey said. The American Society of Clinical Oncology is projecting a shortage of more than 2,200 oncologists by 2025, even though the demand for cancer treatment is expected to grow by 40% over the next 6 years.

However, Artificial Intelligence (AI) doesn’t seem to be the answer, either. IBM’s Watson didn’t work. Neither did other AI entries in the field. And with the current focus on racial disparity, it is clear that the mechanisms and tools that have been created do not represent minorities.

But AI has its place in medicine. The FDA released data on 521 devices they have recently approved over the last few years involving AI. A vast majority of these approvals have come in radiology.

“Most of our comfort is in the field of radiology when it comes to using AI,” Pandey said. “But radiology is just one slice of the cancer puzzle. We have to think broadly, but radiology has made an impact and has paved the way for us to think about it beyond medical imaging.”

A recent survey of radiologists shows that AI can have significant benefits in certain areas including breast, oncologic, thoracic, and neuro imaging, and AI saved time that could then be spent with patients. But AI can’t do it alone, either.

“Half of the respondents said that their patients don’t want to be treated by a machine, or diagnosed by a machine,” Pandey said. “There’s no future where a patient comes in and a machine diagnoses them and treats them. There’s a hybrid approach between the two. And what AI has done in radiology is it’s given radiologists a little more time to spend with their patients and dive deeper.”

Currently, AI is being applied in areas such as patient engagement (eg, chatbots), drug development, data and analysis, and diagnostics.

“But this doesn’t get to the heart of the problem, which is what pathways has been trying to solve,” Pandey said. “How do we do better for patients when it comes to treatment planning?”

Pandey presented an example of a patient with HER2-positive infiltrating ductal carcinoma. A survey of 130 oncologists returned no fewer than eight different treatment regimens.

“The same patient could get six rounds of treatment or could get 24 rounds of treatment, depending on which doctor they saw,” Pandey said. “When you think about it from an economic standpoint, there’s a huge variation amongst these. From an efficacy standpoint, there’s a huge variation. But there is a right patient for one of these regimens. The key and the future of AI in cancer care is identifying how to get there.”

There are additional challenges in AI in terms of quality of the data used to train the model, the perceptions of those entering the data, and morality.

Looking to the future, perhaps the best use of AI in cancer care is through guided intelligence, which involves integration of multimodal data and individual components to present causality, the need to be transparent and explainable, and the opportunity for clarity on decisions and to derive insights justified by science.

“Guided intelligence is taking components of AI and bringing it together with traditional mathematical modeling approaches to be able to support physician decisions,” Pandey said.

There is more and better data available, but data alone is not the answer. The value of AI is not the technology, but in how it can be embedded in systems to help change clinical workflow and operational processes. And when the value and limitations of AI are understood, the fear of the unknown vanishes.

“It’s time to embrace the AI revolution beyond medical image interpretation,” Pandey concluded.