How AI can optimise early cancer treatment

Male doctor in surgical clothes looking at vertebral mri scan headshot
Image © megaflopp | iStock

Early cancer treatment is vital to sustaining high cancer survival rates, but could AI improve clinician capabilities and predict worse cases?

For a person living with lung cancer whose diagnosis comes at the earliest stage, your odds of successful cancer treatment are almost 60%. If your diagnosis arrives at stage 4, however, the probability of success dramatically decreases to less than 10%.

With a window this small and stakes this high, there often is no time for a traditional trial-and-error method. Doctors want to find the optimal treatment drugs as soon as possible to shrink a patient’s tumours and improve their quality of life.

The means of cancer treatment are even more varied than the number of patients, with the combination of surgery, drugs, and associated treatments creating a constellation of possibilities. In addition, oncologists and primary care clinicians are under a huge amount of pressure to find the right solution for each patient as soon as they can, picking from a very high combination of drug pairings.

That’s why, as healthcare providers turn to AI’s predictive powers to augment frontline doctors’ capabilities, they’re using generative AI platforms to process millions of data points and identify and test tumour samples for drug matchups. These use cases are already revolutionising cancer treatment by providing unprecedented insights into disease progression and drug efficacy, leading us into the next generation of medical treatment.

Healthcare providers turn to AI’s predictive powers to augment frontline doctors’ capabilities

Augmenting clinician capabilities

With AI playing an ever-growing role in cancer treatment and research, oncologists can use machine learning to predict how a certain tumour will grow or shrink when matched with a given drug or combination of drugs. Knowing how individual patients will respond to drug pairings empowers doctors to design more personalised treatments and conduct automatic analysis of huge datasets: covering past treatments, outcomes, and patient-specific genetic information.

With this head-start, primary care providers have the time to do their best work – having been provided with narrowed-down, specific sets and types of drugs to use by their AI helper. These models use a representation of tumours at the molecular level, including RNA patterns and protein and drug interactions, to cut down on the manual, repetitive work performed by trained doctors. This can significantly improve the speed and accuracy of treatment decisions and help minimise side effects.

cancer patient in the NHS
Image © Fly View Productions | iStock

Maximising the efficacy of chemotherapy

On average, 42% of stage 3 and 4 cancer patients are in chemotherapy. This is a difficult, complicated process that has a measurable and severe impact on the individuals that go through it. The fewer drug cocktails and invasive treatments they must go through, the better for all parties involved. That’s why the aim of predictive modeling, helping doctors to choose the best course of action for a particular patient, is so important in the chemotherapy process.

Given that best practices for cancer treatment and drug prescriptions will also depend on the variety of cancer that a patient is diagnosed with, as rates of success vary as much as 98% to 1% between types, the ability to tailor a chemotherapy programme using the patient’s genetics, alongside the clinician’s best judgment, will be vital in achieving the most beneficial outcomes for patients in coming years.

However, the volume, variety, and velocity of pharmacogenomic data available for analysis are increasing daily due to the influence of high-throughput technologies. This leaves healthcare and research organisations with vast amounts of underutilised, often unstructured, data.

Foundation models allow organisations to reduce thousands of legacy models into one

The key lies in generative AI foundation models. As massive, highly versatile models, they’re easily able to unlock insights trapped in unstructured data through powerful emergent capabilities. Using in-context learning, with zero and few-shot prompts, generative AI foundation models can solve dozens of different tasks with a single model.

The watchword of 2023 is efficiency. That’s why foundation models –  combining accuracy, versatility, and scale – are ideal. They allow organisations to reduce thousands of legacy models into one: examining millions of gene expression and mutation profiles at lightning speed.

What can oncologists expect from AI?

Research organisations, AI companies, and healthcare providers each have a critical role in this process of cancer treatment. In order to provide the end-user with the best experience and outcomes, the industry must listen to the concerns of patients and doctors alike and work to integrate their feedback into the programming of the models themselves.

As the underlying technology advances, so must the sector. The kind of predictive analysis that would have been impossible only a few years ago is now a real way to improve lives. Drug response prediction is no longer limited to per-drug or per-cell line analysis, and can now involve the use of integrative models – using both of these simultaneously as inputs.

Oncologists can expect AI to operate as an extension of their existing capabilities rather than a replacement. They will be able to employ AI models to predict how a certain tumour will grow or shrink when matched with a given drug or combination of drugs, analyse more varieties and volumes of data in a fraction of the time, identify patterns that may have otherwise gone unnoticed, and understand how different treatments may interact with each other – such as in patients with comorbid conditions.

Neural networks’ high capacity enables them to conceptualise complex interactions between drugs and cancer cells at the molecular level, but at the end of the day, the final say is in the hands of the doctor. AI supplements and amplifies their work, offering a deeper understanding of cancer treatment possibilities than ever before. This can only result in more informed decisions and earlier treatments for patients, improving outcomes both immediate and long-term.

AI has a vital part to play in advancing cancer research and accelerating treatments. Its integration into healthcare will enable doctors to make faster decisions with more confidence – ultimately improving the quality of life for those affected by cancer. Early treatment is often the key to success, and AI plays a huge part in making this a reality.

This piece was written and provided by Marshall Choy, SVP of Product at SambaNova Systems


Please enter your comment!
Please enter your name here