AI is set to transform cancer care
Digital technology and artificial intelligence are arming clinicians with more data than ever before and offering the possibility of faster, more accurate diagnosis, more personalised treatment and better outcomes, reports Dr Tim Woodman in the second part of his analysis which began last month.
Digital oncology
Digital oncology refers to the integration of digital technologies and data-driven approaches into the field of oncology.
It involves the application of various technologies, including, but not limited to, digital platforms, data analytics and AI, to enhance cancer research and patient care.
The relationship between AI and digital oncology is closely intertwined, as AI often underpins many of the digital solutions developed.
Digital oncology offers the option to enhance patient-centred aspects of oncology care.
One example is telemedicine, which allows patients to consult with oncologists and receive follow-up care remotely, reducing the need for frequent in-person visits.
Then there is remote screening and diagnostics through devices, tools and technologies ranging from a simple at-home test to more advanced devices.
Precision medicine
It also plays a key role in enabling precision medicine by providing the tools to analyse and interpret complex patient data to determine the most effective treatment options.
Digital oncology enables remote monitoring, meaning healthcare professionals – or credible chatbots in the future – can assess information remotely for symptom management, treatment guidance and to check on progress and detect complications.
Such is its potential that the NHS has introduced its ‘Cancer digital playbook’, which provides support to organisations looking for digital tools that support the delivery of cancer pathways.
It provides case studies across the patient pathway from referral management to end-of-life care as an example of how to implement digital oncology technologies.
The benefits of digital oncology
For clinicians:
Reduced need for in-person visits with oncologists and hospitalisations because tele-medicine, remote monitoring, screening and diagnosis, and digital therapeutics will contribute to more efficient cancer care delivery and closer monitoring by multi-disciplinary teams.
Increased use of automation and technology, which will help distribute or spread out the delivery of services, supporting the healthcare workforce and decentralising care.
For patients:
Greater convenience and access, and the ability to gain greater control over their health. As digital health solutions become widespread, it is likely that patients will be able to benefit from increased adoption.
The ability to self-manage their conditions as data is fed back to them. Also the opportunity to support patients dealing with their diagnosis and reduce suffering from the side-effects of cancer treatment by being able to monitor patients more dynamically, intervene earlier and empower them with ways to understand and apply that data.
More aspects of the care journey will be delivered from the comfort of patients’ homes, including services or procedures that can be conducted remotely or through home visits.
Access to 24/7 cancer support. In future, these technologies will be able to detect changes in the patient’s biology, before they even realise and will flag them to the patient and their clinicians.
For healthcare systems:
Improved co-ordination among healthcare providers, patients and insurers.
Real-time data-sharing and communication will streamline claims processes, reduce administrative tasks and ensure patients receive appropriate care faster.
Integrated patient data in the same platform, such as a centralised personal health record, giving patients control over their data and whom they allow access to it.
Better tailored insurance products, as insurers can leverage the data gathered from
digital oncology tools and technologies to make informed decisions about coverage, pricing and risk assessments.
What next?
In future, digital oncology solutions will become standard practice in cancer care.
As they continue to mature, their proficiency in processing and analysing diverse types of data, including data from wearable devices, will increase.
The integration of a diverse range of digital oncology AI-driven solutions will enhance pro-active risk management in cancer care.
Remote monitoring, screening and early diagnosis of cancer will increasingly involve novel connected devices that leverage AI.
Cutting-edge solutions on the horizon include:
A new test using infrared spectroscopy as a non-invasive and low-cost tool for breast cancer screening.
This test, which uses small liquid plasma samples, could be used as an alternative to mammography.
Advanced ambient intelligence may also be used to monitor patients remotely in their homes and alert professionals of unexpected changes.
This technology makes a person’s surroundings ‘smart’ by using sensors and data to understand and monitor individuals and their environment.
Artificial intelligence in cancer care
Artificial intelligence (AI) will play an important role in the future of oncology.
It will be widely used in various forms across the entire healthcare journey from risk prediction and stratification to diagnosis and helping to inform treatment regimes.
Increasingly over the next five years, AI-driven cancer risk assessments will be comprehensive and dynamic, considering vast amounts of data, including genomic data, to provide personalised risk profiles.
Regular health check-ups that incorporate AI-powered risk assessment will allow for the early detection of potential cancer risk. And people will also receive personalised recommendations for lifestyle modifications to reduce their cancer risk.
As these AI-driven assessments become more sophisticated, they will also be able to provide real-time feedback on how lifestyle factors are impacting cancer risk.
Better understanding of a people’s risk will allow healthcare providers to stratify them and offer tailored screening and early, more personalised interventions to those at the highest risk with the aim of improving cancer outcomes.
This will lead to a more personalised experience and increased satisfaction.
A comprehensive understanding of people’s risk also enables the design of personalised insurance plans that reflect individual risk and better suit their needs.
Types of AI
AI is a broad concept, usually referring to the development of computer systems which can perform tasks that typically require human intelligence.
It encompasses any technique that enables computers to mimic human intelligence, using logic, if-then rules, decision trees and so on.
Subsets of AI are:
Machine learning (ML) which includes complex statistical techniques that enable machines to improve at tasks with experience.
Deep learning (DL) which is a subset of ML composed of algorithms that allow software to train itself to perform tasks like speech and image recognition by exposing multi-layered neural networks to vast amounts of data.
Generative AI (GAI) which is a subset of DL that can produce new content such as text, code and pictures based on the information it is given as input.
Uses of AI
One of the most well-established applications of AI in oncology is the interpretation of medical imagery findings. However, it can support at many different stages of the patient journey.
For example, AI is also being used to help predict people at high-risk of getting cancer and stratifying patients into risk categories.
According to a survey of more than 1,000 cancer experts,1 AI-related developments will have a significant impact in grading and classifying tumours, and providing more reliable diagnoses within the next decade. The survey respondents highly favoured AI applications related to early cancer detection and diagnosis.
Here are some examples of AI-based tools that may support different stages of the cancer patient journey.
Risk prediction and stratification – identifying high-risk individuals: A new AI tool is being developed to predict a woman’s ten-year combined risk of developing and then dying from breast cancer.2 Despite promise, further validation is needed before the model is implemented in clinical practice.
Personalised screening based on risk level: Tempo3 is an AI tool for personalised cancer screening that assesses individual risk profiles and recommends the timing for future mammograms based on the patient’s history and evolving risk.
It could improve early detection while reducing the number of unnecessary mammograms. Scientists are currently working to further refine their model.
Delivering personalised prevention: Google and DeepMind have developed Med-PaLM M, a multi-modal large language model able to interpret and understand connections across different forms of patient data including clinical text, images and genomic data.4
Med-PaLM M outperformed other AI models and was able to diagnose tuberculosis in chest X-rays without being trained with such data previously.
The model is now available to a select group of Google Cloud customers for testing and feedback.
Timely and efficient diagnosis: An AI-based algorithm that analyses chest X-rays to detect potential lung nodules and masses has received FDA 510(k) clearance.5 This tool can help enhance patient care and support the healthcare workforce by providing a second read on X-rays.
Delivering tailored treatment: Researchers at the University of Sussex are developing a new AI prediction tool to assess genetic changes in tumours.6 This may potentially allow oncologists to personalise cancer treatments for patients.
Predicting treatment outcomes and prognosis: Digistain is an AI tool that uses an optical scan of tumour tissue to determine the likelihood of breast cancer recurrence and so inform treatment decisions.
It is gaining adoption worldwide, including in the UK, India and the US. Research concluded that this tool would present huge cost savings for the NHS and reduce environmental impact by 460 tonnes of carbon.
Although promising, AI models are not currently widely implemented in oncology care and are often used as stand-alone activities, so lack integration across other relevant digital systems.
There is also a need for a more experienced workforce to support the appropriate implementation of these tools into clinical practice.
Additionally, the use of these algorithms involves significant ethical, risk and privacy issues.
What next?
AI will play an important role in creating personalised prevention plans, which may include recommendations for lifestyle modifications and interventions to reduce risk factors or preventive medications.
Within the next few years, multi-modal AI models will be able to analyse and consider different types of patient data, including medical records, imaging, genetic, lifestyle and environmental data to determine their impact on cancer risk and the likelihood of customers developing cancer.
The timing and frequency of cancer screening will be optimised based on a person’s risk profile. This will help avoid over-screening low-risk individuals while ensuring high-risk individuals receive appropriate and timely screening.
As well as improving health outcomes, this approach will also optimise use of medical resources.
New drugs
AI also has the potential to significantly influence the discovery of new cancer drugs by detecting trends in data – for example, clinical trials, genomics – and suggest promising drug candidates, predict the efficacy of novel treatments and offer insights into overcoming treatment resistance.
It may also accelerate the process of screening existing drugs for repurposing in cancer treatment, potentially uncovering new applications. This not only speeds up the drug discovery process but also improves the chances of finding more effective and personalised cancer therapies.
Conclusion
Technology and improvements in digital infrastructure are set to revolutionise oncology, supporting a connected care model and making high-quality cancer care increasingly accessible to many more people.
This has the potential to speed up screening, improve diagnosis accuracy, enable personalised treatment and improve patient outcomes.
As digital tools and solutions continue to be introduced, more sources of data will be available and shared in real-time with clinicians and healthcare providers.
This increased volume of data will enable AI-based algorithms to identify irregular patterns, paving the way for pro-active cancer care.
Dr Tim Woodman (right) is medical director, policy and cancer services, Bupa UK Insurance