Artificial Intelligence can shorten cancer radiation treatment duration

Physicians spend between 25 minutes and two hours outlining bones and organs on approximately 100 scan cross-sections for each patient. Researchers say the AI programme operates 1.5 times faster. (Photo: Julien Tromeur via Unsplash)

AI has been used in radiology to analyse and locate tumours on scans to improve cancer screening. AI-based solutions can assist pathologists in making more accurate and consistent cancer diagnoses, lowering case-error rates. Predictive AI models can assess a person's likelihood of developing cancer by identifying risk factors.

At Addenbrooke's Hospital in Cambridge, the National Health Service can now plan radiotherapy treatments more quickly, reducing patient wait times. It assists doctors in determining where to direct therapeutic radiation beams to target malignant cells while sparing as many healthy cells as feasible. Researchers at Addenbrooke's Hospital worked with Microsoft to train the AI programme.

Typically, physicians spend between 25 minutes and two hours "contouring" or outlining bones and organs on approximately 100 scan cross-sections for each patient. However, researchers say the AI programme operates 1.5 times faster.

For example, physicians want to avoid injuring the adjacent bladder or rectum when treating the prostate organ, as this could leave patients with permanent continence problems.

How is AI used in radiotherapy?

AI can be used to supplement typical traditional radiography techniques. (Photo: Craig Cameron via Unsplash)

Specialists can arrange radiotherapy treatments around 1.5 times faster with the help of this AI technology, meaning more patients can undergo treatment sooner, and the possibility of better results is increased.

AI could revolutionise radiation oncology. However, massive curated datasets — frequently incorporating imaging data and accompanying annotations — are necessary to create radiation oncology AI models. According to UT Southwest research, employing AI allows cancer patients to begin therapy sooner since AI can use complex data to design an optimal treatment plan. AI can also develop treatment procedures and dosage calculations and quickly recalculations to change the treatment plan.

The Project InnerEye

Dr Jena's research includes collaboration with Microsoft Research on Project InnerEye, a Microsoft research effort developing Machine Learning and open-source software that enables healthcare organisations and innovators to build solutions to assist clinicians in radiotherapy treatment planning.

The Project InnerEye team conducted peer-reviewed research with eight clinical centres worldwide, demonstrating that clinicians using machine learning assistance can segment images significantly faster than manually, with accuracy within the bounds of human expert variability. To increase access to this study, Microsoft made Project InnerEye toolkits accessible as open-source software in September 2020.

OSAIRIS AI — the new AI system

Dr Jena's team developed OSAIRIS, a novel AI system employing open-source software technologies from Project InnerEye and Azure Machine Learning, with the help of £500,000 funding from the NHS AI Lab. OSAIRIS has undergone rigorous testing and risk assessments to ensure that it can be used in the day-to-day care of radiation patients across the NHS. Doctors could not discern the difference between OSAIRIS' work and the work of a doctor colleague in masked examinations known as "Turing tests." However, the oncologist maintains control throughout.

How do radiologists use AI?

AI can help automate operational duties such as evaluating imaging appropriateness, patient scheduling, selecting examination procedures, and improving radiologists' reporting workflow.

However, there are some obstacles that radiologists should be aware of before integrating AI into regular practice. Regulatory compliance, ethical issues, data protection, cybersecurity, AI training bias, and the safe incorporation of AI into common practice are all barriers.

What are the limitations of AI in oncology?

AI uses computer programmes that analyse massive volumes of data to learn how to make judgements or predictions. AI may aid in cancer screening, diagnosis, and therapy planning in medicine. It may also be employed in medication discovery and development studies. Current AI applications in radiology often evaluate the likelihood of problems for a given patient based on radiological imaging. An AI system may determine, for example, that a particular patient's breast lesion has a 10 per cent chance of being cancerous.

An AI system may determine if a particular patient's breast lesion has a 10 percent chance of being cancerous. (Photo: Victoria Strukovskaya via Unsplash)

The need for varied and all-inclusive data sets for training is one of the main things that keeps AI algorithms and CDSS from being used more widely in cancer care. When these models are implemented and employed, they should be used on the same kind of patients from which the training data came.

Will AI replace radiographers?

AI can be used to supplement typical traditional radiography techniques. It can enable appropriate patient positioning within the gantry and automate image processing. AI will soon be a part of radiologists' daily practice, assisting physicians in improving efficiency and diagnostic capacity. AI can sift through massive amounts of imaging data in seconds, enabling radiologists to assist with worklists and diagnosis prioritisation.

Although AI poses no immediate threat to radiologists, their jobs will shift as technology improves. The question is not whether AI will replace radiologists but whether those who accept and learn from technological improvements will replace those who do not.