Planning before surgery
Imaging plays a phenomenal role in planning for any major breast operation—whether it be from pinpointing the location of a breast tumour to helping a surgeon navigate complex breast anatomy. Radiologists now have access to swarms of imaging data to aid the former, thanks in part due to modern imaging techniques. Nonetheless, it can be time-consuming to process this information before a surgery is planned.
New machine learning technology aims to bring greater efficiency and accuracy to this process. Initial trials suggest that machine learning performs to the same level if not better than a radiologist in detecting cancer, and also shows a higher sensitivity (i.e. a better ability to detect cancer in an individual that actually has cancer). Not only will this provide surgeons with prerequisite knowledge to make smarter treatment decisions but will lead to reduced workloads, a reduced burden on resources, and reduced chance of error.
Making long-term predictions
Clinicians always want to ensure what they do is backed up by strong evidence—one of the reasons why so much time is spent applying traditional statistical ideas to monitor and predict what might happen to patients after their breast surgery operations.
Whilst a relatively new phenomenon, machine learning looks promising as a gamechanger within this field. New trials suggest it predicts five-year mortality after breast cancer operations more accurately than statistical models and have even gone on to suggest it can predict a patient’s chance of developing a complication like lymphoedema (a long-term swelling in the tissues of the body after an operation).
All of this has been down to the creation of a specific type of machine learning dubbed an “artificial neural network”—a type of machine learning modelled after a human brain cell called a neurone.