Google’s AI Beats Doctors at Breast Cancer Detection | TechTree.com

Google’s AI Beats Doctors at Breast Cancer Detection

When they acquired DeepMind in 2014, the obvious intention was to tick the box associated with artificial intelligence and create a sort of cybernetic friend

 

When Larry Page reportedly brushed aside criticism and competition from Facebook to gobble up UK-based artificial intelligence (AI) start-up DeepMind in 2014 for $500 million, sceptics said they were merely checking the boxes in spite of lofty statements around how the cutting-edge expertise would help them build general-purpose learning algorithms for simulations, e-commerce and gaming.

However, five years into the deal, DeepMind has made news of a completely different kind. It has developed an AI model that can identify breast cancer from scans with fewer false positives or false negatives compared to when conducted by medical experts. Yes! You raid it right… the machine and algorithms did better than doctors when it came to detecting breast cancer.

A report published on the BBC website said an international team comprising researchers from Google Health and the Imperial College of London had designed and trained a computer model on X-ray machines from nearly 29,000 women with the algorithm outperforming six radiologists in reading mammograms suggesting that the future of diagnosis could well belong to machines.

Even though breast cancer is amongst the most common type among women, early detection has proven to be difficult due to the high rates of false positives which essentially means that the mammogram is perceived as abnormal even when no cancer is present. Such instances lead to unwanted medical interventions and cause untold misery to the patient and her family.

The research was conducted on patients across the UK and the United States and according to the results false positives reduced by 5.7% and false negatives by 9.4% in the latter geography. So, what did the machine do differently? Unlike doctors, who use patient histories and past mammograms to assess prevalence of cancer, the AI only used the most recent mammogram of each patient. In spite of this, it screened decisions with greater accuracy than what the radiologists could.

However, the research team has clarified that the AI project is still in its early stages and only more such studies and collaboration with healthcare providers could bring forth how effective and efficient it could be in breast cancer detection. There was also no comment about whether anything similar could be done with other forms of cancer.

In the past DeepMind has been used to identify ailments related to the eyes and kidneys though there have been debates over their efficacy compared to human intervention. Of course, the privacy issue had reared its head several times including in 2017 when media reports claimed that the UK health department had shared data without adequate legal safeguards on consent and privacy.

The BBC report further quoted Dominic King of Google Health to suggest that the company was proud of these research findings that indicated the value that AI-based systems can bring to medical diagnosis. Since the process of reading mammograms and coming to the right conclusions is a time-consuming one, the experts feel that if the AI project does reduce errors, it would augur well for patients who can be put on treatment sooner.


TAGS: Breast cancer, Google Health, DeepMind, Research

 
News Corner
IMP IMP IMP
##