According to an international study, artificial intelligence is better at detecting breast cancer than expert radiologists. Not only did the AI system detect more cancers in mammograms, but also ignored features the radiologists had incorrectly identified as possible tumours.
A team of researchers from Google health and the University of London trained the machine learning technology on anonymous mammograms of more than 76,000 women in the UK and more than 15,000 women in the US. The hope is that AI could soon improve the effectiveness and accuracy of breast screening, but more clinical trials will be needed first.
The AI system has already outperformed six radiologists. To put that in perspective, it takes a medical student 10 years to train and qualify as a doctor capable of interpreting mammographs.
Inaccuracies, whether false negatives or false positives, can lead to delays in detection and treatment, needless stress and invasive biopsies. Around one in eight women are diagnosed with breast cancer in their lifetime.
International standards for AI and big data are developed by ISO/IEC JTC 1/SC42, the Subcommittee on Artificial intelligence, of the ISO/IEC Joint Technical Committee on Information Technology. These standards will support the work of IEC TC 62: Electrical equipment in medical practice and that of other IEC TCs.
Most regions face significant healthcare challenges. These include aging populations, lack of physicians or inadequate healthcare infrastructure.
AI can provide important benefits in some domains, such as research, diagnosis, finding the most effective treatments, telemedicine, or developing entirely new and better drugs, to achieve what is known now as personalized or precision medicine.
To be of any use in any domain, AI needs to process and interpret extensive data sets to draw relevant conclusions. Agreements and safeguards permitting the exchange of healthcare information at country and international levels, provide an unequalled volume of data.
Since the use of AI in healthcare means access to a large volume of data about individuals, the data must remain private and be protected from cyber threats.
This can be achieved, to a certain extent, though the anonymization of data, making it impossible to trace back information to individuals or clusters of individuals.
This is first and foremost the responsibility of healthcare professionals and service providers using best practices and, among others, standards developed by ISO/IEC JTC 1/SC 27: Information security, cyber security and privacy protection.