Menu

Comment: does AI threaten the biomedical workforce?

Comment: does AI threaten the biomedical workforce?
7 November 2019
Dr Perry Maxwell says artificial intelligence (AI) can aid diagnosis, but biomedical scientists are still essential

As more healthcare laboratories turn to digital tools, artificial intelligence is seen as having vast potential for the future of healthcare. Just last year, St James’s Teaching hospital in Leeds was the first pathology laboratory in the world to go 100% digital

However, there are concerns in the professional community as to how AI may affect the biomedical workforce, which were raised at Dr Perry Maxwell’s talk on digital cytology at IBMS Congress 2019.

Dr Maxwell (pictured above) is Clinical Lead in Tissue Hybridisation and Digital Pathology at the Precision Medicine Centre of Excellence and an Honorary Senior Lecturer at Queen’s University Belfast. He commented:

“At IBMS Congress in Birmingham, I presented on ‘Digital cytology: it’s all about the sample’. The Topol report on the digital future of healthcare identified several key technologies including artificial intelligence.  We have described AI as representing a third revolution in pathology following on from the introduction of immunohistochemistry and genomic testing1
AI, in the form of deep learning neural networks, requires digitisation of laboratory services and digitalisation of pathology slides and reporting workflows. In my presentation, I looked at what implications these have for cytology and look at how AI can affect and contribute to cytology practice. 
One concern expressed on the day was whether AI would put at risk future employment of staff.  To address this, let us look at a recent article ‘Digital assistants aid disease diagnosis’ by Neil Savage in the journal, Nature.
Savage highlighted two deep learning tools, which show promise in helping the radiologist or pathologist in their work. For the radiologist, it was the promise of highlighting cases from their caseload, which needed urgent attention; for the pathologist, it was the promise of helping to identify aggressive types of lung cancer.  The take-home message from these reports and others2 is that such tools are performing to a level equivalent to the medical practitioner. 
Medical practitioners and the public at large may see such tools as a threat to their jobs but Savage highlights the misconception of how these neural networks may outperform humans.  As Savage reports, it is in the integration of multiple datasets such as those from ‘pathology, radiology, genomics, electronic health records and even lifestyle data’ where such tools can outperform humans, not single tests. In pathology, further advances may be made in the quantitation of predictive biomarkers e.g. immune checkpoint biomarkers such as PD-L1, which are used to predict response to specified immunotherapies. 
A limiting factor in the development of deep learning tools is the need for robust labelled data, in the case of CT scans and radiology, the need for greater than the 96,000 cranial CT scans the radiologist used as cited by Savage.  Real world data is required to train the deep learning networks, which the UK government has recognised for diagnostic test use with the release of projects for digitalisation and AI development under the Innovate UK Industrial Strategy Challenge Fund. 
An example of this is the £14 million PathLAKE consortium of Warwick, Coventry, Belfast, Oxford and Nottingham in partnership with Philips and others to develop a robust data-lake of both images and clinical metadata for industry to use in training AI tools.
As Savage reports, ‘for now, however, AI systems will be there to help doctors to make a diagnosis, not to put them out of a job’.”

For a wider discussion of AI in pathology, see the Biomedical Scientist magazine’s article, ‘The rise of the robots’.

 

References

  1. Salto-Tellez M, Maxwell P, Hamilton P.  Artificial Intelligence – the third revolution in pathology.  Histopathology 2019;74:372-376 http://dx.doi.org/10.1111/his.13760
  2. Liu X, Faes L, Kale AU et al.  A comparison of deep learning performance against health-care professionals in detecting diseases from medical imaging: a systematic review and meta-analsyis. Lancet Digital Health. Published online September 24, 2019 https://doi.org/10.1016/S2589-7500(19)30123-2
  3. Savage, Neil. 2019. "Digital Assistants Aid Disease Diagnosis". Nature.Com. https://www.nature.com/articles/d41586-019-02870-4.
  4. 2018. “The Rise of the Robots”. The Biomedical Scientist. https://thebiomedicalscientist.net/science/rise-robots.
 
Back to news listing