I work in the areas of machine learning, unsupervised machine learning and applications, looking for potential collaboration opportunities in Canada, North America and Europe for several Covid-19 related research and potentially, development projects:
- Applications of unsupervised machine learning in epidemiological data analysis to identify characteristic clusters in distributions of epidemiological data.
- Time-period and geographical cohort study of Covid-19 / BCG correlation in some jurisdictions in NA and Europe.
- Integrated national / regional Covid-19 diagnostics and tracking framework.
- Unsupervised machine learning in Covid-19 genome analysis.
Would appreciate any interest, comments and discussion.
These are a few thoughts to ponder...
Most modelling work in the field of ML/AI on health data are being done by non-clinical teams. Such models will not be applicable to clinical practice. The topics such as posted above must be communicated to clinicians and getting their input. It feels to me that the primary motivation of the above topics choice is just a race for publications, and presenting to the conferences. The first question data scientists need to ask is " So what?". If you think your model is better in identifying say, COVID19 cluster characteristics distributions compared to the earlier developed models, how is this going to help the patients and clinicians? How is this going to improve clinical practice? How is this relevant at all? It looks like that COVID pandemic is a real test to data scientists’ sense of purpose. Can data scientists use their knowledge in a practical way and help to reducing the surge of sick patients? Or one can write a beautiful theoretical paper (as posted above in the discussion topics) that will never get attention of the clinicians.
Until data scientists understand that practical value of their work is not only about algorithms, the insane race for publications will continue. In order to become relevant a partnership is required with those who understand the real meaning of dealing with coronavirus or any other possible clinical outbreak. I wish that we see less topics about COVID-19 that focus only on formal data analysis and algorithms but many more that include the proper clinical motivation and support.
Alexander Kolker, indeed if a result is genuine, it should give us a better understanding of some aspect(s) of the problem and that is already a step, an advance that can have some positive applications. More specifically, the analysis of case distributions, identifying the regions of parameters, latent and observable that can be associated with certain outcomes can be valuable for example, in proactive audit of public facilities and policies providing inputs that can be used to avoid more severe outcomes.
Certainly, multidisciplinary teams comprising specialists in medical and information domains can have clear advantages; nevertheless, there isn't to my view, anything extraordinary in the medical information per se that would prevent its analysis by statistical and other information processing methods.
- Badges
- Science topic
- Similar topics
- Comment