Where to find a solid database to start digging....
UnicornDear RM friends,
The Biomed and Pharma industry are considered the slow adopters for analytics implementation but as I see initiatives grow every day, I think the time is right to get your focus on this industry. The Pharma industry is taking the lead with "Precision Medicine" as joined effort to combine the best medical domain knowledge with the most accurate analytics to ultimately give the most individualized treatment a patient deserves. For this effort, the Pharma industry is supported by the Biomed industry as far it relates to molecular and genetic analysis. The Biomed industry also fell in love with imaging such as MRI scans with companies as IBM etc attempting to find cancers faster and better than any radiologist in the country. The knowledge hidden in electronic health records is the next and perhaps most difficult attempt to get the medical reasoning out of the middle ages. Privacy regulations have limited access to this data for decades not to mention any protectionism in the field remembering the death of the Google health initiative in 2011 (https://en.wikipedia.org/wiki/Google_Health).
Luckily some research groups have found the power of datathons to speed up the introduction of analytics in the field. Keep in mind medical records are stuffed with multivariate time series and tons of unstructured data. A succesful initiative with accessible database is managed by the MIT (https://mimic.physionet.org/).
I wanted to share with you their latest report on the datathon in Beijing: http://medinform.jmir.org/2017/4/e43/.
Good luck.
Sven
Answers
Hi RM friends,
Also blockchain is entering the BIOMED and PHARM industry: A colleague of mine published a nice paper on this topic: Converging blockchain and next-generation artificial intelligence technologies to decentralize and accelerate biomedical research and healthcare
Cheers
Sven
One of the things I like about ML in healthcare is at the moment if you have an XRay the radiologist is generally looking specifically for signs of pneumonia when examining your chart. Due to the patient symptoms there is a little bias on what they are looking for & other conditions might be missed.
Once a large enough group of models are created it can help to highlight potential conditions that the physician might not have noticed previously because they might not have been looking due to the patient displaying no prior symptoms.
Particularly as the ML process might also be able to access the patient EHR going back and, although not be correct all the time, could act as an early alert system for the doctor to explore further.
Hi RM community,
A recent example in this context: https://arxiv.org/pdf/1711.05225.pdf
Cheers
Sven