Machine learning finds an improved way to match donor organs with patients

As someone doing ML for a living this development scares me. Trying policy to de-bias dataset against prior policy (which is changing with time by the way) is very difficult. And any new policy will effect the outcomes and new datasets going forward. Positive feedback loops in Reinforcement Learning is a huge problem.

Improvements in organ donation&#x2F;procurement networks in the USA are needed. &quot;In 2020, 21.3 percent of procured kidneys were not transplanted,&quot; per <a href="https:&#x2F;&#x2F;www.washingtonpost.com&#x2F;health&#x2F;2022&#x2F;08&#x2F;03&#x2F;transplant-deaths-mistakes-senate-finance&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.washingtonpost.com&#x2F;health&#x2F;2022&#x2F;08&#x2F;03&#x2F;transplant-...</a> and <a href="https:&#x2F;&#x2F;onlinelibrary.wiley.com&#x2F;doi&#x2F;epdf&#x2F;10.1111&#x2F;ajt.16982" rel="nofollow">https:&#x2F;&#x2F;onlinelibrary.wiley.com&#x2F;doi&#x2F;epdf&#x2F;10.1111&#x2F;ajt.16982</a> We should not let an unknowable and unachievable &quot;perfect&quot; be the enemy of better.

Around 90% of liver disease deaths is from preventable risk factors (alcohol, viral hepatitis and obesity, <a href="https:&#x2F;&#x2F;ukhsa.blog.gov.uk&#x2F;2014&#x2F;09&#x2F;29&#x2F;liver-disease-a-preventable-killer-of-young-adults&#x2F;" rel="nofollow">https:&#x2F;&#x2F;ukhsa.blog.gov.uk&#x2F;2014&#x2F;09&#x2F;29&#x2F;liver-disease-a-prevent...</a> ). We have the tools to fix to these (the main being more even distribution of resources in society) yet we don&#x27;t.

&quot;Sorry, our internal black box system determined you are not a candidate at this time. Please try again later.&quot;<p>Aside from reinforcement loops being an issue, ML is only as good as the programmer makes it. And we will have no idea how it was constructed and neither will non-tech health professionals. Theres no way a system like this can be described in its entirety without mistake to surgeons when one line of code changes everything.<p>So maybe it can help out human decision but I would not want this to have final say for quite some time.

needs a [2020] tag. Mihaela van der Schaar&#x27;s research group is highly productive, I&#x27;d be surprised if they hadn&#x27;t published some update to that methodology by now

In leu of current events - the first thing that crossed my mind is what the CCP will do with this technology.<p>But it&#x27;s good to see more real world application of machine learning - feels like the distance from research to application is very slow.<p>&quot;Machine learning radiology&quot; is something I used to monitor as a sign of improvement - since it&#x27;s a &quot;simple&quot; use case. Too &quot;busy&quot; to google about now - any news on this ?

We could revolutionize not only organ transplants, but all of human health though advances in human cloning.<p>The biggest increase to human health span will be when we start replacing our bodies wholesale as they age.<p>Head transplants onto monoclonal, HLA-non-expressing (antigen free) bodies at age 50 will renew cardiovascular, pulmonary, and immune health (eg new thymuses). It resets almost all of the genetic cancer clocks in our bodies. Coupled with drugs to combat brain tissue waste aggregation and plaque formation, it could see humans through to multiple hundred years life spans.<p>As for ethics, the monoclonal bodies will be grown in a lab, without brains, born out of chimeric pig uteruses and artificially innervated for endocrine regulation.<p>I talk about this a lot in my HN post history.<p><a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=26673243" rel="nofollow">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=26673243</a><p><a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=28534681" rel="nofollow">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=28534681</a><p><a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=30622820" rel="nofollow">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=30622820</a><p><a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=30407908" rel="nofollow">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=30407908</a>

Seems like a difficult problem. Do you let 15 people have a subpar match or give 14 a perfect match and 1 person dies.