Look for Watson IBM. More information: 

http://jop.ascopubs.org/content/9/3/155.full

http://www.ibm.com/smarterplanet/us/en/ibmwatson/implement-watson.html

http://www.aaai.org/ojs/index.php/aimagazine/article/view/2303/2165

Look for Watson IBM. More information: 

http://jop.ascopubs.org/content/9/3/155.full

http://www.ibm.com/smarterplanet/us/en/ibmwatson/implement-watson.html

http://www.aaai.org/ojs/index.php/aimagazine/article/view/2303/2165

Thank you Bruno for your response. I will check those out!

From legal (and ethical) point I think you can look also in area of drones robots and decision making. What are the implications as we as humans or not able anymore  to judge the outcome of analyses and decisions?

have o look at http://blogs.law.stanford.edu/robotics/

I think the work of prof. Van der Gaag may be interesting for you. She uses Bayesian networks as a decision support tool for medical diagnosis. For example:

"E.M. Helsper, L.C. van der Gaag. Ontologies for probabilistic networks: a case study in the oesophageal-cancer domain. The Knowledge Engineering Review, vol. 22, pp. 67–86, 2007." and "P.L. Geenen, L.C. van der Gaag. Developing a Bayesian network for clinical diagnosis in veterinary medicine: from the individual to the herd. Proceedings of the Third Bayesian Modelling Applications Workshop, held in conjunction with the Twenty-first Conference on Uncertainty in Artificial Intelligence, Edinburgh, 2005."

Also Dan Lizotte's work may be interesting. He does a lot with Markov Decision Processes in medical applications: http://www.csd.uwo.ca/~dlizotte/publications/

e-HPA, or electronic-Healthcare Predictive Analytics now encompasses diagnostic (Dx) technologies (both inferential and connectionist models), prognostic (Px) technologies, and the suggestion of optimal treatment (Tx) interventions.  In the mid-1990s, John Pollock and I wrote a paper (in proceedings of AAAI Conference - link on my info page) about an emergency toxicology system we build that used defeasible reasoning.  Things have progressed considerably since then, and models have changed considerably in two ways (1) they incorporate not only Dx, but also Px and Tx, with a view towards using data-driven outcome predictions (prognostic analytics) to drive optimal treatment.  Dx is less important now than it used to be, as arguably Dx is irrelevant if Px can be optimized with best Tx.  In other words, if you know (probabilistically) that your prognosis is optimal, and that your treatment is optimal, the diagnosis and etiology of the disease is certainly interesting, but in the extreme case, only the outcome matters.  Assuming the technology works, that is - it optimizes the Px by suggesting the optimal Tx, Dx doesn't matter as much (or at all, again in the hypothetical extreme case).  This is a shift towards outcome based medicine.  (2) There is much more data available now to drive these models - the Ai that I work on uses predictive analytics for both Px and Tx, but not Dx (Dx is a classification, not predictive, problem) and a hybrid model of time-series SVMs to create percepts that are analyzed by a defeasible reasoner.  In any case, it is much more useful to build clinical decision support systems based on AI that looks at all three - Dx, Px, and Tx - as the ultimate goal is to optimize Px.  Sometimes you can't diagnose something, or understand its etiology, but nevertheless you can predict its prognosis, and suggest interventions to improve that prognosis, knowing NOTHING about the diagnosis or the etiology.  I think that in the medical informatics field, a seismic shift is occurring towards OUTCOME-BASED medicine as an evolution of evidence-based medicine.  For a good review of the ethical implications of this approach, see the article IMPLEMENTING ELECTRONIC HEALTH CARE PREDICTIVE ANALYTICS, at: http://content.healthaffairs.org/content/33/7/1148.full.html 

Full Cite:

Ruben Amarasingham, Rachel E. Patzer, Marco Huesch, Nam Q. Nguyen and Bin Xie Implementing Electronic Health Care Predictive Analytics: Considerations And

Challenges

Health Affairs, 33, no.7 (2014):1148-1154

doi: 10.1377/hlthaff.2014.0352

 Thank you all for your answers. They are all very helpful!

https://www.google.com/search?q=artificially+intelligent+machines+that+aid+in+medical+diagnoses&num=50&source=lnms&tbm=nws&sa=X&ei=2lw3Vc76Hoq8sAW73oHwCg&ved=0CAcQ_AUoAQ&biw=1375&bih=725

you can use Fuzzy cognitive maps for medical diagnosis. there are many articles on fuzzy cognitive maps for mdss