Google just announced the forthcoming end of their personally controlled health care record Google Health due to lack of widespread adoption. We have been working with the service since its launch in 2008 and used it as a back-end for various projects (most recently for a medication reconciliation tool tied to the SMART platform). Google Health was offering a very decent user interface, a fair amount of API support and some really good interfacing ideas; I hate to see this one go.
[UPDATE] You might deduct how important the project was in the end from the fact that it was not even worth a shutdown announcement for itself.
I gave the following presentation during a set of BMI related talks; it covers the cornerstones of cloud computing development and explores potential application fields in BMI.
It was created using Prezi, a neat little tool that allows you to break out of the power point cage. I am not entirely familiar with it (you will see some fairly uninspired slide clones in there), but the response of the audience was overall very positive, so I would recommend it.
A typical problem when benchmarking a clinical system is establishing a ground truth. Let us assume a clinical support system that supports a physician in interpreting radiology images (e.g., by recognizing tumors). The only intuitive method we have to evaluate the performance of the system is to test it on a set of labeled images, or in other words, through asking questions we already know the answers to. However, creating this ground truth will require us to rely on an analysis through the very process we attempt to improve, namely the “manual” analysis by a physician.
The scenario can be frequently observed wherever we try to recognize patterns. Another example is research oriented extracting of knowledge from patient records, where we would attempt to recognize adverse drug effects or develop best practices. A note might be indicating the negative impact on the patients health, although it is not recognized as such be the medical expert that is the referee for the benchmark due complexity, illegibility or counterintuitive nature.
There are methods to damp the effect, such as increasing the number and competence of the referees or implementing a round of reconsideration of results (e.g., the physicians can be confronted another time with images that have been recorded as false positives during the automated tumor detection), but those methods are often expensive and time consuming or, in the worst case, just not available.
Therefore we have to keep in mind that when dealing with a highly complicated and often intuition driven field such as medicine we have to constantly account for possible human errors, and that there is always the possibility that we have our job well enough to outperform the quality of the human decision. Or to coin it less optimistically: sometimes even great results can become a problem.
When developing biomedical systems that have to be placed into clinical care we often times have to face the enormous challenge of either giving a new face to existing and familiar processes or even restructure and replace them completely with new applications. In this context, building user interfaces for an discipline that is inherently sensitive to time constraints requires careful thought, skilled design, and a profound understanding of how health care personal can be supported in their daily work. While some of our past approaches did not exactly fuel the enthusiasm for computer support at the bedside (e.g., awkward data input via overlayed keyboards), technology is constantly moving forward and the progressing availability of mobile devices (especially those equipped with reasonably functional touchscreens) allows us today more than ever to create applications that are intuitive, safe and rapid in executing everyday tasks.
For health care professionals, however, taking steps towards health information technologies, such as implementing an electronic health record in a physicians practice, will always include change that requires an open mind, professional training, costs, and a certain level of discomfort in the transition time. From a software engineering point of view, our responsibility is to make this transition as easy and efficient as possible, but eliminating its downsides completely is an impossible task. Therefore, going the whole distance to highly available, safe, and cost-efficient health care through structures like national health networks will also require policy makers to stimulate progress and, where necessary, to enforce unpleasant steps if they lead to benefit for patients.
An interesting but also discomforting thought that came up last during a discussion regarding Biomedical Informatics: Our very computational idea of evidence based medicine is with its maybe ten years of of presence still a fairly new concept to the field. For decades, the procedures applied to patients were largely based on experience, intuition and a somewhat vague and very encapsulated “it worked quite well last time”-approach rather than on globally proven facts.
We only just started to imagine and implement systems for collecting health care information and delivering it back to health care professionals in forms that empower them to optimal decision making. Keeping in mind how young this concept of Health Information Exchange is, we can only assume that there is certainly still room for massive improvement.
Today the UConn CSE department was hosting a guest lecture by Dr. Vladimir Vapnik who was presenting Learning using privileged information (LUPI), an extension to his popular support vector machine (SVM) method. While the underlying Math is rather uninviting, the concept itself is stunningly convincing. It it based around the the idea that the training phase for a learning problem can be supplemented and improved with additional data from a privileged information space. Intuitively, this can be described as similar to a teacher giving a student subjective explanations on top of the text book examples.
The surprising point is to see that this information does not necessarily make any sense in the actual problem space. A provoking example Vapnik used during his talk was an experiment where poems describing the graphically representation of 8 and 5 were used to improve the pattern recognition of those digits. Although those informal descriptions appear to be nonsense when one looks at the graphically oriented problem, the LUPI extension of SVM is actually able to leverage them to improve the training phase of the recognition problem.
Details and several other examples related to biomedical informatics and statistics can be found in the related publication. I recommend at least skipping through it.