Health Data, Self-serve, Visualization, Semantic Analysis and Collective Intelligence

Notes from the Biomedical Data Mining Camp Session led by Dr. Irene Gabashvili and other health-related discussions at Data Mining Camp and Transparency Camp 2010.

The title for the session was "Biomedical Data Mining: Successes, Failures, and Challenges" (streamed online from Fireside C).
The topic stemmed from the similarly named last year's session - Biomedical Data Mining: Dimensionality, Noise, Applications - now split into several discussions including Bioinformatics & Genome Sequencing organized by Raymond McCauley, Dimensionality Reduction moderated by Luca Rigazio (HLDA / HDA; LDPP; Core Vector Machines; Sparse Proj SQ; Random Projection and Feature Selection).

Main sub-topics of the biomedical session were:

• Reality Mining
• Visualization
• Imaging
• Signal Processing

Reality mining is expected to improve public health and medicine. It was named as one of "10 emerging technologies that could change the world". This is not only about mining data pertaining to social behavior - although social factors do impact well-being and social data provides valuable health predictors. It's also about health data collected in real and near real time. Audience asked about ways to collect health data and their limitations. Some of the questions reflected earlier Q&A with the Data Mining Camp expert panel , especially Dr. Michael Walker, author of numerous FDA and CLIA-approved products to diagnose and treat disease. He commented on the need to have tools well beyond the data available to take cell samples every two minutes instead of laboratory testing every few months - in order to allow cellular simulations and obtain parameters for differential equations. Sampling frequencies don't come close to allow this kind of modeling. Irene Gabashvili agreed that first-principles cellular modeling for predicting health won't be possible (although there are engineers that believe a platform for real-time in-vivo measurements of most cells can be developed). Yet, good predictors could be and will be developed - based on sensors measuring macro-level observations and missing value estimators. Genetic information is not enough, we need to capture environmental risk factors. How can we separate genes and environment?, asked one of the participants. Aurametrix' initial focus is on chemicals in our food - and even though some may argue that our taste and satiety mechanisms are dictated by genes, food analytics provides insights into non-genetic components of our health. Other questions were on the time line for body sensor networks and data growth. Jeffrey Nick of EMC estimates that personal sensor data will balloon from 10% of all stored information to 90% within the next decade. Irene Gabashvili thinks that this will happen rather sooner than later, perhaps in the next two years.

Another interesting aspect of reality mining is crowdsourcing or collective intelligence - in order to get useful information from all the data (temporospatial location, GPS, activity, food, symptoms, behavior, communication content, proximity sensing), we need to analyze it not only on individual but also group level. We need to share more, without sacrificing privacy and security. Collective contributions can be reliable - Shamod Lacoui's answer to this is in selecting those who contribute, restricting inputs to domain. It would help to “filter out the dross”, while “saving the best”. It is needed to suppress noise, to infer intelligence from the collection of facts, clicks, steps, whatever one can contribute. This resonates with discussions at the Transparency Camp - one of the useful tools is SwiftRiver - free, open source software platform to validate and filter news. Swift relies on Natural Language Processing, Machine Learning and Veracity Algorithms to track and verify the accuracy of reports and suppress noise (like duplicate content, irrelevant cross-chatter and inaccuracies). Transparency Camp also posed a question on whether there is a need for an FDA-like institution to ensure information safety and healthy information consumption.

Self-serve was a topic of a smaller Data Mining Camp Session. Even though it was aimed at sales reps that need to go beyond Excel spreadsheets to mine private data of their interest, self-service is currently the only option for health care consumers. People need to analyze everyday life for health implications. They need better tools to not focus on metrics that are easy to collect instead of metrics we need to collect.

In order to mine high-dimensional health space, many disparate types of data should be mashed and validated, gaps should be bridged and structured metadata added to data. Randy Kerber talked about data formats and approaches to make it happen. Semantic web discussions involved NoSQL experts that mentioned limitations of gaining popularity technologies such as MongoDB, Cassandra and HBase. Another relevant session - on cloud computing - discussed its (sometimes over-rated ?) performance and Hadoop technologies.

Visualization techniques provide one of the most effective methods of extracting knowledge from health data. Remember who invented the pie chart? That's right, it was Florence Nightingale, a nurse who needed a way to better represent her data. one of the most famous examples of visualizing epidemiological data was Dr. John Snow's map of deaths from a cholera outbreak in London, Many other techniques and software tools exist, but maps remain popular - especially google maps API. One of popular tools for epidemiological data is Google Maps API. For example, it embeds Google Maps into healthmap.org with JavaScript.

One of the participants of Biomedical Session developed kidsdata.org (@kidsdata on twitter). It provides insights into geospatial autism statistics and visualizes trends and other useful health-related information.

Another way to display geospatial data is Dynamic Choropleth Maps. Complex networks can be also explored with alluvial diagrams and other approaches. More visualization techniques and tools can be applied to health data - to look at the data in new ways and gain useful insights.

Some of the questions from the audience were on the availability of data. Sources discussed included CDC (see, for example, NHANES laboratory files; eHealth metrics) and Entrez Life Sciences databases.

Signal Detection and Signal Processing for Mining Information was another discussion topic.
Questions were on data mining versus simple tracking and signal monitoring. It was agreed that data mining is the key to health management. Cardionet, body sensors (see posts on teletracking, M-health, Telemedicine: part 1; Telemedicine: part 2; Health 2.0 Software tools, Devices to keep you healthy), SNP detection, telemedicine applications, random and rare electrocardiographic events and other applications were also discussed.




See other materials from Data mining Camp 2010:

• ACM Data Mining Camp, March 20, 2010 (sfbayacm.org)
• Data Mining Camp Report (Zemanta)
• Mining Data mining Camp Impressions (Aurametrix)
  
• My Experience at ACM Data Mining Camp #DMcamp (bytemining.com)

• Transparency Camp, March 27-28, 2010 (The Sunlight Foundation)
  
• Notes from Transparency Camp (David “Oso”, Rising Voices)
  

Biomedical Data Mining: Dimensionality, Noise, Applications

Image by juhansonin via Flickr

ACM Silicon Valley Data Mining Camp on November 1, 2009 has attracted more than 200 people with different backgrounds and interests. It was held at Hacker’s Dojo, sponsored by REvolution Computing, KXEN (Knowledge Extraction Engines), and LinkedIn (See notes on this event by @Andraz of Zemanta, Ken's open source tools, and relevant #dmcamp twits) . Biomedical/Healthcare data mining topic was suggested by Junling Hu and Irene Gabashvili and supported by A.J. Chen, Greg Makowski, Sukanta Ganguly, and 40 other participants of the Data Mining Camp. Below is a brief transcript of the discussion. The session started from introductions, here are some of them:

• Irene, with background in biophysics, medical informatics and CS, pursuing a personal health management venture, interested in data mining to advance personalized medicine;
• Lawrence, with background in physics and software engineering and interest in health IT. He is the organizer of Google Wave meetup (you may know about Google Health Wave);
• Hua, formerly with Kaiser, interested in medical scheduling and web development;
• Liana, interested in Natural Language Processing for biomedical knowledge mining;
• Maura, interested in Health IT, medical engineering and security;
• Magnus, developing Medical Databases;
• Kevin, interested in medical startups;
• Watson, with background in genomics and machine learning;
• Peters, working on medical devices and software embedded systems;
• Steve, formerly of Applied Biosystems;
• Roy of Codexis, focusing on data mining and pattern recognition in multivariate time series
• Jima, with background in medical informatics;
• Karsleep, interested in biomedical data mining;
• Deena, scientific analyst interested in how data mining technologies could be applied to healthcare;
• Junling Hu, scientist at Bosch, working on a device and software collecting and analyzing patients' information, based on daily questionnaires and other collected data.

There also were people interested in the subject merely as healthcare consumers (aren't we all?). Junling started the session from mentioning a recently published paper on computer technologies for healthcare determining strategic directions in the area. Irene also suggested to check the mHealth Summit focusing on mobile technologies to improve research data collection, healthcare delivery, and health outcomes. Junling described the project she was working on - inexpensive device collecting data and sending it to a "coaching" nurse that monitors stay-at-home patients. Next step is to mine the data automatically, thus reducing the load on healthcare professionals without sacrificing patients' well-being. Junling also mentioned some of the challenges such as compliance of participants who are typically not eager to fill out the 20-question surveys. This is especially bad for obesity studies. The data mining challenges mentioned during the session were: (1) Missing Data We are not talking about sparse data (discussed in one of the previous sessions on data mining with R), but actually missing data. Data is sparse if only a small fraction of the attributes are non-null - like the number of items we typically buy in a grocery store is much less than the number of products they offer. Data is missing if the values were never entered or the member combination is not meaningful (for example, obstetrics/gynecolgy values not meaningful for men) . One of the suggestions from the experts in the audience was to utilize "multiple imputation". Other suggestions included "once-a-week" questioning instead of daily surveys. Irene mentioned the 7D-PAR (Seven-Day Physical Activity Recall) , one of standardized questionnaires developed in the 80s (1,2) and other established methods. Questions and comments from the audience:

• Data mining methods utilized for Chronic Disease Assesment and Elderly monitoring. Junling talked about unsupervised classification algorithms and two supervised learning methods she found to be most useful for her work - SVM and logistic regression. Both were equally good in predicting hospitalization events
• Indicators of Goodness of Model Predictions. Suggested events were hospitalizations, mortality... It was noted that good indicators are yet to be found.

(2) Performance Criteria This was another health data mining challenge emphasized during the discussion. All standard methods can be applied such as accuracy, precision, recall, true positives, false positives and especially combinations of the last 2 measures. Junling mentioned breast cancer classifier developed by Siemens and other algorithms predicting emergency situations with 90% accuracy. Irene noted that one of the problems of digital mammography and other cancer predictors is a high rate of false positives. From 30 to 40% of cancers are overdiagnosed (3), thus increasing healthcare costs. This has to be changed. Several people in the audience emphasized that existing methods are averaging the population. Medicine needs to be truly personalized, we need better methods and more data. (3) Large Number of Input Features One of the main problems of health data mining is coping up with large number of input features. Obviously, a 20-question test is not sufficient. Should it rely on thousand questions or trillion inputs? And how to select a subset of relevant features to build robust learning models? Junling's preffered approaches are logistic regression and singular value decomposition. She would add features one by one and check if the overall accuracy for predictions remains good. Questions from the audience included:

• A 3-5 year Vision for Health Data Mining: what do we expect to achieve? Participants expressed an optimistic outlook
• Ray: Are most input variables discrete or continuous? The answer was: mixed

(4) Very-Large Scale Data The good thing about pattern recognition is that the more patterns you have, the better it performs. Google translator is a good proof of this assertion (although this translator needs even more patterns to do a decent job). Biomedical data sets such as CT, MRI, PET scans and other image data, gene expression, genetic variation are very large scale in nature. The challenge for data miners is to integrate and extract information from data of such scale. Questions from the audience:

• What are the other large-scale studies trying ot mine patterns in health data, outside of US? Studies in China and Taiwan using similar devices and models; also in Europe

Adding to this interesting discussion that was unfortunately interrupted because of the lack of time, I'd like to mention a few other challenges facing biomedical data mining.

• We should not underestimate the complexity of relationships between causative and effect variables in human health. Simplistic approaches are deemed to fail. Over-fitting could be a problem too
• Integration between heterogeneous data sources and types,and putting content in context (semantic integration) remains a challenge.
• Privacy Concerns associated with the Sharing of Individual Health Information.

References

1. Blair S. How to assess exercise training habit and physical fitness. In: Behavioral Health, edited by Matarazzo JD. New York: Wiley, 1984, p. 424-447.
2. Rauramaa R., Tuomainen P., Väisänen S., and Rankinen T. Physical activity and health- related fitness in middle-aged men. Med Sci Sports Exerc 27: 707-712, 1995.
3. Gøtzsche, P.C., Jørgensen, K.J., Mæhlen, J. and Zahl, P.-H. Estimation of lead time and overdiagnosis in breast cancer screening. British Journal of Cancer (2009) 100, 219–219.

Related articles by Zemanta

• Health IT: What's the Future? (rwjfblogs.typepad.com)
• Thoughts on the SF Bay Data Mining Camp (revolution-computing.com)