2009 News

Prepare for the PhysioNet/Computers in Cardiology Challenge 2010, Mind the Gap! A set of training data is now available. The Challenge begins on 1 January.

The entire European ST-T Database, a landmark set of expert-annotated recordings for study of the ECG manifestation of myocardial ischemia and for development and evaluation of methods for quantifying ST and T changes in the ECG, is now available on PhysioNet. Roughly half of the database was previously contributed to PhysioNet in September, 2000; this new contribution makes the entire database available for free use by researchers for the first time.

If you have any questions or comments regarding this challenge, please post it directly in our Community Discussion Forum. This will increase transparency (benefiting all the competitors) and ensure that all the challenge organizers see your question.

Papers presented at CinC 2009 by challenge participants, scores, and sources for the open-source entries, are now available.

PhysioNet is seeking contributions of data from EMG and nerve conduction studies. The Margret and H.A. Rey Institute for Nonlinear Dynamics in Medicine has offered an annual award for the best such contribution.

A set of 64-channel EEGs from subjects who performed a series of motor/imagery tasks has been contributed to PhysioNet by the developers of the BCI2000 instrumentation system for brain-computer interface research.

The Challenge has concluded, and the correct classifications for event 1 and event 2 are now available (follow the links). In addition, the c records (the data following T₀ for each of the test set records, which have been withheld for the duration of the Challenge) are now available in the same locations as the a and b records that have been available since April.

Thanks to all of the Challenge participants, many of whom will discuss their work during dedicated scientific sessions of Computers in Cardiology next week. A number of participants were able to classify all 10 cases in event 1 without errors; many were able to classify at least 80% of cases in event 2, and the best result achieved in event 2 was correct classification of 93% (37 of 40) cases. The final scores, and the open source software developed and contributed by participants in the open source division of the Challenge, will be posted shortly after CinC.

Special thanks to Franco Chiarugi, whose invaluable feedback at every stage prompted corrections in the training set and improvements in the design of the challenge that contributed significantly to its success.

We are pleased to announce that a limited distribution of the entire MIMIC-II Clinical Database, containing data for over 26,000 subjects, has been contributed to PhysioNet and is now available to qualified researchers. Details are available in the paper by Mohammed Saeed et al entitled Multiparameter Intelligent Monitoring in Intensive Care II (MIMIC-II): A public-access intensive care unit database.

Over 1300 records have been added to the MIMIC-II Waveform Database, bringing the total to over 4000 multi-parameter records from ICU patients, each typically several days in length, with many that are much longer. The new records include data from 249 neonatal ICU patients, a population not previously represented in PhysioNet. [Numbers of records have been corrected as of 30 July 2009.]

The clinical data that accompany the test set records are now available (in the same locations as the test set waveforms and numerics records). Also note that the CinC abstract deadline has been extended to 8 May 2009 and we have thus been able to extend the first entry deadline to 6 May 2009. If you miss these deadlines, you are still welcome to participate, but you will be ineligible for an award. If you submit your first entry on or before the deadline, you may revise it (see restrictions below) until the final entry deadline of 31 August 2009.

PhysioBank's Automated Teller Machine, a self-service facility for exploring PhysioBank using a web browser, is now on-line. Replacing the original PhysioBank web services introduced in 2000, the ATM offers a variety of tools for visualizing PhysioBank data and converting them to text, EDF, or .mat files, and more.

The WFDB Software Package, versions 10.4.17 and later, includes wfdb2mat, a program that converts all or any desired portion of the signals in any PhysioNet or compatible record to a .mat file that can be read by Matlab or Octave. If the original signal files have not been downloaded from PhysioNet, <tt>wfdb2mat</tt> can read them directly from the PhysioNet server. The <tt>.mat</tt> files generated by <tt>wfdb2mat</tt> are also directly readable by dozens of other applications in the WFDB Software Package, including WAVE.

A few days ago, the first 60 patient records from the MIMIC II Clinical Database were posted, as part of the training set for our annual challenge (see the news item immediately below). A new tutorial, Using the MIMIC II Database, is now available to help challenge participants and others interested in this collection and in linking it with the recorded physiologic signals and time series of vital signs from the same patients in the MIMIC-II Waveform Database.

The 10th annual PhysioNet/Computers in Cardiology Challenge is underway. Participants are invited to identify which ICU patients are likely to experience sudden and severe drops in blood pressure. Successful prediction of these acute hypotensive events may lead to improved care and survival of patients at risk. The Challenge culminates in September with presentations by participants at a dedicated scientific session of Computers in Cardiology in Park City, Utah.

Heart Rate Variability Analysis with the HRV Toolkit: Basic Time and Frequency Domain Measures is a new tutorial that describes how to use the HRV toolkit (an accompanying software package created for PhysioNet by Joe Mietus) to select and prepare time series of inter-beat intervals and to calculate measurements of the basic time- and frequency-domain HRV statistics that are widely used in the literature. Particular attention is given to techniques for identifying and dealing with outliers, in order to permit reliable determination of measurements.