3 The planned approach and methods for solving the problem

3.1 State of the art

A literature review of the last ten years is being conducted to assess state of the art for ECG automatic processing collecting the following points if available : (1) The memory and space used to perform the primary goal of the algorithm (sound an alarm, for ex.). (2) The type of algorithms used to identify ECG anomalies. (3) The type of algorithms used to identify specific diagnoses (like a flutter, hyperkalemia, and others). (4) Their performance (accuracy, ROC, etc.)

A broad search will be conducted on Pubmed, Scopus, Google Scholar, device manuals, and other specific sources.

Keywords:

  • ECG AND monitoring AND ICU
  • ECG AND[time series]
  • ECG AND automatic AND interpretation

Articles published after “The PhysioNet/Computing in Cardiology Challenge 2015: Reducing False Arrhythmia Alarms in the ICU” will also be analyzed.

3.2 Research plan and methods

This research is being conducted using the Research Compendium principles15:

  1. Stick with the convention of your peers;
  2. Keep data, methods, and output separated;
  3. Specify your computational environment as clearly as you can.

Data management is following the FAIR principle (findable, accessible, interoperable, reusable)16.

Currently, the dataset used is stored on a public repository17, the source code is publicly open and stored on Github18, while the reports and reproducibility information on each step is found on a public website19.

3.2.1 Type of study

This thesis will be a diagnostic study as the algorithm must classify the change in pattern as positive or negative for life-threatening.

3.2.2 The data

Initially we will use the CinC/Physionet Challenge 2015 dataset that is publicly available on Physionet. This dataset is a good start for exploring the main goal of reduce false alarms. This dataset was manually selected for this challenge and the events were labeled by experts, so it is not RAW data. All signals have been resampled (using anti-alias filters) to 12 bit, 250 Hz and have had FIR bandpass [0.05 to 40Hz] and mains notch filters applied to remove noise. Pacemaker and other artifacts may be present on the ECG6. Furthermore, this dataset contains at least two ECG derivations and one or more variables like arterial blood pressure, photoplethysmograph readings, and respiration movements.

These variables may or may not be helpful for increasing the sensitivity or specificity of the algorithm. It is planned to use the minimum set of variables as it is known in multi-dimensional analysis that using just two (or some small subset) of all the dimensions can be much more accurate than either using all dimensions or a single dimension20.

It is desirable that real data extracted from Portuguese ICU could be used in further stage to assess the validity of the model in real settings and robustness (using RAW data instead of filtered data). The variables available on Physionet’s dataset are commonly available on Portuguese ICU’s.

3.2.3 Workflow

All steps of the process will be managed using the R package targets21 from data extraction to the final report, as shown in Fig. .

The report will then be available on the main webpage19, allowing inspection of previous versions managed by the R package workflowr22, as we can see in Fig. .

3.2.4 Statistical analysis

The Statistical analysis will be performed using R language v4.0.5 or greater and it will be computed the ROC curve for the algorithm.

The experiment will be conducted using the Matrix Profile concept23, the state-of-the-art for time series analysis. It will be conducted several experiments to identify the best algorithm for this task. One of such algorithms is the online semantic segmentation for multi-dimensional time series20.

In addition, we will combine the fading factors24,25 strategy to minimize the memory and space consumption lowering the processor power needed, allowing this algorithm to be used in almost any device.

3.2.5 Research Team

  • Thesis Author: Francisco Bischoff
  • Supervisor: Professor Pedro Pereira Rodrigues
  • Co-supervisor: Professor Eamonn Keogh (UCR, Riverside)

3.3 Expected results and outcomes

We expect the following results: (1) Identify the obstacles of identifying life-threatening ECG changes within memory, space, and CPU constraints. (2) Be able to reduce ECG monitor’s false alarms using the proposed methodology. (3) Assess the feasibility of implementing the algorithm in the real world and other settings than ICU monitors.

And outcomes: (1) To achieve a better score of false alarm reduction than the best on Physionet’s 2015 challenge. (2) To push forward the state-of-the-art technology on false alarms reduction, maybe even being domain agnostic. (3) To draw more attention to fading factors as a reliable, fast, and cheap approximation of the true value. (4) To draw more attention to the matrix profile concept as an efficient, agnostic, and almost parameter-free way to analyze time series. (5) To draw more attention of the Patient Monitorization industry on solving the false alarm problem.