Knit figures to PDF

Author: Max Czapanskiy

analysis/figures/bw-bcg-1.pdf

@@ -52,6 +52,7 @@ knitr::opts_chunk$set(
   fig.path = "../figures/",
   fig.height = 6,
   fig.width = 6,
+  dev = "pdf",
   dpi = 300
 )
 library(cetaceanbcg)
@@ -279,7 +280,7 @@ The 3-dimensional BCG exhibited increasing heart rates over the course of dives.
 
 ## Limitations and considerations for future applications
 
-While the BCG method presented here holds the potential to mine existing and future marine mammal bio-logging datasets for information about cardiovascular function, it has several limitations compared to ECG methods. Most importantly, BCGs are highly sensitive to movement artifacts [@inanBallistocardiographySeismocardiographyReview2015], so only motionless periods are valid for analysis. This limits the behavioral and physiological contexts in which heartrate may be measured. For example, the BCG is probably an inappropriate method for quantifying the magnitude of surface tachycardia [@goldbogenExtremeBradycardiaTachycardia2019] and exercise modulation of bradycardia [@noren2012], due to movement artifacts during those activities. Secondarily, we did not test whether the BCG is robust to tag placement location. The blue whale data presented in this study was collected when a dorsally-deployed tag slipped to the lateral chest cavity behind a flipper, where it is reasonable to expect greater accelerations caused by heart beats than from a tag farther from the animal's center of mass. It is possible that the ballistic forces generated by heart beats are strong enough to produce an interpretable BCG for a variety of potential tag deployment locations, but this likely varies with animal body size, as well as accelerometer sampling rate and sensitivity.
+While the BCG method presented here holds the potential to mine existing and future marine mammal bio-logging datasets for information about cardiovascular function, it has several limitations compared to ECG methods. Most importantly, BCGs are highly sensitive to movement artifacts [@inanBallistocardiographySeismocardiographyRemovview2015], so only motionless periods are valid for analysis. This limits the behavioral and physiological contexts in which heartrate may be measured. For example, the BCG is probably an inappropriate method for quantifying the magnitude of surface tachycardia [@goldbogenExtremeBradycardiaTachycardia2019] and exercise modulation of bradycardia [@noren2012], due to movement artifacts during those activities. Secondarily, we did not test whether the BCG is robust to tag placement location. The blue whale data presented in this study was collected when a dorsally-deployed tag slipped to the lateral chest cavity behind a flipper, where it is reasonable to expect greater accelerations caused by heart beats than from a tag farther from the animal's center of mass. It is possible that the ballistic forces generated by heart beats are strong enough to produce an interpretable BCG for a variety of potential tag deployment locations, but this likely varies with animal body size, as well as accelerometer sampling rate and sensitivity.
 
 When auditing existing bio-logging data and planning future tag deployments for BCG analysis, careful consideration should be paid to sampling rate. As a rule of thumb in signal processing, the sampling rate should be at least twice the frequency of the phenomenon of interest. In the case of the BCG, the relevant frequency is that of the BCG waveform, *not* the heartrate. In humans, the power of the IJK-complex (the part of the BCG waveform used for heart beat detection) is concentrated between 4-7 Hz [@moukadem2018]. It is unlikely that marine mammal BCG waveforms have a higher frequency than humans, owing to their generally larger body sizes. Therefore, it is possible that BCGs may be generated for accelerometer sampling rates as low as 10-15 Hz. Conservatively, the authors recommend a sampling rate of no less than 50 Hz (i.e., twice the upper cut-off frequency of the widest bandpass filter used in this study).