Digital Biomedical Signals and Signal Processing - An Introduction
Raja Selvaraj
Introduction, Analog versus Digital signals
Biomedical signals are part of our daily work
ECG
EGMs
Echo
CAG
Analog signals
Photo from film camera
Sound stored on grooves in Vinyl record
Analog signal - zooming in
Digital signals
ECG today
X-ray in PACS
Angiogram
Digital photo
Music CD
Digital signal - zooming in
Analog vs Digital
Analog - record of signal directly to a recording medium
Digital - signal converted to bits - stored / displayed / printed
Analog has higher resolution
But practically limited by the recording medium
Digital easier to store and reproduce
Digital can be manipulated easier
Analog to digital conversion
Sampling rate
Heart rate 60 bpm (1 beat / sec), measured HR with sampling at 1 Hz
Heart rate 60 bpm (1 beat / sec), measured HR with sampling at 1.25 Hz
Heart rate 60 bpm (1 beat / sec), measured HR with sampling at 2 Hz
Adequate sampling
Sampling frequency should at least be twice that of frequency of interest
Higher frequencies than that sampled may show aliasing
Amplitude and Bit resolution
Bit resolution, LSB and A/D conversion
ECG conversion
12 bits, voltage range -5mV to 5mV
10,000 microV / 2 ** 12
LSB = 2.44 microV
Display of signals - Time Domain vs Frequency domain
Prescription - Time domain
Monday 8am - Metoprolol 25 mg
Monday 8pm - Metoprolol 25 mg
Monday 9pm - Omeprazole 20 mg
Tuesday 8am - Metoprolol 25 mg
Tuesday 8pm - Metoprolol 25 mg
Tuesday 9pm - Omeprazole 20 mg
Prescription - Frequency domain
Metoprolol 25 mg BID (starting 8 am)
Omeprazole 20 mg OD (starting 9 pm)
HRV - Frequency domain
Time to frequency domain
Fourier transform
Fourier transform
Discrete Fourier transform
Fast Fourier transform
Microvolt TWA
Storage / export
Binary format
bit - 0 / 1
8 bits = 1 byte
1024 bytes = 1 kilobyte (KB)
1024 kilobytes = 1 megabyte (MB)
More compact storage
Single ECG lead
10 seconds
1000 Hz
12 bits
Will need 15 kb
Text format
CSV
XML
Larger size
Same single lead
50 kb
Binary vs ASCII
Binary smaller
Faster to read
But format needs to be known
Often proprietary
ASCII human readable
Slight slower to read and larger
FDA recommends XML for ECG
Signal processing
Filter
Reducing signal amplitude in selected frequencies
Time domain (moving average)
Frequency domain
Frequency domain filter
Filters
Low pass
High pass
Bandpass
Notch
Bidirectional filter to avoid ringing artifact
Differentiation
Averaging
Beat detection (QRS)
Mean / Median of multiple beats
Increases signal to noise ratio
Average beat