Biomedical engineering cheat sheet

Biological systems

A biomedical model links physical or computational variables with a biological question. State what the model leaves out.

A simplified flow model does not represent every feature of circulation.

Measurement chain

A sensor, conditioning stage and recording system each influence the measured signal.

Noise can enter at more than one stage.

Sampling

Sampling converts a changing signal into discrete measurements. Sampling rate and signal bandwidth affect interpretation.

For an ideal band-limited signal, the sampling rate must exceed twice its highest frequency.

Evaluation

Evaluate measurement accuracy, repeatability, uncertainty and suitability for the intended use as distinct questions.

A repeatable measurement can still have systematic bias.

Accuracy, precision and calibration

Accuracy concerns closeness to a reference; precision concerns repeatability. Calibration relates an instrument’s output to known reference inputs under specified conditions.

Repeated readings can agree closely while sharing the same systematic offset.

Signal and noise

A measurement contains the desired signal plus unwanted contributions. Filtering can reduce some noise but can also remove or distort useful information.

Specify which frequency components matter before choosing a filter.

Sampling and aliasing

A sampled signal can misrepresent frequencies above the supported range. The Nyquist condition concerns a band-limited signal and requires an appropriate analogue anti-aliasing strategy.

To represent a 40 Hz band-limited signal, the sampling rate must exceed 80 samples/s under the ideal sampling model.

Biocompatibility and context

A material’s interaction with the body depends on contact type, duration and the device’s intended use. A successful bench test alone does not establish clinical safety or effectiveness.

Evaluate a device as a complete system, including users and the intended environment.

Worked example: sampling interval

A recorder samples at 200 samples/s. The interval between samples is 1/200 s = 0.005 s, or 5 ms. This does not by itself establish the sensor’s bandwidth.

Sample rate and measurement accuracy describe different properties.

Common mistake: correlation validates a device

A high correlation can coexist with systematic bias. Agreement, calibration range, uncertainty and performance on the intended population also need evaluation.

Plot differences against a reference rather than reporting correlation alone.

Check your understanding

Can a sensor be precise but inaccurate?

Yes. Its repeated readings may be close to one another while all are offset from the reference.

Curriculum connections

Sources