Biomedical Imaging and Diagnostic Systems Exercises

These exercises practise biomedical imaging and diagnostic systems as task-based measurement systems. They cover contrast-to-noise ratio, spatial sampling, dose-noise trade-offs, ultrasound wavelength, acquisition latency, detector quantization, optical signal quality, diagnostic performance metrics, measurement uncertainty, workflow throughput, and post-deployment quality assurance.

The goal is not to make an image look sharper in isolation. The goal is to decide whether an imaging system preserves the information needed for a declared diagnostic, measurement, monitoring, or guidance task while controlling risk and uncertainty.

Assume simplified engineering models unless an exercise states otherwise. Real imaging and diagnostic systems should also check intended use, patient or sample population, modality-specific safety, calibration, reconstruction settings, operator workflow, software version, display configuration, artifacts, reference method, and validation evidence.

How to Use These Exercises

For each imaging calculation, define:

1. the diagnostic or measurement task;
2. the modality, source, detector, geometry, and processing chain;
3. the image-quality or diagnostic-performance metric;
4. the safety, uncertainty, or workflow constraint;
5. the validation evidence needed before the result supports a claim.

The common mistake is optimizing a visual output without proving task performance. A useful imaging metric must connect to the decision the image is supposed to support.

Treat every numerical result as a task-specific screen. A CNR, pixel size, latency, sensitivity, or throughput value becomes engineering evidence only when the acquisition protocol, reconstruction version, user workflow, reference method, and acceptance criterion are controlled.

Exercise 1: Contrast-to-Noise Ratio

A phantom image has mean target intensity:

mean background intensity:

and background noise standard deviation:

Use:

Calculate contrast-to-noise ratio.

Solution

Difference in mean intensity:

CNR:

Engineering Comment

The CNR is 4.0 for this phantom, acquisition mode, reconstruction setting, and background region. That context is essential. CNR can change with dose, gain, reconstruction algorithm, filter, phantom material, motion, detector calibration, and display processing.

If the task is low-contrast detection, CNR should be tied to observer performance or a task-based validation method, not treated as a universal pass by itself.

Exercise 2: Pixel Size and Target Sampling

An imaging system has field of view:

and matrix size:

Calculate pixel size. Then estimate how many pixels span a 1.5 mm target.

Solution

Pixel size:

Pixels across the target:

Engineering Comment

The 1.5 mm target spans about 3.2 pixels. That may be enough for visibility in a high-contrast phantom but weak for precise measurement, segmentation, or low-contrast detection. Partial-volume effects, point-spread function, reconstruction blur, and motion can reduce effective resolution.

Spatial sampling should be judged against the task, not only the matrix size.

Exercise 3: Dose-Noise Trade-Off Screen

In a simplified x-ray imaging model, image noise is inversely proportional to the square root of relative exposure:

If exposure is reduced to 64 percent of the baseline:

estimate the noise increase factor.

Solution

Noise ratio:

The simplified model predicts a 25 percent noise increase.

Engineering Comment

The calculation illustrates a basic exposure-noise trade-off, not a clinical dose recommendation. Real imaging review must include diagnostic task, patient size, reconstruction algorithm, detector performance, motion, artifacts, and applicable dose-management controls.

Lower exposure is not automatically better if the resulting image no longer supports the intended decision.

Exercise 4: Ultrasound Wavelength

An ultrasound probe operates at:

Assume sound speed in soft tissue:

Estimate wavelength:

Solution

Convert frequency:

Wavelength:

In millimeters:

Engineering Comment

The wavelength is about 0.205 mm. Higher frequency reduces wavelength and can improve spatial detail, but it usually reduces penetration depth. Probe choice should match anatomy, depth, acoustic window, frame rate, and safety constraints.

Ultrasound resolution is not determined by wavelength alone; aperture, focusing, bandwidth, beamforming, speckle, and operator technique also matter.

Exercise 5: Imaging Latency for Guidance

A guidance system acquires images at:

The reconstruction pipeline adds:

Display update adds:

Estimate approximate end-to-end latency using one frame interval plus reconstruction and display time.

Solution

Frame interval:

End-to-end latency:

Engineering Comment

The estimated latency is about 140 ms. For static review this may be acceptable, but for guidance, navigation, or motion compensation it may be too high depending on motion speed and clinical task.

Latency validation should include acquisition, reconstruction, display, tracking, network transfer, buffering, and worst-case processing load.

Exercise 6: Detector ADC Quantization

A detector channel uses a 14-bit ADC with full-scale range:

Calculate quantization step size.

Solution

Number of codes:

Step size:

Engineering Comment

The step size is 0.153 mV at the ADC input. More bits do not guarantee better image quality if detector noise, analog gain, dark current, calibration drift, saturation, or reconstruction bias dominate.

ADC resolution should be evaluated in the detector chain, not as an isolated specification.

Exercise 7: Optical Signal-to-Noise Ratio

An optical diagnostic channel measures signal current:

The RMS noise current in the measurement bandwidth is:

Calculate SNR as a current ratio and in decibels.

Solution

Current ratio:

SNR in decibels for equal-impedance amplitude ratio:

Engineering Comment

The SNR is 20, or 26.0 dB, in the stated bandwidth. Optical measurements can change with ambient light, tissue scattering, source drift, contact pressure, skin tone, motion, detector saturation, and temperature.

The validation question is whether this SNR supports the intended diagnostic task, not whether it looks large in isolation.

Exercise 8: Sensitivity and Specificity from a Validation Dataset

A diagnostic algorithm is evaluated against a reference method on 220 cases:

Result	Count
True positive	72
False negative	8
True negative	126
False positive	14

Calculate sensitivity and specificity.

Solution

Sensitivity:

Specificity:

Engineering Comment

Both sensitivity and specificity are 90.0 percent for this dataset. That does not automatically prove clinical adequacy. The review should check case mix, prevalence, reference method quality, confidence intervals, subgroup performance, acquisition conditions, user workflow, and failure consequences.

Diagnostic performance metrics are evidence only when the validation dataset represents the intended use and the uncertainty around the estimate is acceptable for the claim.

Exercise 9: Measurement Uncertainty for Vessel Diameter

An imaging system estimates vessel diameter. Standard uncertainty components are:

Component	Standard uncertainty
Pixel calibration	0.04 mm
Segmentation repeatability	0.08 mm
Motion blur contribution	0.06 mm
Reference phantom uncertainty	0.03 mm
Operator cursor placement	0.05 mm

Estimate combined standard uncertainty.

Solution

Combined standard uncertainty:

Engineering Comment

The combined standard uncertainty is about 0.122 mm. If the clinical or engineering decision threshold is close to this value, the measurement may not be reliable enough without improved acquisition, segmentation, calibration, or repeatability evidence.

Quantitative imaging should report uncertainty with the measurement context, not only the displayed diameter.

Exercise 10: Diagnostic Throughput and Queue Risk

An imaging room has:

of usable scanning time. Average room time per case, including setup and cleaning, is:

The daily scheduled demand is 28 cases. Calculate daily capacity and whether demand fits.

Solution

Available minutes:

Capacity:

Demand comparison:

Demand exceeds nominal capacity by:

Engineering Comment

The schedule is over capacity by 3 cases per day before considering late arrivals, repeat scans, equipment faults, contrast delays, patient transport, cleaning exceptions, or reporting workflow.

Diagnostic system performance includes throughput and workflow reliability. A technically excellent scanner can still fail the service need if operational capacity is unrealistic.

Exercise 11: Post-Deployment QA Drift

A weekly phantom QA program tracks CNR. The baseline CNR after commissioning was:

The current four-week average is:

The local investigation trigger is a CNR decrease greater than 10 percent from baseline. Calculate the decrease.

Solution

Relative decrease:

Comparison:

The result exceeds the investigation trigger.

Engineering Comment

The QA trend should be investigated. Possible causes include detector drift, calibration change, reconstruction update, source output change, phantom positioning, environmental effects, or user workflow change.

Post-deployment quality assurance is part of validation maintenance. A system can pass at release and later drift away from the evidence basis that justified clinical use.

Review Checklist

When reviewing biomedical imaging evidence, ask:

• Is the metric tied to the diagnostic, measurement, monitoring, or guidance task?
• Are acquisition settings, reconstruction version, display settings, and calibration state controlled?
• Does image quality include contrast, noise, resolution, artifacts, uncertainty, and workflow?
• Are dose, acoustic output, optical exposure, heat, electricity, infection, and data-integrity risks controlled?
• Does validation represent anatomy, motion, operator skill, patient or sample variability, and site workflow?
• Are algorithm outputs tested for bias, failure modes, and version changes?
• Does QA monitoring detect drift after installation, updates, maintenance, and routine use?
• Are acceptance criteria defined before validation data are reviewed?
• Can the image, measurement, or diagnostic output be traced back to configuration and evidence?

Biomedical imaging is trustworthy when physics, computation, safety, calibration, workflow, and validation all support the declared task.