Using Statistical Process Control to Monitor Medical Linacs

Reliable treatment delivery depends on knowing whether a machine is performing within its established clinical range. Statistical process control (SPC) provides a structured way to detect meaningful changes in a linear accelerator, imaging system, treatment planning workflow or radiation measurement process before a small drift becomes a patient safety concern.

For medical physicists, the value of SPC lies in separating ordinary measurement variation from a genuine special cause. This makes daily quality assurance more informative than a simple pass-or-fail exercise. It also creates a clear record for service meetings, accreditation reviews, equipment replacement decisions and communication between physicists, radiation therapists and engineers.

Why Process Control Matters In Radiotherapy

Traditional quality assurance often compares a result with a tolerance limit. That approach is essential, but it can miss gradual movement inside the permitted range. A beam output reading that changes slightly each week may remain technically acceptable while signalling an emerging calibration, environmental or hardware issue.

SPC adds a time-based view. Control charts show the expected centre line, the natural spread of the process and limits based on observed performance. A stable system produces random variation around its average. A run of results on one side of the mean, a trend, or an abrupt shift may justify investigation even when no tolerance has been breached.

Establishing A Defensible Baseline

A useful baseline begins with consistent measurement conditions. The same detector, setup geometry, beam energy, phantom, software version and acquisition method should be used wherever practical. Records should include temperature, pressure, machine state and any maintenance that could influence the result.

Baseline data should represent routine operation rather than an unusually short period after commissioning. Depending on the test, several weeks or months may be appropriate. The medical physics team can then calculate the mean and standard deviation, review outliers and confirm that the process is sufficiently stable before setting statistical limits.

Clinical tolerances still apply. Control limits describe the behaviour of the local process; they do not replace national guidance, manufacturer specifications, departmental action levels or professional requirements. A process may be statistically stable yet clinically unacceptable, or clinically compliant yet statistically unstable.

Selecting Measures That Reveal Drift

The strongest SPC programme uses measurements linked to patient risk and machine function. Daily output, flatness, symmetry, imaging coincidence and laser alignment may be suitable for frequent review. Monthly checks can examine more complex characteristics such as multileaf collimator positioning, dose-rate response or image-guided radiotherapy accuracy.

Data quality matters as much as chart design. A missing result should be marked as missing rather than silently replaced, and changes in equipment or technique should be annotated. If a detector is recalibrated or a new software release changes image reconstruction, the team should decide whether the baseline remains valid.

Useful measures should be sensitive enough to detect change without generating constant false alarms. A high-frequency test with excessive noise may need a better measurement method, a larger subgroup or a revised sampling interval. The goal is a decision system that supports safe action rather than producing charts no one trusts.

Signals Worth Tracking

Different chart rules are suited to different data patterns. Individual measurements are often appropriate for daily output checks, while averages or ranges can be useful when several repeated readings are collected during one session. A physicist should select the chart type after considering the sampling method and the expected distribution.

A practical dashboard might include:

• Photon and electron beam output by energy and treatment unit
• Beam symmetry, flatness and profile indicators
• Imaging and treatment isocentre coincidence
• Multileaf collimator and jaw positioning accuracy
• Couch, laser and image-guidance alignment
• Detector readings, environmental conditions and measurement uncertainty

Control limits should be recalculated carefully when a sustained process change is accepted. For example, a hardware upgrade may produce a new stable mean. Freezing the old limits indefinitely could create unnecessary alerts, while immediately accepting every shift could conceal a fault. Documented review by a qualified medical physicist provides the necessary clinical context.

Designing Control Charts For Daily Use

A control chart must fit the workflow of the department. Automated data capture from a machine log, QA platform or electronic worksheet reduces transcription errors and makes review faster. Visual displays should show the latest result, current control limits, tolerance limits and any annotations connected to service events.

Western Australian and Queensland services may operate across multiple campuses, while larger networks in Sydney or Melbourne can have several treatment units sharing a common protocol. In these settings, each machine needs its own baseline even when the same model is installed across the network. Pooled data can hide a unit-specific problem.

The chart should also state who reviews it and when. A radiation therapist may identify an unusual daily result, while the physicist determines whether a repeat measurement, machine hold or engineering callout is required. Clear escalation rules prevent a warning from sitting unnoticed in a spreadsheet.

Responding To Special Causes

An out-of-control signal is a prompt to investigate, not automatic proof of equipment failure. The first step is to verify the measurement: repeat the test, check setup and confirm that the correct energy, detector and calibration factors were used. If the result is reproducible, compare it with recent maintenance, environmental conditions and related QA tests.

Common special causes include detector damage, incorrect phantom positioning, beam steering changes, software updates, mechanical wear and interruptions to cooling or power. A single unusual point may have a straightforward explanation, whereas a gradual trend can indicate deterioration requiring planned intervention.

The response should be recorded with the finding, decision, responsible person and follow-up evidence. This creates a useful audit trail for the department and supports communication with hospital governance. A concise department report can also translate technical findings into clear information for managers and clinical colleagues.

Making SPC Work Across Australian Services

Australian departments need to adapt SPC to local staffing, procurement and regulatory arrangements. A metropolitan service with several linacs may have dedicated QA software and engineering support, while a regional hospital may rely on a smaller physics team and scheduled vendor visits. The method should remain rigorous without assuming identical resources.

Practical priorities include:

• Aligning action rules with local radiation safety and quality systems
• Recording daylight-saving changes and time-zone differences in shared databases
• Considering power, temperature and humidity effects in regional or remote sites
• Checking vendor support arrangements under the Australian market
• Training staff to recognise trends rather than chase every isolated variation

The Australian context also affects communication. A result reviewed in a morning huddle in Brisbane may need to be handed over to an afternoon team in Perth, while public-holiday rosters can delay specialist review. Standardised annotations, escalation contacts and documented coverage make the monitoring system resilient.

Professional meetings remain valuable for comparing methods across services. The 2014 COMP meeting in Banff brought together scientific presentations, business discussions, examinations and practical conference information. Its emphasis on sharing technical knowledge reflects the same principle behind effective SPC: local measurements become more useful when they are interpreted within a broader professional community.

Turning Data Into Clinical Confidence

SPC succeeds when it becomes part of routine decision-making rather than a separate statistical exercise. Monthly QA meetings can review trends, repeated warnings, corrective actions and changes in measurement uncertainty. Over time, the record helps identify which tests provide early warning and which generate little actionable information.

Communication should extend beyond the physics office. Radiation therapists need to know what a warning means for the day’s schedule, engineers need precise evidence when a service call is raised, and managers need a concise explanation of risk, downtime and resource requirements. Sharing conference news can help departments maintain professional awareness and encourage staff to discuss emerging approaches.

Build a small, well-defined charting system around high-value measurements, review it with the clinical team and refine it using real incidents and maintenance records. When statistical signals are connected to clear responsibilities, SPC turns routine machine checks into an active safeguard for consistent, accurate patient treatment.