Digital Planning in Breast Reconstruction: What the Tools Do Not Do

Digital planning tools in breast reconstruction attract two opposite and equally unhelpful reactions: that they will shortly make part of the surgical judgement unnecessary, and that they are a marketing layer over decisions surgeons were already making perfectly well. Both misread what the tools are for.

A planning tool is an instrument for making a judgement explicit and checkable. Its value lies in what it forces to be written down, and its risk lies in how convincing its output looks.

Planning is a loop

The most common failure in digital planning is not an inaccurate model. It is a plan that is never compared with what happened.

Measure, model, decide, verify — and then the verification has to return to the model, or the loop is a line. A planning tool whose predictions are never checked against outcomes cannot improve, and neither can the surgeon using it. Worse, an unverified tool accumulates confidence without accumulating accuracy: it keeps producing output, the output keeps looking authoritative, and nothing ever contradicts it.

Verify is the step that gets skipped, because it happens months later, in a different clinic, when the plan is no longer interesting. Building it in — standardised photography at fixed intervals, volumetric assessment where available, patient-reported measures as in BREAST-Q — is what separates a planning system from a visualisation.

What each tool answers, and what it does not

Different tools address genuinely different questions, and a lot of confusion comes from treating them as one category.

The third column is the important one, and it is the column vendors are least inclined to print. A tool that lays out reconstruction routes does not choose between them. A simulator does not replace the assessment made in theatre with the tissue in view — the judgement described in reading the mastectomy flap cannot be made in advance by anything. A device library knows the geometry of a device but not what a particular flap will tolerate. A cost model prices a pathway but not the complication that was avoided, which is frequently the largest number in the comparison.

None of this makes the tools less useful. It makes them useful for something specific, which is the only way anything is useful.

Simulation and the consent conversation

Visual simulation is where the benefit and the hazard are closest together. A patient who can see a plausible representation of a proposed reconstruction understands the discussion better, asks better questions, and forms expectations grounded in something concrete. That is a real gain over a verbal description of a result the patient has never seen.

The hazard is that a rendered image is read as a promise. It is smooth, symmetrical, unbruised and unswollen; it has no scar maturation, no reabsorption, and no staging. A patient shown an image without those qualifications has been shown an outcome that nobody agreed to deliver.

The mitigations are unglamorous and they work: state explicitly what the simulation does not model, show it alongside real photographs of real results including the intermediate stages, and document what was said. Questions to ask your surgeon is written to support that conversation from the patient’s side.

Where the data actually comes from

A model is a claim about a population, and its accuracy for a given patient depends on how well that patient resembles it. A tool trained or validated on one body habitus, one reconstruction type or one surgical technique will be less accurate outside it — and it will not usually say so. Output tends to arrive at the same apparent confidence regardless of how far outside its validation range the input sits.

Which makes the same questions asked of a device in the regulatory and evidence article worth asking of a planning tool: on what population, validated how, against what outcome, and reported where. Where a tool is itself regulated as a medical device, it has a technical file and that file is answerable.

Somebody has to own the data

The loop in the first figure only closes if the verification step has an owner, and in most units it does not. Preoperative measurements are captured because they are needed for the operation. Postoperative outcomes are captured only if someone whose job it is captures them, at intervals, for patients who may have been discharged.

Which makes this an organisational question rather than a software one. The units that get value from structured planning tend to have decided three things explicitly: who enters data and when, what the minimum dataset is, and who looks at the aggregate and how often. A tool adopted without those decisions produces a database that is entered into enthusiastically for a quarter and then abandoned, which is worse than no database, because it will later be mistaken for a complete record.

The minimum dataset question is where most of the value sits. A small number of fields captured consistently on every patient is more useful than a comprehensive form captured on the interesting ones, because selection is exactly what makes a record unable to answer questions. The temptation runs the other way, since every field looks worth having at the point of designing the form.

There are also governance obligations attached to all of this — patient images and outcome data are personal data, and simulation images generated from a patient’s own photographs particularly so.

The practical case for using them anyway

Set against all that, the argument for structured planning is straightforward. It makes reasoning explicit, so it can be reviewed, taught and disagreed with. It standardises what gets recorded, which is what makes outcomes comparable across a unit rather than a matter of individual memory. It surfaces options that might not have been considered. And it gives the patient something concrete to respond to.

Those benefits are all about structure and communication, not about prediction — and a unit that adopts planning tools for the first set of reasons tends to get more from them than one that adopts them for the second.

The broader landscape is covered in med tech for breast reconstruction, the industry landscape and the future of breast reconstruction.

The short version

Planning tools make judgements explicit; they do not make them. The loop only works if verification returns to the model, and the most valuable thing any vendor can tell you about a tool is what it does not do.

This article is educational material for clinicians. Descriptions of tool categories are general; the capabilities, validation and regulatory status of any specific product are matters for its manufacturer’s documentation.