← Back to home
Digital Twin · Oncology Monitoring

Sense. Simulate. Decide. Repeat.

A digital twin keeps a living, virtual model of a tumor's biology and treatment response in sync with imaging, liquid biopsy, and pathology — then runs the simulations a quarterly scan never has time for, so treatment decisions happen before visible progression, not after. Walked through below using HER2-positive breast cancer as the working example.

BEFORE THE LOOP STARTS

Building the Twin

TUMOR IMAGING CT · MRI · PET TUMOR GENOMICS & BIOMARKERS HER2 status · hormone receptors · tumor sequencing TREATMENT HISTORY prior regimens · surgery · radiation PERSONALIZED TUMOR TWIN baseline model of this tumor's biology & treatment response
Before any monitoring starts, the twin needs a starting shape: imaging, tumor genomics and biomarkers, and treatment history combine into a baseline virtual model. That baseline is what the continuous loop below keeps refined.
PHYSICAL — REAL PATIENT DIGITAL TWIN — VIRTUAL MODEL 1 PATIENT Lives their life — signal collected along the way liquid biopsy · symptom diary 4 TUMOR BOARD Reviews the flagged trend, not raw data oncologist · oncology nurse 2 DIGITAL TWIN Virtual tumor, updated every reading growth & treatment response model 3 SIMULATION Tests several regimens on the model 3 forecasts, ranked by risk continuous 1.4% ctDNA · 3 symptoms NEW IMAGING SCAN e.g. PET SUV 6.2 runs parallel what-if scenarios continue regimen? · switch to T-DM1? · add immunotherapy? PROGRESS- ING? progression — alert sent stable — loop continues adjusted treatment plan & coaching
Two update paths feed the same twin: a continuous stream from liquid biopsies and a symptom diary, and a separate, dashed path for whenever a new imaging scan comes back — both land on the model, not just the initial build. The twin then simulates, checks the result against progression risk every cycle, and most cycles simply loop quietly; only real progression risk reaches the tumor board.
THE SAME LOOP, ZOOMED IN

Four Stages, Up Close

1

Sense

ctDNA FRACTION 1.4% CA 15-3 34 U/mL SYMPTOM SCORE 3 / 10 ctDNA FRACTION — LAST 6 WEEKS now

One of the biweekly liquid-biopsy draws, plus a symptom diary and periodic imaging feeding the same model on a slower cadence.

2

Model

TREATMENT SENSITIVITY GROWTH RATE RECURRENCE RISK population average this tumor

The twin isn't a textbook tumor — it's calibrated to how this one tumor actually behaves.

3

Simulate

NOW · WEEK 6 continue regimen — progresses switch to T-DM1 — toward remission add immunotherapy — gradual decline

Every regimen option is tried on the model first. The full forecast is below.

4

Decide & act

ELEVATED RISK Suggest switch to T-DM1 next infusion cycle sends to app Regimen change next: T-DM1 infusion

One alert, two audiences: the tumor board sees the risk, the patient sees the plan.

STAGE 3, IN DETAIL

Where the Twin Pulls Ahead

The gauges above show why the twin's forecasts are personal. Here's the actual simulation it ran from a single week-6 ctDNA draw, testing three treatment courses against this tumor's model before anyone committed to one.

The tumor board never sees the raw branching — only the ranked outcome, so their time goes to the decision, not the data.

remission band <0.1% 3.0% 2.0% 0.8% 0% wk 1 3 6 8 10 11 wk 12 now · week 6
Actual reading Continue current regimen Switch to T-DM1 Add immunotherapy
Illustrative simulation from one liquid-biopsy draw — from the week-6 reading, the twin forecasts three plausible regimen paths so the tumor board compares outcomes, not raw numbers. Not real patient data.

Why the Loop Matters