Confirm your Synapse account is Certified. Docker push and submission fail silently if your account is not a Certified User, with no error message explaining why.

Check Account Settings → Trust & Credentials. If it shows "Incomplete," take the certification quiz (roughly 15 minutes) before doing anything else below.

Submission is not made via a button on the Challenge page — there isn't one. Docker submissions are initiated from your own Synapse project, not the challenge project:

1. Push your image to Synapse's Docker registry: docker.synapse.org
2. Go to your project's Docker tab
3. Click Docker Repository Tools → Submit Docker Repository to Challenge
4. Select the evaluation queue matching your task

Submit only to the queue(s) for the task(s) you are entering. There is no validation leaderboard or interim feedback; submitted containers are run against the hidden test set by the organisers after the deadline closes.

Prepare your predictions and container. Both the prediction format and the container harness contract are now final and published.

Prediction format. The scoring code in the evaluation repository defines the exact CSV your model must produce per task, keyed by (patient_id, timepoint), covering every exam in the ground-truth set:

github.com/EchoRisk-MICCAI/echorisk-benchmark/tree/master/evaluation

• Task 1: lvef_pred, float in [0, 100]
• Task 2: dysfunction_pred, calibrated probability in [0, 1], not a hard label
• Task 3: cardiotoxicity_pred, calibrated probability in [0, 1], baseline (T1) only

Run the official scorer locally on the validation set before packaging, so your offline numbers match what the leaderboard will compute:

Container harness contract. Your container must read a manifest and DICOM files mounted read-only at /input, and write a single /output/predictions.csv covering every case in the manifest, using no arguments and no runtime environment variables. Full details, exact directory structure, and a working reference container are here:

• Harness specification (HARNESS_SPEC.md)
• Working reference container (submission-example/)

Evaluation hardware (confirmed): NVIDIA A40, 48 GB VRAM · 32 CPU cores · 48 GB system RAM (40 GB available per container run) · 1 TB shared storage. Wall-clock limit: 15 minutes per case, timed from container start — this includes model-loading time, not just active inference.

Test offline before submitting. Per Challenge Rule 6, containers must run fully automatically with no manual intervention, no internet access, and no external API calls at inference time. Validate this locally before submission:

This is the exact invocation used at evaluation time, not an approximation. --gpus '"device=0"' pins a single GPU device — do not write code that enumerates or selects among multiple devices. --network none reproduces the offline evaluation environment; if your container depends on a runtime download (model weights, tokeniser files, API calls), it will fail evaluation. --shm-size 512m raises Docker's default 64 MB shared memory pool — if your inference pipeline uses PyTorch (or similar) multi-process data loading, test against this exact value locally, since a larger pool on your own machine may hide a crash that only appears under the evaluation server's actual limit. The mount paths shown are local placeholders for your own testing — inside the container, your code must always read from /input and write to /output.

Submit via your project's Docker tab using the steps in Step 2 above — push to docker.synapse.org, then Docker Repository Tools → Submit Docker Repository to Challenge. You may resubmit at any point before the deadline to update your container; only your last submission is evaluated, consistent with Challenge Rule 2.

Only one designated submitter per team. Submission access is granted to a single contact per team, matching the Synapse Profile URL supplied on your registration form. Only that person needs access — your team's official entry is whichever container they submit from their own project.

If your designated submitter has not yet accepted the Synapse Team invitation granting this access, check Synapse notifications and accept it as soon as possible. Submission is not possible until this is done.

To change who your designated submitter is, email echorisk.miccai@gmail.com with the new Synapse Profile URL.

Recommended naming convention (optional). Organisers recommend naming Docker repositories and submissions using the format TeamName_TaskX_vN (e.g. CardioVisionLab_Task1_v3). This is a recommendation for clarity only and is not enforced.

Deadline and what happens after. The submission window closes on 20 August 2026. After the deadline, organisers run all final containers against the hidden test set. No test-set metrics are visible to participants before or during this process. Results are released on 27 September 2026 during the MICCAI Annual Meeting.