Beyond the
Last-Frame
Idiap Research Institute · MedAI

Beyond theLast-Frame

Teaching a retinal-scan AI to read time — not just the last snapshot.

A graduation project on temporal deep learning for the automated grading of retinal inflammation in fluorescein angiography.

AuthorTymo van Rijn
Student no.1057297
HostIdiap — Martigny, CH
ProgrammeHBO-ICT · Software Eng.
The premise

A film,not a photo.

A fluorescein angiography exam is closer to a short film than a photograph. A dye is injected and tracked through the retinal vasculature over ten to fifteen minutes — and clinicians diagnose by how the brightness moves, not how it looks at one instant.

The deep-learning pipeline at Idiap’s MedAI group treated each exam as a single static image — the last frame. For several disease patterns, that discards the entire basis of the diagnosis. This project set out to fix that.

Grows & blurs

Leakage

Dye escapes from vessels; the bright area grows and blurs as the sequence runs.
Fills & sharpens

Pooling

Dye collects in a defined space — often with a sharper border than leakage.
Brightens

Staining

Tissue takes up dye and brightens — without the expanding pattern of leakage.
Stays stable

Window defect

Brighter signal appears early and stays geographically stable throughout.
The problem, in one line

Current FA classification pipelines insufficiently utilise the temporal information present in the examination — limiting diagnostic performance, robustness and interpretability.

The work5 findings

What thisproject found.

Each is a self-contained piece of the work — open any one to go as deep as you like. Together they trace the project from raw data to a working temporal pipeline.

State of play

For whoevercontinues it.

Single-frame baseline0.82

ROC-AUC — last frame only (prior work)

Temporal model — max-pooled tokens0.86

Test AUC — state at time of writing

Honest framing: this is a working method and pipeline, not a finished clinical product. The numbers are real but provisional. The value for whoever picks this up is the groundwork — a re-runnable benchmark, a calibrated method, and 0 validated frame timestamps added to the dataset.

Delivered
  • A working, MedNet-integrated temporal evaluation pipeline.
  • A reproducible LDA phase-calibration probe and separability benchmark.
  • 32,844 validated frame timestamps, reusable by the whole group.
  • A two-layer automated unit-test suite.
Open — for the next phase
  • The explicit time-embedding track (the recommended next direction).
  • HOJG-scale external validation — the key open research question.
  • Integration with production clinical workflows.

15-minute overview

Watch the Idiap TAM presentation

A recorded, institute-wide talk walking through the whole project end to end.

YouTube ↗

Thank you fortaking the time.

A real research contribution — built carefully, reviewed honestly, and made reusable for the team that continues it.

AuthorTymo van Rijn
Student no.1057297
HostIdiap — MedAI
Year2025 — 2026