JANUS: Anatomy-Conditioned Gating for Robust CT Triage Under Distribution Shift

TL;DR

JANUS is a physiology-guided dual-stream architecture that enhances CT triage accuracy and reliability across diverse pathologies and institutional shifts by conditioning visual embeddings on macro-radiomic priors.

Contribution

The paper introduces JANUS, a novel anatomy-conditioned model that improves robustness and calibration in CT triage under distribution shift.

Findings

JANUS achieves macro-AUROC 0.88 and AUPRC 0.74 on MERLIN test set.

JANUS generalizes well to external datasets with AUROC 0.87.

JANUS reduces false positives more effectively under domain shift.

Abstract

Automated CT triage requires models that are simultaneously accurate across diverse pathologies and reliable under institutional shift. While Vision Transformers provide strong visual representations, many clinically significant findings are defined by quantitative imaging biomarkers rather than appearance alone. We introduce JANUS, a physiology-guided dual-stream architecture that conditions visual embeddings on macro-radiomic priors via Anatomically Guided Gating. On the MERLIN test set (N=5082), JANUS attains macro-AUROC 0.88 and AUPRC 0.74, outperforming all reproduced baselines. It generalizes to an external dataset N=2000; AUROC 0.87), with the largest gains on findings defined by size and attenuation as well as improved calibration on both datasets. We further quantify prediction suppression using the Physiological Veto Rate (PVR), showing that under domain shift JANUS reduces high-confidence false positives substantially more often than true positives. Together, these results are consistent with physically grounded conditioning that improves both discrimination and reliability in CT triage. Code is made publicly available at github repository https://github.com/lavsendahal/janus and model weights are at https://huggingface.co/lavsendahal/janus.