HyperCT: Low-Rank Hypernet for Unified Chest CT Analysis

TL;DR

HyperCT introduces a parameter-efficient, hypernetwork-based framework with low-rank adaptation for unified chest CT analysis, outperforming baselines on diverse tasks.

Contribution

It presents a novel Hypernetwork approach with Low-Rank Adaptation for efficient multi-task chest CT analysis using Vision Transformers.

Findings

Outperforms strong baselines on large-scale radiological and cardiological tasks.

Provides a unified, parameter-efficient solution for holistic patient assessment.

Demonstrates effectiveness of hypernetworks with low-rank updates in medical imaging.

Abstract

Non-contrast chest CTs offer a rich opportunity for both conventional pulmonary and opportunistic extra-pulmonary screening. While Multi-Task Learning (MTL) can unify these diverse tasks, standard hard-parameter sharing approaches are often suboptimal for modeling distinct pathologies. We propose HyperCT, a framework that dynamically adapts a Vision Transformer backbone via a Hypernetwork. To ensure computational efficiency, we integrate Low-Rank Adaptation (LoRA), allowing the model to regress task-specific low-rank weight updates rather than full parameters. Validated on a large-scale dataset of radiological and cardiological tasks, \method{} outperforms various strong baselines, offering a unified, parameter-efficient solution for holistic patient assessment. Our code is available at https://github.com/lfb-1/HyperCT.