Curriculum Learning for Few-Shot Domain Adaptation in CT-based Airway Tree Segmentation

TL;DR

This work proposes integrating Curriculum Learning into airway segmentation networks, distributing the training set into batches according to ad-hoc complexity scores derived from CT scans and corresponding ground-truth tree features, and investigates few-shot domain adaptation.

Abstract

Despite advances with deep learning (DL), automated airway segmentation from chest CT scans continues to face challenges in segmentation quality and generalization across cohorts. To address these, we propose integrating Curriculum Learning (CL) into airway segmentation networks, distributing the training set into batches according to ad-hoc complexity scores derived from CT scans and corresponding ground-truth tree features. We specifically investigate few-shot domain adaptation, targeting scenarios where manual annotation of a full fine-tuning dataset is prohibitively expensive. Results are reported on two large open-cohorts (ATM22 and AIIB23) with high performance using CL for full training (Source domain) and few-shot fine-tuning (Target domain), but with also some insights on potential detrimental effects if using a classic Bootstrapping scoring function or if not using proper scan sequencing.

Figures (4)

• Figure 1: Proposed CT scoring functions.
• Figure 2: Source2Target strategy to form $\mathbf{CL\,Batches}$.
• Figure 3: Complexity scores on ATM22 and AIIB23: Bootstrapping and our ML-based scoring functions are compared with histograms (top) and for ordering of CT scans (bottom).
• Figure 4: Qualitative results using 2 CL strategies on the Target domain.