TUS-REC2024: A Challenge to Reconstruct 3D Freehand Ultrasound Without External Tracker

TL;DR

The paper tackles trackerless 3D freehand ultrasound reconstruction by introducing the TUS-REC2024 benchmark, which provides a large, publicly available in vivo dataset and a standardized evaluation framework. It reviews the landscape of trackerless methods, presents the challenge design, and compares participating approaches across global and local reconstruction accuracy using dense displacement fields. The results highlight progress from learning-based models and ensemble strategies, while exposing persistent challenges in drift, long-sequence robustness, and cross-domain generalization. The benchmark, with openly available data and code, aims to accelerate reproducibility and spur clinically viable trackerless ultrasound solutions. Overall, TUS-REC2024 establishes a rigorous, evolving platform to drive methodological innovation and practical deployment in resource-limited settings.

Abstract

Trackerless freehand ultrasound reconstruction aims to reconstruct 3D volumes from sequences of 2D ultrasound images without relying on external tracking systems. By eliminating the need for optical or electromagnetic trackers, this approach offers a low-cost, portable, and widely deployable alternative to more expensive volumetric ultrasound imaging systems, particularly valuable in resource-constrained clinical settings. However, predicting long-distance transformations and handling complex probe trajectories remain challenging. The TUS-REC2024 Challenge establishes the first benchmark for trackerless 3D freehand ultrasound reconstruction by providing a large publicly available dataset, along with a baseline model and a rigorous evaluation framework. By the submission deadline, the Challenge had attracted 43 registered teams, of which 6 teams submitted 21 valid dockerized solutions. The submitted methods span a wide range of approaches, including the state space model, the recurrent model, the registration-driven volume refinement, the attention mechanism, and the physics-informed model. This paper provides a comprehensive background introduction and literature review in the field, presents an overview of the challenge design and dataset, and offers a comparative analysis of submitted methods across multiple evaluation metrics. These analyses highlight both the progress and the current limitations of state-of-the-art approaches in this domain and provide insights for future research directions. All data and code are publicly available to facilitate ongoing development and reproducibility. As a live and evolving benchmark, it is designed to be continuously iterated and improved. The Challenge was held at MICCAI 2024 and is organised again at MICCAI 2025, reflecting its sustained commitment to advancing this field.

Figures (23)

• Figure 1: (a): Schematic illustration of three coordinate systems: the image coordinate system, the tracker tool coordinate system, and the camera (or world) coordinate system. (b) Schematic illustration of the calibration setup for freehand ultrasound calibration, where an ultrasound probe with an attached tracker tool images a pinhead submerged in a water bath. The pinhead acts as a calibration target, allowing computation of the spatial transformation between the ultrasound image coordinate system and the tracker tool coordinate system.
• Figure 2: Annual and cumulative publications on deep learning-based trackerless freehand ultrasound reconstruction.
• Figure 3: Experimental setup for freehand ultrasound data acquisition. The setup consists of a tracked ultrasound probe, an ultrasound scanner, an optical tracker, and an acquisition laptop. The optical tracker monitors the probe's transformation during scanning, while a volunteer is scanned using predefined probe trajectories.
• Figure 4: Timeline of the TUS-REC2024 Challenge. Key milestones include the release of training data, baseline code, and validation resources, followed by the submission phase and final Challenge event at MICCAI 2024.
• Figure 5: Participant and team statistics summarising engagement across TUS-REC2024 Challenge, including 101 registered participants from 43 teams, and participation by 25 individuals grouped into 6 teams.