SonoSelect: Efficient Ultrasound Perception via Active Probe Exploration

2026-04-07

AI Summary

This work addresses the inefficiency and redundancy in ultrasound diagnosis caused by blind acquisition of multi-view images and the absence of intelligent guidance toward probe positions with maximal information gain. The authors formulate active view exploration in ultrasound as a sequential decision-making problem, dynamically fusing 2D images into a 3D spatial memory and introducing an ultrasound-specific exploration objective that jointly optimizes organ coverage, reconstruction uncertainty, and scanning redundancy. By integrating 3D spatial memory, a sequential decision mechanism, and a tailored exploration strategy, the proposed method enables adaptive probe motion. Experiments demonstrate that only two views suffice to achieve high performance in multi-view organ classification in simulation, while in renal cyst detection, the method attains kidney and cyst coverage rates of 54.56% and 35.13%, respectively, with trajectories precisely focused on lesions.

Abstract

Ultrasound perception typically requires multiple scan views through probe movement to reduce diagnostic ambiguity, mitigate acoustic occlusions, and improve anatomical coverage. However, not all probe views are equally informative. Exhaustively acquiring a large number of views can introduce substantial redundancy, increase scanning and processing costs. To address this, we define an active view exploration task for ultrasound and propose SonoSelect, an ultrasound-specific method that adaptively guides probe movement based on current observations. Specifically, we cast ultrasound active view exploration as a sequential decision-making problem. Each new 2D ultrasound view is fused into a 3D spatial memory of the observed anatomy, which guides the next probe position. On top of this formulation, we propose an ultrasound-specific objective that favors probe movements with greater organ coverage, lower reconstruction uncertainty, and less redundant scanning. Experiments on the ultrasound simulator show that SonoSelect achieves promising multi-view organ classification accuracy using only 2 out of N views. Furthermore, for a more difficult kidney cyst detection task, it reaches 54.56% kidney coverage and 35.13% cyst coverage, with short trajectories consistently centered on the target cyst.

Problem

• ultrasound perception
• active probe exploration
• view redundancy
• anatomical coverage
• diagnostic ambiguity

Innovation

• active view exploration
• ultrasound perception
• sequential decision-making
• 3D spatial memory
• adaptive probe guidance