SAMe: A Semantic Anatomy Mapping Engine for Robotic Ultrasound

2026-04-28

AI Summary

Current robotic ultrasound systems lack the anatomical understanding necessary to autonomously identify scanning targets and determine initial probe poses, often requiring expert intervention. This work proposes SAMe (Semantic Anatomical Mapping engine), which introduces, for the first time, an explicit, lightweight, and control-compatible anatomical prior layer that enables end-to-end mapping from ambiguous clinical complaints to a six-degree-of-freedom initial probe pose. The method constructs a patient-specific anatomical representation from a single surface image, eliminating the need for preoperative CT or MRI registration. Real-robot experiments demonstrate initialization success rates of 97.3% for liver and 81.7% for kidney scans, significantly outperforming surface-based heuristic baselines.

Abstract

Robotic ultrasound has advanced local image-driven control, contact regulation, and view optimization, yet current systems lack the anatomical understanding needed to determine what to scan, where to begin, and how to adapt to individual patient anatomy. These gaps make systems still reliant on expert intervention to initiate scanning. Here we present SAMe, a semantic anatomy mapping engine that provides robotic ultrasound with an explicit anatomical prior layer. SAMe addresses scan initiation as a target-to-anatomy-to-action process: it grounds under-specified clinical complaints into structured target organs, instantiates a patient-specific anatomical representation for the grounded targets from a single external body image, and translates this representation into control-facing 6-DoF probe initialization states without any additional registration using preoperative CT or MRI. The anatomical representation maintained by SAMe is explicit, lightweight (single-organ inference in 0.08s), and compatible with downstream control by design. Across semantic grounding, anatomical instantiation, and real-robot evaluation, SAMe shows strong performance across the full initialization pipeline. In real-robot experiments, SAMe achieved overall organ-hit rates of 97.3% for liver initialization and 81.7% for kidney initialization across the evaluated target sets. Even when restricted to the centroid target, SAMe outperformed the surface-heuristic baseline for both liver and kidney initialization. These results establish an explicit anatomical prior layer that addresses scan initialization and is designed to support broader downstream autonomous scanning pipelines, providing the anatomical foundation for complaint-driven, anatomically informed robotic ultrasonography.

Problem

• robotic ultrasound
• anatomical understanding
• scan initialization
• patient-specific anatomy
• autonomous scanning

Innovation

• semantic anatomy mapping
• robotic ultrasound
• anatomical prior
• scan initialization
• patient-specific modeling