Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep Kernel and Image Quality Estimators for Optimizing Robotic Ultrasound Controller using Bayesian Optimization

Deepak Raina, SH Chandrashekhara, Richard Voyles, Juan Wachs, Subir Kumar Saha

TL;DR

This work tackles autonomous robotic ultrasound optimization under high‑dimensional probe control by introducing a learned deep kernel for Gaussian processes to enable sample‑efficient Bayesian optimization of six‑dimensional probe poses. Two image quality estimators based on deep CNNs, one for classification and one for segmentation, provide real‑time quality feedback to guide optimization. The approach is validated on three urinary bladder phantoms, demonstrating over 50% gains in sample efficiency and robust inter‑patient adaptability compared with traditional kernels. The results highlight practical potential for fully autonomous ultrasound procedures and suggest avenues for extending to more complex anatomy and in‑vivo validation. Overall, the framework combines data‑driven embedding, image‑quality feedback, and reinforcement of high‑quality imaging to streamline autonomous robotic ultrasound control.

Abstract

Ultrasound is a commonly used medical imaging modality that requires expert sonographers to manually maneuver the ultrasound probe based on the acquired image. Autonomous Robotic Ultrasound (A-RUS) is an appealing alternative to this manual procedure in order to reduce sonographers' workload. The key challenge to A-RUS is optimizing the ultrasound image quality for the region of interest across different patients. This requires knowledge of anatomy, recognition of error sources and precise probe position, orientation and pressure. Sample efficiency is important while optimizing these parameters associated with the robotized probe controller. Bayesian Optimization (BO), a sample-efficient optimization framework, has recently been applied to optimize the 2D motion of the probe. Nevertheless, further improvements are needed to improve the sample efficiency for high-dimensional control of the probe. We aim to overcome this problem by using a neural network to learn a low-dimensional kernel in BO, termed as Deep Kernel (DK). The neural network of DK is trained using probe and image data acquired during the procedure. The two image quality estimators are proposed that use a deep convolution neural network and provide real-time feedback to the BO. We validated our framework using these two feedback functions on three urinary bladder phantoms. We obtained over 50% increase in sample efficiency for 6D control of the robotized probe. Furthermore, our results indicate that this performance enhancement in BO is independent of the specific training dataset, demonstrating inter-patient adaptability.

Deep Kernel and Image Quality Estimators for Optimizing Robotic Ultrasound Controller using Bayesian Optimization

TL;DR

This work tackles autonomous robotic ultrasound optimization under high‑dimensional probe control by introducing a learned deep kernel for Gaussian processes to enable sample‑efficient Bayesian optimization of six‑dimensional probe poses. Two image quality estimators based on deep CNNs, one for classification and one for segmentation, provide real‑time quality feedback to guide optimization. The approach is validated on three urinary bladder phantoms, demonstrating over 50% gains in sample efficiency and robust inter‑patient adaptability compared with traditional kernels. The results highlight practical potential for fully autonomous ultrasound procedures and suggest avenues for extending to more complex anatomy and in‑vivo validation. Overall, the framework combines data‑driven embedding, image‑quality feedback, and reinforcement of high‑quality imaging to streamline autonomous robotic ultrasound control.

Abstract

Ultrasound is a commonly used medical imaging modality that requires expert sonographers to manually maneuver the ultrasound probe based on the acquired image. Autonomous Robotic Ultrasound (A-RUS) is an appealing alternative to this manual procedure in order to reduce sonographers' workload. The key challenge to A-RUS is optimizing the ultrasound image quality for the region of interest across different patients. This requires knowledge of anatomy, recognition of error sources and precise probe position, orientation and pressure. Sample efficiency is important while optimizing these parameters associated with the robotized probe controller. Bayesian Optimization (BO), a sample-efficient optimization framework, has recently been applied to optimize the 2D motion of the probe. Nevertheless, further improvements are needed to improve the sample efficiency for high-dimensional control of the probe. We aim to overcome this problem by using a neural network to learn a low-dimensional kernel in BO, termed as Deep Kernel (DK). The neural network of DK is trained using probe and image data acquired during the procedure. The two image quality estimators are proposed that use a deep convolution neural network and provide real-time feedback to the BO. We validated our framework using these two feedback functions on three urinary bladder phantoms. We obtained over 50% increase in sample efficiency for 6D control of the robotized probe. Furthermore, our results indicate that this performance enhancement in BO is independent of the specific training dataset, demonstrating inter-patient adaptability.
Paper Structure (19 sections, 11 equations, 9 figures, 2 tables)

This paper contains 19 sections, 11 equations, 9 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Autonomous Robotic Ultrasound (A-RUS) systems
  4. Bayesian optimization for ultrasound robots
  5. High-Dimensional Bayesian Optimization
  6. Methodology
  7. Bayesian optimization
  8. Deep Kernel
  9. Acquisition Function
  10. Image quality estimation:
  11. Image classification
  12. Image segmentation
  13. Dataset
  14. Robotic controller
  15. Results and Discussions
  16. ...and 4 more sections

Figures (9)

  • Figure 1: Robotic ultrasound system with probe attached to its end-effector raina2021comprehensive, conducting an in-vivo ultrasound.
  • Figure 2: Overview of the Bayesian optimization (BO) framework for optimizing 6D robotic ultrasound controller using deep kernels in Gaussian process model and image quality estimators as feedback.
  • Figure 3: Dataset and neural network structure for Deep Kernel
  • Figure 4: D-CNNs for image quality estimators
  • Figure 5: Experimental setup of robotic ultrasound system with three different phantoms of the urinary bladder.
  • ...and 4 more figures