Papers by Maria Tirindelli
2018 IEEE International Conference on Robotics and Automation (ICRA), 2018
IEEE Robotics and Automation Letters, 2022
Ultrasound imaging is becoming more prevalent in clinical practice and research. To counteract th... more Ultrasound imaging is becoming more prevalent in clinical practice and research. To counteract the drawbacks of high user-dependency and difficult interpretability, ultrasound probes can be attached to robotic arms, enabling an increase in accuracy and repeatability. Currently, robotic ultrasound scans are mainly performed in a perpendicular manner. However, these scans create shadows below high-attenuation structures like bones due to acoustic shadowing, leading to an information loss in the scan. To counteract this and improve the compounding quality of robotic ultrasound scans, we introduce an ultrasound pose optimization method. In this initial work, we focus on the volume coverage of a region of interest and an acoustic shadowing reduction in this volume. Our proposed method is compared against perpendicular scans and random scans. Results show that the volume coverage sweep achieves higher coverage with the trade-off of more performed poses. In addition, the acoustic shadow reduction consistently leads to a higher coverage and confidence of the volumes when applied after a random or perpendicular scan, with a relatively small number of additional poses. Such context-aware volume scanning and path optimization can pave the path to standardized, fully automatic high-quality robotic ultrasound scans without the need for pre-acquired data and systematically reduces the occurrence of acoustic shadows.
Medical Image Computing and Computer Assisted Intervention – MICCAI 2021, 2021
Medical Ultrasound (US), despite its wide use, is characterized by artefacts and operator depende... more Medical Ultrasound (US), despite its wide use, is characterized by artefacts and operator dependency. Those attributes hinder the gathering and utilization of US datasets for the training of Deep Neural Networks used for Computer-Assisted Intervention Systems. Data augmentation is commonly used to enhance model generalization and performance. However, common data augmentation techniques, such as affine transformations do not align with the physics of US and, when used carelessly can lead to unrealistic US images. To this end, we propose a set of physics-inspired transformations, including deformation, reverb and Signal-to-Noise Ratio, that we apply on US B-mode images for data augmentation. We evaluate our method on a new spine US dataset for the tasks of bone segmentation and classification.
IEEE Robotics and Automation Letters, 2020
Spine injections are commonly performed in several clinical procedures. The localization of the t... more Spine injections are commonly performed in several clinical procedures. The localization of the target vertebral level (i.e. the position of a vertebra in a spine) is typically done by back palpation or under X-ray guidance, yielding either higher chances of procedure failure or exposure to ionizing radiation. Preliminary studies have been conducted in the literature, suggesting that ultrasound imaging may be a precise and safe alternative to X-ray for spine level detection. However, ultrasound data are noisy and complicated to interpret. In this study, a robotic-ultrasound approach for automatic vertebral level detection is introduced. The method relies on the fusion of ultrasound and force data, thus providing both "tactile" and visual feedback during the procedure, which results in higher performances in presence of data corruption. A robotic arm automatically scans the volunteer's back along the spine by using force-ultrasound data to locate vertebral levels. The occurrences of vertebral levels are visible on the force trace as peaks, which are enhanced by properly controlling the force applied by the robot on the patient back. Ultrasound data are processed with a Deep Learning method to extract a 1D signal modelling the probabilities of having a vertebra at each location along the spine. Processed force and ultrasound data are fused using both a non deep learning method and a Temporal Convolutional Network to compute the locations of the vertebral levels. The benefits of fusing force and image signals for the identification of vertebrae locations are showcased through extensive evaluation.
2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2020
In this paper we introduce the first reinforcement learning (RL) based robotic navigation method ... more In this paper we introduce the first reinforcement learning (RL) based robotic navigation method which utilizes ultrasound (US) images as an input. Our approach combines state-of-the-art RL techniques, specifically deep Q-networks (DQN) with memory buffers and a binary classifier for deciding when to terminate the task. Our method is trained and evaluated on an in-house collected data-set of 34 volunteers and when compared to pure RL and supervised learning (SL) techniques, it performs substantially better, which highlights the suitability of RL navigation for US-guided procedures. When testing our proposed model, we obtained a 82.91% chance of navigating correctly to the sacrum from 165 different starting positions on 5 different unseen simulated environments.
Uploads
Papers by Maria Tirindelli