3D Ultrasound Imaging of Residual Limbs With Camera-Based Motion Compensation

• Main Authors: Ranger, Bryan J. (Author), Feigin, Micha (Author), Zhang, Xiang (Author), Moerman, Kevin M. (Author), Herr, Hugh M. (Author), Anthony, Brian W. (Author)
• Other Authors: Massachusetts Institute of Technology. Institute for Medical Engineering & Science (Contributor), Massachusetts Institute of Technology. Media Laboratory (Contributor)
• Format: Article
• Language: English
• Published: Institute of Electrical and Electronics Engineers (IEEE), 2020-04-08T15:43:05Z.
• Subjects: Article
• Online Access: Get fulltext

 Ultrasound is a cost-effective, readily available, and non-ionizing modality for musculoskeletal imaging. Though some research groups have pursued methods that involve submerging the transducer and imaged body segment into a water bath, many limitations remain in regards to acquiring an unloaded volumetric image of an entire human limb in a fast, safe, and adequately accurate manner. A 3D dataset of a limb is useful in several rehabilitative applications including biomechanical modeling of soft tissue, prosthetic socket design, monitoring muscle condition and disease progression, bone health, and orthopedic surgery. This paper builds on previous work from our group and presents the design, prototyping, and preliminary testing of a novel multi-modal imaging system for rapidly acquiring volumetric ultrasound imagery of human limbs, with a particular focus on residual limbs for improved prosthesis design. Our system employs a mechanized water tank setup to scan a limb with a clinical ultrasound transducer and 3D optical imagery to track motion during a scan. The iterative closest point algorithm is utilized to compensate for motion and stitch the images into a final dataset. The results show preliminary 2D and 3D imaging of both a tissue-mimicking phantom and residual limbs. A volumetric error compares the ultrasound image data obtained to a previous MRI method. The results indicate potential for future clinical implementation. Concepts presented in this paper could reasonably transfer to other imaging applications such as acoustic tomography, where motion artifact may distort image reconstruction. ©2019