Enhanced needle localization in ultrasound using beam steering and learning-based segmentation

• Published in: Computerized medical imaging and graphics Vol. 41; pp. 46 - 54
• Main Authors: Hatt, Charles R, Ng, Gary, Parthasarathy, Vijay
• Format: Journal Article
• Language: English
• Published: United States Elsevier Ltd 01-04-2015
• Subjects: Algorithms; Animals; Chickens; Classification; Humans; Image Enhancement - methods; Image Interpretation, Computer-Assisted - methods; Imaging, Three-Dimensional - methods; In Vitro Techniques; Injections - instrumentation; Injections - methods; Internal Medicine; Machine Learning; Needle guidance; Needles; Nerve Block - methods; Other; Pattern Recognition, Automated - methods; Regional anesthesia; Reproducibility of Results; Segmentation; Sensitivity and Specificity; Swine; Ultrasonography, Interventional - methods; Ultrasound; linear pre-segmentation method from; FI; intensity-based segmentation method; Segmentation; targeting error; R T; Needle guidance; proposed boosted classifier based segmentation method using T features; IS; filter-based segmentation method; FS; TE; beam-steered ultrasound image with enhanced needle appearance; Regional anesthesia; fused image combining needle enhanced image and B-mode image; Classification; BCS- T; NLSR; NI; rectified classification confidence using T features; ULS; Ultrasound; needle localization success rate; RT; BCS-T
• ISSN: 0895-6111; 1879-0771
• DOI: 10.1016/j.compmedimag.2014.06.016

Highlights • Beam steering is used to enhance the appearance of needles in ultrasound images. • Needle segmentation is challenging, even when using beam steering-based needle enhancement. • Needle segmentation is posed as a machine learning problem. A boosted classifier using a bank of log-Gabor wavelets is proposed for needle pixel classification. The classifier is trained on ex vivo and clinical nerve block datasets. • As a result of improved segmentation performance compared to previous methods, needle localization accuracy and robustness are increased using the proposed method.