Automated ultrasound image acquisition to aid criterion-based examiner training and reliability testing
Harris-Love MO, Ismail C, Monfaredi R, Hernandez HJ, Pennington D, Woletz P, McIntosh V, Adams B, Blackman MR, 2016 (doi:10.7717/peerj.2146)
Our group has utilized automated B-mode image acquisition featuring a portable ultrasound unit (SonoSite Titan M-Turbo) with a 6 MHz linear array sound transducer with a custom interface integrated with a load cell (FC22 Compression Load Cell; 0–44.48 ± 0.45 N). Automated image acquisition and sound transducer positioning were performed with the Kuka light weight arm (LWA) robot (7 degrees of freedom; motion error, ±0.05 mm) to attain uniform force and angle targets.
In determining the interrater reliablity of our novice and experienced staff using force feedback augmented imaging, no differences in material dimension measures were detected among the 6 examiners. Additionally, there was a high degree of association among the material thickness values obtained by the examiners and the Kuka LWA robot (R squared = .86-.97, p < .001). The initial data suggests that sonography with augmented force feedback allows examiners with varied experience levels to exhibit high interrater reliability.
Changes in serial sonographic image characteristics based on examiner force and sound transducer orientation
Harris-Love MO, Monfaredi R, Ismail C, Blackman MR, Cleary K, 2014. (doi:10.3389/fnagi.2014.00172)
The top panels of the figure (A–C) depict transverse views of a muscle tissue mimetic phantom with a progressive magnitude of stress imposed on the phantom surface by the sound transducer. The material deformation (thickness, centimeter) secondary to the stress progression was as follows: (A) 3.78 cm, (B) 3.45 cm, and (C) 3.21 cm.
The bottom panels (D–F) depict similar sonographic views as the preceding panels. The echointensity observed in the serial images is based on a progressively increasing cranial/caudal tilt angle of the sound transducer applied to the phantom surface. The changes in echointensity (grayscale, unitless, 0–255) secondary to the angle progression were as follows: (D) 56.64, (E) 48.10, and (F) 36.90. (All images were acquired using a 6 MHz linear array sound transducer and a muscle mimetic phantom with anechoic gel via automated image capture by the Kuka LWA robot.)
These serial images illustrate that progressive shifts in cranial/caudal tilting of 10° resulted in a >15% decrease in echointensity, and suggest that force feedback may enhance the consistency of manual image acquisition.
Kuka LWA robot provided courtesy of the Sheikh Zayed Institute for Pediatric Surgical Innovation at Children's National Hospital
Understanding how the reliability of rehabilitative ultrasound is affected by the method used to select of the region of interest
Harris-Love MO, Seamon BA, Teixeira C, Ismail C., 2016. (doi.org/10.7717/peerj.1721)
The reliability of quantitative methods used during rehabilitative ultrasound procedures may be affected by many factors after image acquisition. We examined the impact of two image analysis factors: 1) the method used to select the region of interest (ROI), and 2) the software platform used for the grayscale histogram analyses.
The Bland-Altman plots in the figure depict the agreement between the grayscale measurements across the two image analysis platforms for Examiner 1. The limits of agreement for grayscale measures obtained using the Free Hand method (FHT) for ROI selection were wider than comparable measures obtained using the Rectangular Marquee Tool (RMT) method. The mean difference between the echogenicity estimates obtained with the RMT and FHT methods was .87 grayscale levels (95% CI [.54–1.21]; p < .0001) using data obtained with both Photoshop and ImageJ.
The RMT method facilitates a rapid selection of the region of interest, and use of the ImageJ software platform yielded lower measurement errors. However, the measurement differences were minimal between the methods, and the FHT method may be advantageous for muscles or tissues with irregular boundaries.
(RMT, Rectangular Marquee Tool; FHT, Free Hand Tool; diff., difference; grayscale level range = 0–255.)