A pilot study of video-motion analysis in endovascular surgery: development of real-time discriminatory skill metrics.

OBJECTIVES: Accurate assessment and credentialing of physicians is essential. Objective motion analysis of guide-wire/catheter manipulation to assess proficiency during endovascular interventions remains unexplored. This study aims to assess its feasibility and its role in evaluation of technical ability. MATERIALS AND METHODS: A semi-automated catheter-tracking software was developed which allows for frame-by-frame motion analysis of fluoroscopic videos and calculation 2D catheter tip path-length. 21 interventionalists (6 cardiologists, 8 interventional radiologists, 7 vascular surgeons; 14/21 had performed >500 endovascular procedures) performed an identical carotid artery stenting procedure (CAS) on a VIST simulator (Mentice, Gothenburg, Sweden). Operators were sub-divided into four categories according to CAS experience: 6 inexperienced (0 CAS-group A), 3 low-volume (1-20 CAS-group B), 5 moderate-volume (21-50 CAS-group C) and 7 high-volume (>50 CAS-group D) CAS experience. Total PL was calculated for each case and comparisons made between groups. PL was correlated with: quantitative, simulator-derived metrics and qualitative performance scores (generic and procedure-specific) derived from post-hoc video analysis by three blinded observers. RESULTS: Group D used 5160.3 (inter-quartile range- IQR 4046.4-7142.9) pixels of movement, compared to 6856.7 (5914.4-8106.9) for group A (p = 0.046); 10,905.1 (7851.1-14,381.5) for group B (p = 0.017); and 9482.6 (8663.5-13,847.6) for group C (p = 0.003). Statistically significant inverse correlations were seen between total PL and qualitative performance scores (rho = -0.519 for generic (p = 0.027) rho = -0.567 for procedure-specific (p = 0.014) scores). PL did not correlate with any of the simulator-derived metrics (errors, contrast volume, total procedure and fluoroscopy times, cine-loops used). CONCLUSION: Endovascular instrument video motion analysis is feasible and may represent a valuable tool for the objective assessment of endovascular skill.

High-resolution Doppler-velocity estimation techniques for processing of coherent heterodyne pulsed lidar data.

On the basis of an analysis of the autocovariance of the complex heterodyne signal, some novel algorithms are derived and are investigated for use in determining, with high spatial resolution, Doppler-velocity coherent-lidar profiles in the case of rectangular and rectangularlike sensing laser pulses. These algorithms generalize other known Doppler-velocity estimators for the more complex case of nonuniform scattering and Doppler-velocity distribution within the pulse length. Algorithm performance and efficiency are studied and are illustrated by computer simulations. It is shown that the Doppler-velocity profiles can be determined with essentially better resolution in comparison with the use of other known estimation approaches, but at the expense of some increase in the number of statistical realizations (number of laser shots) required to reduce the speckle-noise effect. The minimum achievable resolution interval is shown to be much shorter than the pulse length.

Using the spectral asymmetry of TEA CO2 laser pulses to determine the Doppler-shift sign in coherent lidars with low frequency stability.

We develop a method for determinating the relative positions of the lidar transmitter (LT) and the local oscillator (LO) frequencies in Doppler CO2 lidars. It uses the weak spectral asymmetry of TEA CO2 laser pulses, defined by a number of secondary peaks at the high-frequency side of the main spectrum peak. Depending on the sign of the beat frequency, these peaks may appear in the demodulated spectrum at either the high- or the low-frequency side. Each laser pulse spectrum is compared with reference spectra with two types of asymmetry, with the cross-correlation coefficients used as criteria. The performance of the method at different values of signal-to-noise ratio is analyzed numerically. The method is also applied to raw data from the lidar reference channel and demonstrates good performance at noise levels lower than the secondary peaks in the pulse spectrum or at a signal-to-noise ratio of > or = 20 dB. Application of the pulse spectrum asymmetry for lidar frequency stabilization is analyzed. Lidar operation without frequency stabilization is considered as well. The method offers a simple Doppler lidar hardware for the creation of low-cost coherent lidars, velocimeters-rangefinders, etc.

Reduction of the pulse spike-cut error in Fourier-deconvolved lidar profiles.

A simple approach is analyzed and applied to the National Oceanic and Atmospheric Administration (NOAA) Doppler lidar data to reduce the error in Fourier-deconvolved lidar profiles that is caused by spike-cut uncertainty in the laser pulse shape, i.e., uncertainty of the type of not well-recorded (cut, missed) pulse spikes. Such a type of uncertainty is intrinsic to the case of TE (TEA) CO(2) laser transmitters. This approach requires only an estimate of the spike area to be known. The result from the analytical estimation of error reduction is in agreement with the results from the NOAA lidar data processing and from computer simulation.

Lidar receivers without overlap of the photomultiplier's single pulses.

Lidar receivers of linear performance without overlap of the single pulses are analyzed. The statistics of output signals are investigated by using a model of the photoreceiving system that is based on the conversion of secondary electron trains into single-frequency decayed oscillations. I show that the lidar profile can be separated into nonuniform and uniform parts and a background and sampled by low-speed analog-to-digital converters (12-16 bits/0.01-100 kHz) to provide high amplitude and temporal resolution. The nonstationary background is reduced without chopping to the square root of its intensity. The use of the lidar receivers is discussed.