C. Lamberti

Validation of an ECG-derived respiration monitoring method

The method proposed here includes automatic optimal lead(s) selection from multiple lead systems, calculates the EDR-values from QRS-area variations and obtains a respiratory waveform through interpolation. The respiratory frequency and the presence of apnea were calculated in the time-domain, identifying the respiratory cycles on the waveform through a min-max detection method. Validation of the method has been performed using eleven patients affected mostly by obstructive respiration disorder in a sleep laboratory; gold standard respiration data was taken from oral/nasal flow, EDR was calculated using 12-lead ECG synchronized with the sleep data. Summarized results for the automatic process are: respiration cycle detection sensitivity/specificity was 98%/90% respectively. Apnea detection sensitivity/specificity was 87%/85%, respectively.

Evaluation of algorithms for real-time ECG data compression

After having analyzed the recent electrocardiogram (ECG) data compression techniques, three algorithms (TRIM, AZTEC-VT, SAPA-2), which can be implemented on a real-time ambulatory ECG recording system, have been tested. The most suitable methods for reconstruction of the compressed signal have also been investigated. Particular attention has been paid to the definition of the performance indexes. Considerations on the filtering power of the algorithms have been made, using an ECG signal synthesizer. The results show that while TRIM and SAPA-2 may be considered at the same level, AZTEC-VT does not achieve similar performance. It is convenient to give a preference to SAPA-2, for online processing, because TRIM requires the input of four dimensional parameters in order to adapt the algorithm to the signal, while SAPA-2 requires only one input parameter. It is concluded that compression algorithms, considering processing time, compression ratio and the error introduced, allow performance that is adequate for real-time data compression in new generation Holter systems.<<ETX>>

Real-time 3D echocardiographic data analysis for left ventricular volume estimation

A new computerized semi-automatic method for left ventricular chamber segmentation is presented. The left ventricle is imaged by real-time three-dimensional echocardiography. This real-time data acquisition method allows accurate evaluation of chamber size and shape, even in case there are cavities with irregular geometry. The surface detection model is based on a partial differential equation that propagates interfaces with curvature dependent speeds. The equation is solved by applying numerical methods for conservation laws. The method consists of one step only. Initial conditions are manually established on a small subset of slices of the entire volume. The solution obtained is a surface corresponding to the surface between LV cavity and LV endocardium. This mathematical model is applied to sequences of frames of human hearts (volume range: 34-109 ml) imaged by real-time 3D echocardiography. Volume estimates show an excellent correlation with those obtained by manual tracing (r=0.99).

Estimation of right ventricular volume without geometrical assumptions utilizing cardiac magnetic resonance data

Object of this study is to provide a new method to estimate right ventricular volume. The right ventricle (RV) is a complex crescent-shaped structure and, contrary to the left ventricle, it cannot be suitable with any simple geometric model existing in medical literature. To test RV volume estimation based on level set method, we performed RV endocardial surface detection on reconstructed cardiac magnetic resonance (MR) studies. Level set algorithm has been integrated into a package for 3D reconstruction of medical images, navigation and processing. The semi-automatic procedure has been applied to end-systolic (ES) and end-diastolic (ED) frames of MRI data belonging to 5 patients affected by dilative and infiltrative pathologies and 5 affected by congenital heart diseases. This study demonstrates the feasibility of accurate evaluation of RV chamber size and shape, even in case of irregular geometry.

Dynamic characterization of aorta morphology and function in presence of an aneurysm

Evaluation of aorta morphology and function in presence of aneurysms or dissection is crucial for a correct treatment choice between surgical resection and percutaneous stent-graft deployment. We developed and tested a new method for automated dynamic aorta segmentation from computed tomography (CT) images from which static and dynamic parameters of aortic morphology and function can be automatically extracted. To detect the aortic surface in a 3D domain we applied a level set segmentation scheme that incorporates gradient-based, weighted expansion and mean curvature dependent regularizers. Three subjects were imaged using a multi-detector CT scanner (Siemens, Sensation Cardiac): one normal and two patients affected by an aneurysm in the ascending and descending aorta respectively. Extracted parameters showed significant differences between them. This preliminary study proves feasibility for an accurate and dynamic aorta segmentation from which several indexes of aortic morphology and function can be automatically extracted. This may be of benefit to patients with aortic aneurysms and dissection.

Quantification of left ventricular modification in weightlessness conditions from the spatio-temporal analysis of 2D echocardiographic images

Two-dimensional echocardiography (2DE) performed during flights with a parabolic trajectory to simulate weightlessness provides a unique means to study left ventricular (LV) modifications to prevent post-flight orthostatic intolerance in astronauts. However, conventional analysis of 2DE is based on manual tracings and depends on experience. Accordingly, the aim was objectively to quantify, from 2DE images, the LV modifications related to different gravity levels, by applying a semi-automated level-set border detection technique. The algorithm validation was performed by the comparison of manual tracing results, obtained by two independent observers with 20 images, with the semi-automated measurements. To quantify LV modifications, three consecutive cardiac cycles were analysed for each gravity phase (1 Gz, 1.8 Gz, 0 Gz). The level-set procedure was applied frame-by-frame to detect the LV endocardial contours and obtain LV area against time curves, from which end-diastolic (EDA) and end-systolic (ESA) areas were computed and averaged to compensate for respiratory variations. Linear regression (y=0.91x+1.47, r=0.99, SEE:0.80 cm2) and Bland-Altman analysis (bias=−0.58 cm2, 95% limits of agreement=±2.14 cm2) showed excellent correlation between the semi-automatic and manually traced values. Inter-observer variability was 5.4%, and the inter-technique variability was 4.1%. Modifications in LV dimensions during the parabola were found: compared with 1 Gz values, EDA and ESA were significantly reduced at 1.8 Gz by 8.8±5.5% and 12.1±10.1%, respectively, whereas, during 0 Gz, EDA and ESA increased by 13.3±7.3% and 11.6±5.1%, respectively, owing to abrupt changes in venous return. The proposed method resulted in fast and reliable estimations of LV dimensions, whose changes caused by different gravity conditions were objectively quantified.

Fast interactive real-time volume rendering of real-time three-dimensional echocardiography: an implementation for low-end computers

Real-time three-dimensional echocardiography (RT3DE) is an innovative cardiac imaging modality. However, partly due to lack of user-friendly software, RT3DE has not been widely accepted as a clinical tool. The object of this study was to develop and implement a fast and interactive volume renderer of RT3DE datasets designed for a clinical environment where speed and simplicity are not secondary to accuracy. Thirty-six patients (20 regurgitation, 8 normal, 8 cardiomyopathy) were imaged using RT3DE. Using our newly developed software, all 3D data sets were rendered in real-time throughout the cardiac cycle and assessment of cardiac function and pathology was performed for each case. The real-time interactive volume visualization system is user friendly and instantly provides consistent and reliable 3D images without expensive workstations or dedicated hardware. We believe that this novel tool can be used clinically for dynamic visualization of cardiac anatomy.

Topology of optical flow in 3D echocardiography

A new approach for quantification of global strain and twist of ventricular wall is presented, it utilizes the theoretical framework of dynamical systems faking into account the large time-varying displacements that occur during the beating of the heart. Topological features of the 3D optical flow have been analysed; such features summarize the global behaviour of the motion field and can be used as cardiac function indexes. In particular the eigenvalues and eigenvectors of the vector field critical points give useful information about deformation and rotation of the entire left ventricle during cardiac cycle. The eigenvectors directions are related to the principal axis of deformation and to the rotation axis. The eigenvalues are related to deformation and rotation magnitudes, therefore they provide an estimate of the relative variation of volume and twisting for each time step.

Semi-automatic detection and tracking of mitral and aortic annuli from real-time 3D transesophageal echocardiographic images

The recently developed echocardiographic matrix array transesophageal (mTEE) transducer provides real-time 3D images of high spatial and temporal resolution that may be suitable for detailed simultaneous study of functional anatomy of the mitral and aortic valves. We developed software that detects and tracks throughout the cardiac cycle mitral and aortic annuli (MA and AoA) and tested it in 15 patients with normal valves. Following manual initialization of each annulus in 15 planes rotated around the valvepsilas axis, the position of each annulus was tracked using a two-step 3D feature tracking algorithm based on maximum likelihood and Lucas-Kanade optical flow techniques and parameters of valve geometry were automatically measured throughout the cardiac cycle. Frame-by-frame detection and tracking of MA and AoA was possible in all patients. This approach allowed for the first time non-invasive quantitative measurements of the 3D dynamic geometry of normal MA and AoA and their coupling from mTEE data.

Nearly automated left ventricular long axis tracking on real time three-dimensional echocardiographic data

The measurement of left ventricular (LV) long axis length (LAL) is an integral part of echocardiographic evaluation of LV volume, the most important determinant of LV systolic function. LAL measurements from 2D echocardiography (2DE) are highly dependent on the ability to obtain non-foreshortened LV images. Real-time 3D echocardiography (RT3DE) could potentially overcome the effects of long-axis foreshortening, but LAL measurements on RT3DE data are currently based on manual analysis and are time-consuming. We developed and tested a nearly automated method based on optical flow techniques for the measurement of the LV LAL throughout the cardiac cycle from RT3DE data. Results of comparisons on 10 patients with manual tracing on RT3DE data showed good agreement and no significant bias (r=.99, bias=-1.8 mm). The proposed method allowed fast and accurate quantification of the LAL throughout the cardiac cycle with minimal user interaction and short computational time