Computer-derived Measurements Research Articles

0738 Clinical and Physiological Relevance of Computational Studies of Obstructive Sleep Apnea: A Systematic Literature Review

Abstract Introduction Structural interventions for obstructive sleep apnea (OSA) have unpredictable success rates. Anatomically accurate computer simulations of airflow and soft tissue dynamics may be used in future virtual intervention planning tools to identify the optimal patient interventions. The objective of this study is to review the existing literature on the correlation between computer-derived biomechanical variables and clinical measures of OSA severity. Methods Scientific papers written in English that correlated the apnea-hypopnea index (AHI) with computer-derived biomechanical variables were identified by searching on the PubMed and SCOPUS databases the search phrase “sleep apnea” AND “computational fluid dynamics” OR "finite element” OR “fluid structure interaction”. Results A total of 19 articles were identified that reported correlations between computer-derived biomechanical variables and AHI, which was the metric of OSA severity reported in most studies. These studies demonstrated that several anatomic and physiologic variables correlate with OSA severity, including airspace cross-sectional areas, airspace volumes, and airflow resistance. No studies were found that correlated computer-derived dynamic measures of upper airway mechanical stability, such as tissue compliance, to OSA severity. Conclusion Computer-derived anatomic and physiologic variables may serve as useful predictors of surgical outcome or mandibular device treatment response in OSA patients. Further research is needed to test the hypothesis that virtual surgery planning based on computer-derived measures of upper airway stability can improve outcomes of OSA interventions. Support (If Any) This project was funded in part by the Advancing a Healthier Wisconsin Endowment.

Breast Density: Comparison of Chest CT with Mammography

To (a) perform a pilot study comparing radiologists' reading of breast density at computed tomography (CT) of the chest with breast density readings from mammography performed in the same patient and (b) compare a subset of these with computer-derived measurements of breast density at CT. The institutional review board waived informed consent for this HIPAA-compliant retrospective review of mammograms and chest CT scans from 206 women obtained within 1 year of each other. Two radiologists with expertise in interpreting mammographic and CT findings independently reviewed the mammograms and CT scans and classified each case into one of the four breast density types defined by the Breast Imaging Reporting and Data System of the American College of Radiology. Interreader agreements for the mammographic density types and CT density grades were determined by using the Cohen weighted κ statistic. The intrareader correlation coefficient was determined in a subset of CT images. In another subset of 40 cases, the agreement of the semiautomated computer-derived measurements of breast density with the consensus of the two radiologists was assessed. Interreader agreement was higher for the CT density grades than for the mammographic density types, with 0.79 (95% confidence interval [CI]: 0.73, 0.85) versus 0.62 (95% CI: 0.54, 0.70). The intrareader reliability of breast density grades on CT images was 0.88. The computer-derived breast density measurements agreed with those of the radiologists in 36 (90%) cases. When four cases were manually adjusted for the complex anatomy, there was agreement for all cases. Preliminary results suggest that on further validation, breast density readings at CT may provide important additional risk information on CT of the chest and that computer-derived measurements may be helpful in such assessment.

Single photon emission computed tomography myocardial perfusion imaging to detect cardiac allograft vasculopathy

Cardiac allograft vasculopathy (CAV) is a major cause of morbidity and mortality in cardiac transplant recipients. This study evaluates the usefulness of single photon emission computed tomography (SPECT) and various SPECT-derived diastolic variables to detect CAV in heart transplant patients. A retrospective review of 141 SPECT studies with corresponding coronary angiograms within 12 months was performed on 99 transplant recipients. Diastolic function was assessed using computer-derived measures of peak filling rate (PFR), time to peak filling rate (TPFR), and mean first one-third filling rate (MFR/3). Angiography identified CAV in 53 of the 141 studies (38%). Of the 53, SPECT identified 7 with reversible myocardial defects (sensitivity 13%) and stress-induced electrocardiographic evidence of ischaemia was seen in one patient (sensitivity 2%). SPECT imaging was negative in 86 of the 88 negative coronary angiograms (specificity 98%). The positive predictive value and negative predictive value were 78 and 65%, respectively. If a more stringent definition of CAV was used (≥70% stenosis), the sensitivity and specificity were unchanged (14 and 98%, respectively). There was no statistical difference in diastolic variables between patients with or without angiographic evidence of CAV in regard to PFR (3.57 ± 1.14 vs. 3.18 ± 1.21 EDV/s, P = 0.90), TPFR (149 ± 32 vs. 153 ± 43 ms, P = 0.33), or MFR/3 (1.37 ± 0.43 vs. 1.27 ± 0.42 EDV/s, P = 0.94). Adenosine stress/rest technetium-99m tetrofosmin-gated SPECT is not a sensitive test for detection of CAV in heart transplant recipients. Diastolic dysfunction, as assessed by SPECT, was not shown to be associated with development of CAV.

The inaccuracy of simple visual interpretation for measurement of carotid stenosis by arteriography

To determine intraobserver and interobserver variability of carotid arteriography interpretation as well as the reliability of simple visual interpretation (SVI) or "eyeballing" of arteriography in the measurement of internal carotid artery stenoses. Intraobserver and interobserver measurements of 200 carotid arteriograms were performed in a blinded fashion by two vascular surgeons (VS1 and VS2) using a digital caliber computer program similar to software available in catheterization laboratories. The distal normal internal carotid artery was used as a frame of reference. These computer-derived measurements were compared with previous SVI measurements, found by retrospective chart review, that were performed at the initial time of arteriography. Intraobserver agreement (VS1a vs VS1b and VS2a vs VS2b) within +/-5% using the computer program was 94% and 92%. Interobserver agreement within +/-5% using the computer program for the four possible combinations ranged from 43% to 48%. Interobserver agreement using the computer program increased to 83% to 88% for correct stenosis interpretation within +/-20%. In the 16% to 49% category (by computer measurement), SVI would have placed the stenosis in a higher category 40% to 56% of the time. Likewise, in the 50% to 79% category, comparing SVI with the four different computer caliber measurements, SVI overestimated the stenosis to the 80% to 99% category by 30% to 44%. In the 80% to 99% category, SVI overestimated lesions in 27% to 51% of the cases. All occlusions seen on SVI correlated with computer program measurements. The computer readings in many cases downgraded the degree of carotid stenosis into a lower category and in some cases, may have led to a different treatment paradigm. SVI never underestimated carotid stenosis compared with all matched computer program measurements. Compared with a method of objective measurement similar to that used in a catheterization laboratory, SVI overestimated most carotid artery stenoses. Given the coming era of carotid stenting and a renewed need for arteriography before carotid intervention, knowledge of variability and correct interpretation of carotid stenosis using available technology remains paramount to warranted treatment.

Predictive power of irregular border shapes for malignant melanomas

The Irregularity Index is a measure of border irregularity from pigmented skin lesion images. The measure attempts to quantify the degree of irregularity of the structural indentations and protrusions along a lesion border. A carefully designed study has shown that the parameters derived from the Irregularity Index were highly correlated with expert dermatologists' notion of border shape. This paper investigates the predictive power of these parameters on a set of data with known histological diagnosis. A set of 188 pigmented skin lesions (30 malignant melanomas and 158 benign lesions) was selected for the study. Their images were segmented and their border shapes were analysed by the Irregularity Index, producing four border irregularity parameters. The predictive power of these four parameters was estimated by a series of statistical tests. The mean values of the four border irregularity parameters were significantly different between the melanoma group and the benign lesion group. When using the four parameters to predict its disease status, the leave-one-out classification rate was 82.4%, and the area under the receiver operating characteristic curve was 0.77. A malignant melanoma was 8.9 times more likely to have an irregular border than a benign lesion. This study confirmed that border irregularity is an important clinical feature for the diagnosis of malignant melanoma. It also indicates that the computer-derived measures based on the Irregularity Index capture to certain extent the kind of irregularity which is exhibited by melanomas.

Spatial QRS-T angle predicts cardiac death in a clinical population

The purpose of this study was to validate the prognostic value of computer-derived measurements of the spatial alignment of ventricular depolarization and repolarization from the standard 12-lead ECG in a general medical population. Analyses were performed on the first ECG digitally recorded from 46,573 consecutive patients since 1987. QRS and T vector were synthesized by deriving XYZ leads from the 12 leads using the inverse Dower weighting matrix. Subset analyses were considered in patients with and those without standard ECG diagnoses (i.e., atrial fibrillation, Q waves, left ventricular hypertrophy, prolonged QRS duration). The main outcome measure was cardiovascular mortality. During a mean follow-up of 6 years, 4,127 cardiovascular deaths occurred. After adjusting for age, heart rate, and gender in a Cox regression analysis, spatial QRS-T angle was the most significant predictor of cardiovascular mortality, outperforming all other ECG measurements and diagnostic statements. In the subset with ECGs free of any standard diagnoses, annual cardiovascular mortality was 0.8% for normal (0-50 degrees ), 2.3% for borderline (50-100 degrees ), and 5.1% for abnormal (100-180 degrees ) QRS-T angle groups. The borderline and abnormal angle groups had 1.5- and 1.9-fold higher risk, respectively, relative to the normal QRS-T angle group after adjustment for age, gender, and heart rate. Similar results were found when patients with standard ECG diagnosis were included or compared. Spatial QRS-T angle is a significant and independent predictor of cardiovascular mortality that provides greater prognostic discrimination than any of the commonly utilized ECG diagnostic classifications.

Effect of Selective Aortic Arch Perfusion on Median Frequency and Peak Amplitude of Ventricular Fibrillation in a Canine Model

Study objective: To determine whether the computer-derived measures of median frequency or peak amplitude of ventricular fibrillation (VF), obtained by fast Fourier transform of the VF waveform, change during selective aortic arch perfusion in a canine model of cardiac arrest. Methods: Eight mongrel dogs (including 4 control animals) were sedated, intubated, catheterized, and instrumented to record the electrocardiogram (digitally at 100 Hz, filtered with a finite impulse response filter at 2 Hz), right atrial pressure, and aortic pressure during resuscitation in a model of VF-induced cardiac arrest. After 10 minutes of VF-induced arrest, cardiopulmonary resuscitation (CPR) with a mechanical chest compression device was initiated. Beginning 2 minutes later, the 4 study animals received, every 2 minutes, 45 seconds of selective aortic arch perfusion (SAAP) with autologous blood infusions under high pressure. Defibrillation was attempted after 3 minutes of CPR and every minute thereafter. Both study and control groups received standard-dose epinephrine (.01 mg/kg) every 3 minutes by means of an intraaortic catheter. The median frequency, peak amplitude, and coronary perfusion pressure (CPP) during the 5-second period just before defibrillation were obtained with the use of computer algorithms. Results: All SAAP animals and 1 control animal were resuscitated. Baseline measures of median frequency (8.4±1.5 versus 6.6±1.0 Hz) and peak amplitude (.18±.05 versus .36±.13 mV) were not different between the SAAP and control groups, respectively, at the start of CPR. SAAP infusion resulted in significant increases in the SAAP group compared with the control group: median frequency, 9.6±.4 versus 7.3±1.4 Hz; peak amplitude, .74±.21 versus .39±.15 mV; and CPP, 40.5±7.1 versus 18.0±15.0 mm Hg, respectively. Median frequency correlated with CPP ( r 2 =.67). Peak amplitude did not correlate with CPP ( r 2 =.06). Conclusion: Median frequency and peak amplitude increase with SAAP during cardiac arrest in a canine model. This method of resuscitation was reliable in allowing restoration of a stable perfusing rhythm after defibrillation. Changes in measures of peak amplitude and median frequency may reflect interventions that enhance the likelihood of successful defibrillation and may thereby offer a noninvasive means of monitoring interventions during cardiac arrest. [Barton C, Manning JE, Batson N: Effect of selective aortic arch perfusion on median frequency and peak amplitude of ventricular fibrillation in a canine model. Ann Emerg Med May 1996;27:610-616.]

Variation of movement characteristics with washing and capacitation of spermatozoa. II. Multivariate statistical analysis and prediction of sperm penetrating ability

Fifty semen samples were studied by computer-assisted semen analysis before testing in the sperm penetration assay (SPA). Twenty-one concentration and movement measurements were obtained from each sample on sperm in semen, after washing and swim-up, and again after an 18-hour capacitation period. Discriminant analysis was then used to define a function from these measurements that would classify the SPA results as above or below a 10% penetration rate. A significant function was identified using the following variables: sperm concentration and motility in semen, and mean curvilinear velocity, linearity, and amplitude of lateral head displacement of washed sperm. Movement measurements of capacitated spermatozoa were not useful predictors in this analysis. The overall accuracy of this function for predicting SPA results was 72%. These findings demonstrate that computer-derived measurements of sperm movement provide biologically useful information regarding sperm function, and, in addition, emphasize the importance of multivariate techniques in the analysis and description of human sperm motion.

Variation of movement characteristics with washing and capacitation of spermatozoa. II. Multivariate statistical analysis and prediction of sperm penetrating ability

Fifty semen samples were studied by computer-assisted semen analysis before testing in the sperm penetration assay (SPA). Twenty-one concentration and movement measurements were obtained from each sample on sperm in semen, after washing and swim-up, and again after an 18-hour capacitation period. Discriminant analysis was then used to define a function from these measurements that would classify the SPA results as above or below a 10% penetration rate. A significant function was identified using the following variables: sperm concentration and motility in semen, and mean curvilinear velocity, linearity, and amplitude of lateral head displacement of washed sperm. Movement measurements of capacitated spermatozoa were not useful predictors in this analysis. The overall accuracy of this function for predicting SPA results was 72%. These findings demonstrate that computer-derived measurements of sperm movement provide biologically useful information regarding sperm function, and, in addition, emphasize the importance of multivariate techniques in the analysis and description of human sperm motion.

Validation of a computerized edge detection algorithm for quantitative two-dimensional echocardiography.

An edge detection algorithm used in conjunction with digitized two-dimensional echocardiograms was applied to validate computerized two-dimensional echocardiographic (2DE) quantitation of cross-sectional areas of canine left ventricular chambers. Images were enhanced by space-time smoothing and dynamic range expansion, after which automatic edge detection was performed by convolving a Laplacian operator with the enhanced image. In an in vitro study of 29 myocardial slabs, computer-derived 2DE measurements of short-axis sections of the left ventricle were compared with manually derived 2DE data and validated against direct measurements of intraluminal areas of myocardial slabs. Correlations of both manually and computer-derived 2D echocardiograms vs direct measurements were equally satisfactory (r = .95 for both). Computer-derived measurements of perimeters tended to underestimate actual perimeters of the endocardial outlines of left ventricular sections. In 13 closed-chest anesthetized dogs, manually and computer-derived left ventricular short-axis areas measured by 2DE techniques showed a good correlation at both end-diastole (r = .91) and end-systole (r = .92). Left ventricular volumes reconstructed from 2DE images were compared with angiographically determined volumes. The computer-enhanced 2DE method correlated against angiography, with r = .93 for end-diastolic and r = .93 for end-systolic volumes. Left ventricular volume correlations between manually and computer-derived 2D echocardiograms were satisfactory, with r = .87 for end-diastole and r = .87 for end systole. We conclude that computerized enhancement and edge detection of 2D echocardiograms obtained in dogs provided accurate analysis of actual left ventricular cross-sectional areas and left ventricular volumes.