An Integrated Machine Learning Approach for Congestive Heart Failure Prediction.

Congestive heart failure (CHF) is one of the primary sources of mortality and morbidity among the global population. Over 26 million individuals globally are affected by heart disease, and its prevalence is rising by 2% yearly. With advances in healthcare technologies, if we predict CHF in the early stages, one of the leading global mortality factors can be reduced. Therefore, the main objective of this study is to use machine learning applications to enhance the diagnosis of CHF and to reduce the cost of diagnosis by employing minimum features to forecast the possibility of a CHF occurring. We employ a deep neural network (DNN) classifier for CHF classification and compare the performance of DNN with various machine learning classifiers. In this research, we use a very challenging dataset, called the Cardiovascular Health Study (CHS) dataset, and a unique pre-processing technique by integrating C4.5 and K-nearest neighbor (KNN). While the C4.5 technique is used to find significant features and remove the outlier data from the dataset, the KNN algorithm is employed for missing data imputation. For classification, we compare six state-of-the-art machine learning (ML) algorithms (KNN, logistic regression (LR), naive Bayes (NB), random forest (RF), support vector machine (SVM), and decision tree (DT)) with DNN. To evaluate the performance, we use seven statistical measurements (i.e., accuracy, specificity, sensitivity, F1-score, precision, Matthew's correlation coefficient, and false positive rate). Overall, our results reflect our proposed integrated approach, which outperformed other machine learning algorithms in terms of CHF prediction, reducing patient expenses by reducing the number of medical tests. The proposed model obtained 97.03% F1-score, 95.30% accuracy, 96.49% sensitivity, and 97.58% precision.

Effect of vascular volume and edema on wave propagation in canine lungs.

The velocities of longitudinal and transverse stress waves transmitted through inflated lung parenchyma depend on the lung stiffness, as defined by the bulk and shear moduli, and the lung density. We examined the relationship between stress wave velocities and lung density. A saline-filled reservoir was connected to the vessels of caudal dog lobes held inflated at 5 cmH2O transpulmonary pressure, and vascular volume and extravascular lung water were increased in steps by increasing vascular pressure. At each step, we measured the transmitted signals at locations 2 and 7 cm from an impulse surface distortion by means of microphones embedded in the lung surface. Longitudinal and transverse wave velocities were computed by using cross-correlation analysis of microphone signal pairs. Both wave velocities decreased as lung density increased: as a first approximation, wave velocities were inversely proportional to the square root of lung density. This behavior is consistent with the propagation of small-amplitude stress waves through an elastic continuum. Estimated bulk and shear moduli were 26 and 3.5 cmH2O, respectively, and were consistent with results from quasi-static deformation tests.

Propagation of stress waves in inflated sheep lungs.

If the lung is an elastic continuum, both longitudinal and transverse stress waves should be propagated in the medium with distinct velocities. In five isolated sheep lungs, we investigated the propagation of stress waves. The lungs were degassed and then inflated to a constant transpulmonary pressure (Ptp). We measured signals transmitted at locations approximately 1.5, 6, and 11 cm from an impulse surface distortion with the use of small microphones embedded in the pleural surface. Two transit times were computed from the first two significant peaks of the cross-correlation of microphone signal pairs. The "fast" wave velocities averaged 301 +/- 92, 445 +/- 80, and 577 +/- 211 (SD) cm/s for Ptp values of 5, 10, and 15 cmH2O, respectively. Corresponding "slow" wave velocities averaged 139 +/- 22, 217 +/- 36, and 255 +/- 89 cm/s. The fast waves were consistent with longitudinal waves of velocity [(K + 4G/3)/p]1/2, where bulk modulus K = 4 Ptp and shear modulus G = 0.7 Ptp. The slow waves were consistent with transverse (and/or Rayleigh) waves of velocity (G/p)1/2, with a G value of 0.9 Ptp. Measured values of K were 5 Ptp and values of G measured by indentation tests were 0.7 Ptp. Thus, stress wave velocities measured on pleural surface of isolated lungs correlated well with elastic moduli of lung parenchyma.

Creatine kinase MM isoforms in serum after cardiac surgery.

We evaluated creatine kinase (CK; EC 2.7.3.2) MM3:MM1 isoform ratios in the serum of cardiac patients immediately after cardiac surgery for the diagnosis of perioperative myocardial injury. The mean ratio was 4.8 (range, 1.4-10.7) in 22 patients who had postoperative myocardial complications and 4.6 (1.3-9.6) in 66 patients who did not. By the first postoperative day the ratio had decreased substantially in both groups of patients. The isoform ratio did not correlate with the concentration of total CK, CK-MB, total lactate dehydrogenase (LD), or the incidence of LD1:LD2 or LD5:LD2 ratio reversal. Of these measurements, CK-MB and LD concentrations differed most between the groups of patients; parallel testing of CK-MB and LD showed an optimized sensitivity and specificity of 77% and 87%, respectively. We conclude that analysis for CK-MM isoforms does not add information in the period immediately after cardiac surgery; concentrations of CK-MB and LD correlate with myocardial injury, but the sensitivity and specificity of these measurements may not be high enough for clinical utility.

Myocardial tissue characterization using pattern recognition procedures on backscattered ultrasonic signals.

Pattern recognition techniques were applied to backscattered signals obtained in vitro from normal and abnormal canine and human heart samples. Orthogonal transforms, in conjunction with the variance criterion, comprised the feature extractors. The minimum-distance (MD) and nearest-neighbor (NN) rules were used as classification. When the MD rule was used, the magnitude of the DFT gave the best performance for both canine and human samples. When the NN rule was used, all the transforms performed comparably. The classification performances were improved for both species when the NN rule was used with feature extractors containing phase information.

Transducers for ultrasonic limb plethysmography.

Shifts in blood content of tissues cause dimension changes which can be measured with an ultrasonic dimension gauge. We have previously described the construction of a limb plethysmograph which uses this principle. Although other workers have described the construction of ultrasonic dimension gauge transducers intended for use inside animals, these transducers are not ideal for plethysmographic application. In this report we describe the design, construction, and performance characteristics of transducers suitable for limb plethysmography. Good signal quality is obtained with minimum care in transducer placement. These transducers incorporate a rugged external housing which provides for simple attachment to the calf by use of double-sided adhesive collars.

A comparison of limb plethysmograph systems proposed for use on the space shuttle.

Comparisons of a Whitney mercury-filled double-stranded strain gauge with two plethysmographs proposed for Shuttle use--an ultrasound and an impedance plethysmograph--were performed on 20 subjects. An occlusive thigh pressure cuff, inflated to 50 mm Hg, caused partial venous occlusion and subsequent blood pooling distal to the cuff. The average maximum volume changes observed in the Whitney/ultrasound test were 2.07% and 3.35%, respectively, and 2.12% and 2.53% for the Whitney/impedance comparison. Applying the ratio of the maximum volume changes to the gain of each test system caused the impedance and ultrasound volume change determinations to be essentially identical to the Whitney gauge. The three different limb parameters measured result in significant magnitude differences but the three systems track their respective changes identically.

A practical ultrasonic plethysmograph.

An ultrasonic plethysmograph, which gives improved performance over the standard Whitney Strain Gauge, is described. This instrument monitors dimension changes in human limbs by measuring the transit times of acoustic pulses across two chords of the limb. In the case of a small uniform expansion, the percentage change in limb volume is shown to be proportional to twice the percentage change in either of the measured chords. Measurement of two chords allows correction for possible non-uniform expansion. In addition, measurement of two chords allows an estimate of the absolute cross-sectional area of the limb. The developed instrument incorporates a microprocessor, which performs necessary calculation and control functions. Use of the microprocessor allows the instrument to be self-calibrating. In addition, the device can be easily reprogrammed to incorporate improvements in operating features or computational schemes.

Ultrasonic characterization of acute renal failure.

The attenuation coefficient, propagation velocity and backscattering coefficient were measured in vitro, on freshly excised and functionally impaired rabbit kidneys. subcutaneous glycerol treatment was used to introduce acute renal failure. Elevated plasma creatinine levels, measured prior to the excision of kidneys, were used as an index of the degree of renal functional impairment. Propagation velocity for the ten kidneys ranged between 1538-1575 m/s with that for the normals being 1540 +/- 4 m/s. Velocity was found to increase with increasing renal damage. The attenuation coefficient for all ten kidneys exhibited a linear frequency dependence over the range 3.5-6.5 MHz. The slope of the attenuation coefficient for the glycerol treated kidneys (0.723 dB/cm/MHz) was found to be higher than the slope for the normals (0.499 dB/cm/MHz). The frequency dependence of the backscattering coefficient was not altered by glycerol treatment leading to the postulate that modification of frequency dependent behavior of the attenuation coefficient in this experimental model is primarily due to absorption.

Microprocessor-base system for ultrasonic tissue characterization.

A microprocessor-based ultrasonic data acquisiton and analysis system has been developed for measuring acoustic parameters of biologic media in a controlled environment. The system consists of commercially available pulse-echo electronics and both minicomputers and microcomputers. Measurements were made on various biologic tissues, both freshly excised and formalin fixed. Propagation velocity measurements were made using the substitution technique. Peak amplitude and log power spectra difference methods were used for attenuation measurements. Measured results indicate that propagation velocity in the soft tissues was in the range of 1530-1580 m/sec at 20 degrees C. Velocity variation among tissues at any given frequency was less than 2%; dispersion in the frequency range 1-10 MHz was approximately 1.5%. Attenuation coefficients showed a marked dependence on frequency increasing in a linear fashion. The slope of the attenuation frequency curve varied with tissue type and could be used for tissue characterization.