Minimum Message Length in Hybrid ARMA and LSTM Model Forecasting.

Modeling and analysis of time series are important in applications including economics, engineering, environmental science and social science. Selecting the best time series model with accurate parameters in forecasting is a challenging objective for scientists and academic researchers. Hybrid models combining neural networks and traditional Autoregressive Moving Average (ARMA) models are being used to improve the accuracy of modeling and forecasting time series. Most of the existing time series models are selected by information-theoretic approaches, such as AIC, BIC, and HQ. This paper revisits a model selection technique based on Minimum Message Length (MML) and investigates its use in hybrid time series analysis. MML is a Bayesian information-theoretic approach and has been used in selecting the best ARMA model. We utilize the long short-term memory (LSTM) approach to construct a hybrid ARMA-LSTM model and show that MML performs better than AIC, BIC, and HQ in selecting the model-both in the traditional ARMA models (without LSTM) and with hybrid ARMA-LSTM models. These results held on simulated data and both real-world datasets that we considered.We also develop a simple MML ARIMA model.

Prognostic Implications of Coronary CT Angiography: 12-Year Follow-Up of 6892 Patients.

OBJECTIVE. The purpose of this study was to add to evidence of the long-term prognostic value of coronary CT angiography (CCTA) when combined with nonimaging risk factors and to investigate how CCTA can contribute to the decision to start statin therapy. MATERIALS AND METHODS. Patients underwent CCTA in an outpatient setting for a variety of indications. The National Death Index from February 2004 through December 2018 was queried to identify the outcomes of death due to any cause (all-cause mortality) and death due to coronary artery disease. Framingham and machine learning risk estimation models were constructed. Machine learning inputs were generated from radiologists' descriptions of the findings on structured report forms and not directly from the image pixels. Kaplan-Meier survival curves and Cox proportional hazards were calculated. Clinical benefit was assessed on the basis of the potential impact on assignment of statin therapy. RESULTS. A total of 6892 outpatients were studied, 4452 (64.6%) of whom were men (mean [± SD] age, 51.2 ± 11.1 years) and 2440 (35.4%) of whom were women (mean age, 57.3 ± 12.2 years). The median follow-up was 11.9 years. Among the 6892 patients, 569 deaths (8.3%) were attributed to all-cause mortality, and 94 deaths (1.4%) were due to coronary artery disease. Survival showed strong dependence on the extent of coronary atherosclerosis. For all-cause mortality, the AUC was 0.85 (95% CI, 0.83-0.86) for the machine learning risk estimation model versus 0.79 (95% CI, 0.78-0.81) for the Framingham risk estimation model (p < 0.001), and for death due to coronary artery disease, the AUC was 0.87 (95% CI, 0.84-0.91) for the machine learning model versus 0.82 (95% CI, 0.77-0.86) for the Framingham model (p = 0.004). Using machine learning risk estimates, the prescription of statins could more accurately be matched to the burden of coronary disease than when Framingham risk estimates were used. CONCLUSION. Compared with the Framingham model, the machine learning model improved risk estimation. Similar models might be useful to better target prescription of statins and reduce their overuse.

Scoring of Coronary Artery Disease Characteristics on Coronary CT Angiograms by Using Machine Learning.

Background Coronary CT angiography contains prognostic information but the best method to extract these data remains unknown. Purpose To use machine learning to develop a model of vessel features to discriminate between patients with and without subsequent death or cardiovascular events. Performance was compared with that of conventional scores. Materials and Methods Coronary CT angiography was analyzed by radiologists into four features for each of 16 coronary segments. Four machine learning model types were explored. Five conventional vessel scores were computed for comparison including the Coronary Artery Disease Reporting and Data System (CAD-RADS) score. The National Death Index was retrospectively queried from January 2004 through December 2015. Outcomes were all-cause mortality, coronary heart disease deaths, and coronary deaths or nonfatal myocardial infarctions. Score performance was assessed by using area under the receiver operating characteristic curve (AUC). Results Between February 2004 and November 2009, 6892 patients (4452 men [mean age ± standard deviation, 51 years ± 11] and 2440 women [mean age, 57 years ± 12]) underwent coronary CT angiography (median follow-up, 9.0 years; interquartile range, 8.2-9.8 years). There were 380 deaths of all causes, 70 patients died of coronary artery disease, and 43 patients reported nonfatal myocardial infarctions. For all-cause mortality, the AUC was 0.77 (95% confidence interval: 0.76, 0.77) for machine learning (k-nearest neighbors) versus 0.72 (95% confidence interval: 0.72, 0.72) for CAD-RADS (P < .001). For coronary artery heart disease deaths, AUC was 0.85 (95% confidence interval: 0.84, 0.85) for machine learning versus 0.79 (95% confidence interval: 0.78, 0.80) for CAD-RADS (P < .001). When deciding whether to start statins, if the choice is made to tolerate treating 45 patients to be sure to include one patient who will later die of coronary disease, the use of the machine learning score ensures that 93% of patients with events will be administered the drug; if CAD-RADS is used, only 69% will be treated. Conclusion Compared with Coronary Artery Disease Reporting and Data System and other scores, machine learning methods better discriminated patients who subsequently experienced an adverse event from those who did not. © RSNA, 2019 Online supplemental material is available for this article. See also the editorial by Schoepf and Tesche in this issue.

Simulating exploration versus exploitation in agent foraging under different environment uncertainties.

For artificial agents trading off exploration (food seeking) versus (short-term) exploitation (or consumption), our experiments suggest that uncertainty (interpreted information, theoretically) magnifies food seeking. In more uncertain environments, with food distributed uniformly randomly, exploration appears to be beneficial. In contrast, in biassed (less uncertain) environments, with food concentrated in only one part, exploitation appears to be more advantageous. Agents also appear to do better in biassed environments.

Combining images and anatomical knowledge to improve automated vein segmentation in MRI.

PURPOSE: To improve the accuracy of automated vein segmentation by combining susceptibility-weighted images (SWI), quantitative susceptibility maps (QSM), and a vein atlas to produce a resultant image called a composite vein image (CV image). METHOD: An atlas was constructed in common space from manually traced MRI images from ten volunteers. The composite vein image was derived for each subject as a weighted sum of three inputs; an SWI image, a QSM image and the vein atlas. The weights for each input and each anatomical location, called template priors, were derived by assessing the accuracy of each input over an independent data set. The accuracy of vein segmentations derived automatically from each of the CV image, SWI, and QSM image sets was assessed by comparison with manual tracings. Three different automated vein segmentation techniques were used, and ten performance metrics evaluated. RESULTS: Vein segmentations using the CV image were comprehensively better than those derived from SWI or QSM images (mean Cohen's d = 1.1). Sixty permutations of performance metric, benchmark image, and automated segmentation technique were evaluated. Vein identification improvements that were both large and significant (Cohen's d > 0.80, p < 0.05) were found in 77% of the permutations, compared to no improvement in 5%. CONCLUSION: The accuracy of automated vein segmentations derived from the composite vein image was overwhelmingly superior to segmentations derived from SWI or QSM alone.

Improved Quantification of Cerebral Vein Oxygenation Using Partial Volume Correction.

Purpose: Quantitative susceptibility mapping (QSM) enables cerebral venous characterization and physiological measurements, such as oxygen extraction fraction (OEF). The exquisite sensitivity of QSM to deoxygenated blood makes it possible to image small veins; however partial volume effects must be addressed for accurate quantification. We present a new method, Iterative Cylindrical Fitting (ICF), to estimate voxel-based partial volume effects for susceptibility maps and use it to improve OEF quantification of small veins with diameters between 1.5 and 4 voxels. Materials and Methods: Simulated QSM maps were generated to assess the performance of the ICF method over a range of vein geometries with varying echo times and noise levels. The ICF method was also applied to in vivo human brain data to assess the feasibility and behavior of OEF measurements compared to the maximum intensity voxel (MIV) method. Results: Improved quantification of OEF measurements was achieved for vessels with contrast to noise greater than 3.0 and vein radii greater than 0.75 voxels. The ICF method produced improved quantitative accuracy of OEF measurement compared to the MIV approach (mean OEF error 7.7 vs. 12.4%). The ICF method provided estimates of vein radius (mean error <27%) and partial volume maps (root mean-squared error <13%). In vivo results demonstrated consistent estimates of OEF along vein segments. Conclusion: OEF quantification in small veins (1.5-4 voxels in diameter) had lower error when using partial volume estimates from the ICF method.

Accuracy of statin assignment using the 2013 AHA/ACC Cholesterol Guideline versus the 2001 NCEP ATP III guideline: correlation with atherosclerotic plaque imaging.

BACKGROUND: Accurate assignment of statin therapy is a major public health issue. OBJECTIVES: The American Heart Association and the American College of Cardiology released a new guideline on the assessment of cardiovascular risk (GACR) to replace the 2001 National Cholesterol Education Program (NCEP) Adult Treatment Panel III recommendations. The aim of this study was to determine which method more accurately assigns statins to patients with features of coronary imaging known to have predictive value for cardiovascular events and whether more patients would be assigned to statins under the new method. METHODS: The burden of coronary atherosclerosis on computed tomography angiography was measured in several ways on the basis of a 16-segment model. Whether to assign a given patient to statin therapy was compared between the NCEP and GACR guidelines. RESULTS: A total of 3,076 subjects were studied (65.3% men, mean age 55.4 ± 10.3 years, mean age of women 58.9 ± 10.3 years). The probability of prescribing statins rose sharply with increasing plaque burden under the GACR compared with the NCEP guideline. Under the NCEP guideline, 59% of patients with ≥50% stenosis of the left main coronary artery and 40% of patients with ≥50% stenosis of other branches would not have been treated. The comparable results for the GACR were 19% and 10%. The use of low-density lipoprotein targets seriously degraded the accuracy of the NCEP guideline for statin assignment. The proportion of patients assigned to statin therapy was 15% higher under the GACR. CONCLUSIONS: The new American Heart Association/American College of Cardiology guideline matches statin assignment to total plaque burden better than the older guidelines, with only a modest increase in the number of patients who were assigned statins.

Coronary artery imaging with single-source rapid kilovolt peak-switching dual-energy CT.

PURPOSE: To evaluate beam-hardening (BH) artifact reduction in coronary computed tomography (CT) angiography with dual-energy CT, to define the optimal monochromatic-energy levels for coronary and myocardial signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) in dual-energy CT, and to compare these levels with single-energy CT. MATERIALS AND METHODS: The study was approved by the institutional review board and/or ethics committee at each site. Patients provided informed consent. Thirty-nine patients were prospectively enrolled to undergo dual-energy CT, and 25 also underwent single-energy CT. Myocardial and coronary SNR, CNR, and iodine concentration were measured across multiple segments at varying monochromatic energy levels (40-140 keV). BH was defined as signal decrease in basal inferior wall versus midinferior wall, and signal increase in midseptum versus midinferior wall. Generalized estimating equation was used to identify optimal monochromatic-energy levels and compare them with single-energy CT. RESULTS: BH was noted at single-energy CT with basal inferior wall mean reduction of 19.7 HU ± 29.2 (standard deviation) and midseptum increase of 46.3 HU ± 36.3. There was reduction in this artifact at 90 keV or greater (1.7 HU ± 18.4 in basal inferior wall and 20.1 HU ± 37.5 in midseptum at 90 keV; P < .05). SNR and CNR were higher in the myocardium and coronary arteries at 60-80 keV than single-energy CT (myocardium: SNR, 3.02 vs 2.39, and CNR, 6.73 vs 5.16; coronary arteries: SNR, 10.83 vs 7.75, and CNR, 13.31 vs 9.54; P < .01). Mean iodine concentration in resting myocardium was 2.19 mg/mL ± 0.57. CONCLUSION: Rapid kilovolt peak-switching dual-energy CT resulted in significant BH reduction and improvements in SNR and CNR in the myocardium and coronary arteries.

Diagnostic accuracy of coronary artery calcium for obstructive disease: results from the ACCURACY trial.

BACKGROUND: Although numerous trials have demonstrated the diagnostic accuracy of coronary artery calcium (CAC) scanning for prediction of obstructive disease, virtually all studies have been performed using Electron Beam CT (EBCT). We evaluated the diagnostic accuracy of CAC by 64-row CT to detect obstructive coronary stenosis compared to quantitative coronary angiography (QCA) in the ACCURACY multicenter trial. METHODS: 16 sites prospectively enrolled 230 patients (pts) [59.5% males, 57yrs] with chest pain referred for invasive coronary angiography (ICA). Pts underwent CAC scan and CT angiography prior to ICA. Total CAC scores were correlated with angiographically documented stenoses using common cutpoints of CAC >0, >100 and >400. Significant obstructive disease was defined as >50% luminal stenosis by QCA. RESULTS: The per-patient accuracy of CAC by 64-row CT compared to QCA demonstrates a high sensitivity and low specificity for the presence of obstructive disease (>50% stenosis on QCA). With CAC >0, >100 and >400, the sensitivities to predict stenosis were 98%, 88%, and 60%, whereas the specificities were 42%, 71%, and 88%, respectively. CONCLUSIONS: Most previous CAC studies have focused on the fact that significant calcium places patients into a higher risk group in terms of future events, and should lead to more aggressive treatment with preventative therapies. This prospective multicenter results comparing 64-row CAC to QCA demonstrate that CAC using 64-row CT scanner, similar to previously published reports using EBCT, is highly sensitive and moderately specific test to predict significant coronary artery stenosis. The presence of abnormal levels of calcium may place patients into a higher risk group in terms of future events, and lead to more aggressive treatment with preventative therapies. However, the detection of calcium does not always help with a clinical diagnosis particularly in the presence of diffuse moderate coronary atheroma. Whether this information is complementary to CTA data remains to be validated.

Universality probability of a prefix-free machine.

We study the notion of universality probability of a universal prefix-free machine, as introduced by C. S. Wallace. We show that it is random relative to the third iterate of the halting problem and determine its Turing degree and its place in the arithmetical hierarchy of complexity. Furthermore, we give a computational characterization of the real numbers that are universality probabilities of universal prefix-free machines.

Extracardiac findings on coronary CT angiograms: Limited versus complete image review.

OBJECTIVE: Cardiac CT often reveals findings outside the heart and great vessels. A few cardiologists have suggested that the field of interpretation be restricted to avoid false-positive diagnoses. Radiologists generally favor a comprehensive review to avoid false-negative findings. The purpose of this study was to examine this tradeoff by comparing broad and focused approaches with viewing coronary CT angiograms. MATERIALS AND METHODS: Outpatient coronary CT angiography was performed on consecutively registered patients. In the broad approach to review, both the large field-of-view and small field-of-view image sets, including lung windows, were evaluated. In the focused approach, attention was centered on the heart, great vessels, and immediately adjacent structures and did not include lung windows. Each finding was classified as necessitating immediate therapy, timely additional workup, longer-term follow-up, or no action. RESULTS: Among 6,920 patients, 1,642 (23.7%) had one or more extracardiac findings for a total of 1,901 findings in the broad viewing scheme. Of the 6,920 patients, 16.2% had a finding necessitating therapy, workup, or follow-up. In the focused viewing scheme, 90.9% of the findings necessitating therapy, 64.1% necessitating workup, and 51.2% necessitating follow-up were missed. Use of the focused approach resulted in fewer false-positive diagnoses, but five malignant tumors of the breast, 88 lung infiltrates, 43 cases of adenopathy, two cases of polycystic kidney disease, one breast abscess, and one case of splenic flexure diverticulitis were missed. CONCLUSION: Almost one fourth of all patients who underwent diagnostic coronary CT angiography in this study had extracardiac findings. Several serious diagnoses were missed with the limited viewing approach, but use of the broad viewing approach led to more workup and follow-up imaging.

The detection of any coronary calcium outperforms Framingham risk score as a first step in screening for coronary atherosclerosis.

OBJECTIVE: The Framingham risk score is often recommended as the starting point for coronary disease screening. We compared the sensitivity of the Framingham risk score for moderate or greater degrees of atherosclerosis to the sensitivity achieved by simple observation of whether any coronary calcium is present. The reference standard was plaque burden as determined by coronary CT angiography. MATERIALS AND METHODS: Of 1,416 men (mean age, 51.4 +/- 9.9 [SD] years) and 707 women (56.9 +/- 10.6 years), most were asymptomatic. Plaque burden (segment plaque score) and stenoses burden (Duke prognostic score) were estimated. A segment plaque score >or= 4 or a Duke prognostic score >or= 3 indicated moderate or greater disease burden. RESULTS: For a segment plaque score >or= 4, the presence of any calcium was 98% sensitive in men and 97% sensitive in women, whereas a Framingham risk score >or= 10% was 74% sensitive in men and 36% sensitive in women. The negative likelihood ratio for the presence of calcium was 0.04 in subjects of either sex, whereas, for a Framingham risk score or= 3, calcium was 97% sensitive in men and 92% sensitive in women, whereas a Framingham risk score >or= 10% was 88% sensitive in men and 35% sensitive in women. The negative likelihood ratio of calcium presence was 0.05 in men and 0.13 in women, whereas the negative likelihood ratio for a Framingham risk score

Left ventricular apical thinning as normal anatomy.

OBJECTIVE: Apical thinning of the left ventricular myocardium has been described by anatomists as a normal feature. Nonetheless, it does not appear in most anatomic atlases. We investigated its presence in healthy patients and patients with left ventricular hypertrophy using coronary computed tomographic arteriography (CCTA). METHODS: Sixty-four patients without a history of cardiac disease and 8 patients with left ventricular hypertrophy were imaged using coronary computed tomographic arteriography. RESULTS: All 64 patients had a focus of myocardial thinning at the left ventricular apex (mean, 1.2 mm [SD, 1.1 mm]). Its average span in the oblique coronal plane was 4.4 mm (2.9 mm), corresponding to a median area of 14.3 mm2 with an interquartile range of 3.9 to 31.6. The focus faced 4.8 degrees (5.9 degrees) toward the diaphragmatic side of the apex. The 8 hypertrophied hearts also had a zone of apical thinning. In a subset of 12 patients in whom functional data were analyzed, this focus did not thicken or move over the cardiac cycle. CONCLUSIONS: A zone of substantial thinning of the left ventricular apex is a normal anatomic feature.

Traditional clinical risk assessment tools do not accurately predict coronary atherosclerotic plaque burden: a CT angiography study.

OBJECTIVE: The objective of our study was to determine the degree to which Framingham risk estimates and the National Cholesterol Education Program (NCEP) Adult Treatment Panel III core risk categories correlate with total coronary atherosclerotic plaque burden (calcified and noncalcified) as estimated on coronary CT angiograms. MATERIALS AND METHODS: Coronary CT angiography was performed in 1,653 patients (1,089 men, 564 women) without a history of coronary heart disease (mean age+/-SD: men, 51.6+/-9.7 years; women, 56.9+/-10.5 years). The most common reasons for the examination were hypercholesterolemia, family history, hypertension, smoking, and atypical chest pain. The coronary tree was divided into 16 segments; four different methods were used to quantify the amount of atherosclerotic plaque or the degree of stenosis in each segment, and segment scores were combined to give total scores. Framingham risk estimates and NCEP risk categories were calculated for each patient. RESULTS: Correlation of plaque scores with the Framingham 10-year risk estimates were modest: Spearman's rho was 0.49-0.55. For all comparisons of NCEP risk categories to plaque score categories, the proportion of raw agreement, p(0), was less than 0.50. Cohen's kappa ranged from 0.18 to 0.20. Overall, 21% of the patients would have their perceived need for statins changed by using the coronary CTA plaque estimates in place of the NCEP core risk categories; 26% of the patients on statins had no detectable plaque. CONCLUSION: Coronary risk stratification using a risk factor only-based scheme is a weak discriminator of the overall atherosclerotic plaque burden in individual patients. Patients with little or no plaque might be subjected to lifelong drug therapy, whereas many others with substantial plaque might be undertreated or not treated at all.

Diagnostic performance of 64-multidetector row coronary computed tomographic angiography for evaluation of coronary artery stenosis in individuals without known coronary artery disease: results from the prospective multicenter ACCURACY (Assessment by Coronary Computed Tomographic Angiography of Individuals Undergoing Invasive Coronary Angiography) trial.

OBJECTIVES: The purpose of this study was to evaluate the diagnostic accuracy of electrocardiographically gated 64-multidetector row coronary computed tomographic angiography (CCTA) in individuals without known coronary artery disease (CAD). BACKGROUND: CCTA is a promising method for detection and exclusion of obstructive coronary artery stenosis. To date, no prospective multicenter trial has evaluated the diagnostic accuracy of 64-multidetector row CCTA in populations with intermediate prevalence of CAD. METHODS: We prospectively evaluated subjects with chest pain at 16 sites who were clinically referred for invasive coronary angiography (ICA). CCTAs were scored by consensus of 3 independent blinded readers. The ICAs were evaluated for coronary stenosis based on quantitative coronary angiography (QCA). No subjects were excluded for baseline coronary artery calcium score or body mass index. RESULTS: A total of 230 subjects underwent both CCTA and ICA (59.1% male; mean age: 57 +/- 10 years). On a patient-based model, the sensitivity, specificity, and positive and negative predictive values to detect > or =50% or > or =70% stenosis were 95%, 83%, 64%, and 99%, respectively, and 94%, 83%, 48%, 99%, respectively. No differences in sensitivity and specificity were noted for nonobese compared with obese subjects or for heart rates < or =65 beats/min compared with >65 beats/min, whereas calcium scores >400 reduced specificity significantly. CONCLUSIONS: In this prospective multicenter trial of chest pain patients without known CAD, 64-multidetector row CCTA possesses high diagnostic accuracy for detection of obstructive coronary stenosis at both thresholds of 50% and 70% stenosis. Importantly, the 99% negative predictive value at the patient and vessel level establishes CCTA as an effective noninvasive alternative to ICA to rule out obstructive coronary artery stenosis. (A Study of Computed Tomography [CT] for Evaluation of Coronary Artery Blockages in Typical or Atypical Chest Pain; NCT00348569).

The case in favor of screening for coronary artery disease with coronary CT angiography.

Coronary computed tomographic angiography is leading a technologic revolution in the field of cardiac imaging. For the first time, it is possible to image the wall of the coronary artery noninvasively to assess plaque burden, characterize plaque, and assess the degree of stenosis. This paper explores the possibility of using coronary computed tomographic angiography as a screening tool. Key screening examination elements, the pros and cons of screening, the limitations of current technology, and potential pitfalls are discussed.

How to create a cardiac CT clinic.

Coronary computed tomography (CT) angiography is taking an exponentially increasing role in the diagnostic algorithm of suspected coronary artery disease. It has the immediate potential of replacing stress tests as the first study a patient receives if suspected of having coronary artery disease. In the near future, it will likely precede all elective, diagnostic cardiac catheterizations secondary to its extraordinary negative predictive value. This paper discusses the 3 building blocks of a successful cardiac CT clinic, image quality, service, and marketing. It then discusses the significant differences in establishing a cardiac CT clinic depending on if the radiologist is hospital based or private office based.

An empirically derived taxonomy of common distress syndromes in the medically ill.

OBJECTIVE: Contemporary psychiatric classifications have not proved to be useful in the understanding and care of people with physical illness. Distress syndromes are common, but classifications fail to differentiate syndromes relevantly. We sought to take a fresh look at the common distress syndromes in the medically ill. METHODS: 312 medical inpatients were interviewed using a structured psychiatric interview [the Monash Interview for Liaison Psychiatry (MILP)] to elicit the presence of mood, anxiety and somatoform symptoms. A previously reported examination of these data using latent trait analysis revealed the dimensions of demoralization, anhedonia, autonomic anxiety, somatoform symptoms and grief. Patients were scored on these dimensions and, on the basis of these, subjected to cluster analysis. Derived classes were compared on a range of demographic and clinical data including psychiatric diagnosis. RESULTS: Six classes were found, distinguished by general levels of distress (measured by demoralization, autonomic anxiety and somatoform symptoms), anhedonia and grief. The most distressed groups were Demoralization and Demoralized Grief. Anhedonic Depression showed moderate levels of distress but the highest level of social dysfunction. Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) categories of mood disorders did not follow any particular pattern other than reflecting severity of distress. The classes of high distress (Demoralization and Demoralized Grief) were significantly associated with younger age, past history of psychiatric treatment, low Global Assessment of Functioning (GAF) scores over the previous 12 months and DSM-IV somatoform disorders. Patients with Demoralized Grief tended to acknowledge their illness as a significant and relevant stressor. Patients with Demoralization identified other stressors as significant. CONCLUSION: Concepts of demoralization, anhedonia and grief differentiate between important clinical syndromes and have informed the development of a taxonomy of common distress syndromes in the medically ill. Research is required to further explore the validity and utility of these concepts.