Deep Learning Using K-Space Based Data Augmentation for Automated Cardiac MR Motion Artefact Detection

Good quality of medical images is a prerequisite for the success of subsequent image analysis pipelines. Quality assessment of medical images is therefore an essential activity and for large population studies such as the UK Biobank (UKBB), manual identification of artefacts such as those caused by unanticipated motion is tedious and time-consuming. Therefore, there is an urgent need for automatic image quality assessment techniques. In this paper, we propose a method to automatically detect the presence of motion-related artefacts in cardiac magnetic resonance (CMR) cine images. We compare two deep learning architectures to classify poor quality CMR images: 1) 3D spatio-temporal Convolutional Neural Networks (3D-CNN), 2) Long-term Recurrent Convolutional Network (LRCN). Though in real clinical setup motion artefacts are common, high-quality imaging of UKBB, which comprises cross-sectional population data of volunteers who do not necessarily have health problems creates a highly imbalanced classification problem. Due to the high number of good quality images compared to the relatively low number of images with motion artefacts, we propose a novel data augmentation scheme based on synthetic artefact creation in k-space. We also investigate a learning approach using a predetermined curriculum based on synthetic artefact severity. We evaluate our pipeline on a subset of the UK Biobank data set consisting of 3510 CMR images. The LRCN architecture outperformed the 3D-CNN architecture and was able to detect 2D+time short axis images with motion artefacts in less than 1ms with high recall. We compare our approach to a range of state-of-the-art quality assessment methods. The novel data augmentation and curriculum learning approaches both improved classification performance achieving overall area under the ROC curve of 0.89.

To evaluate the effect of video information given before cardiovascular magnetic resonance imaging on patient anxiety and to compare patient experiences of cardiovascular magnetic resonance imaging versus myocardial perfusion scintigraphy. To evaluate whether additional information has an impact on motion artefacts. Cardiovascular magnetic resonance imaging and myocardial perfusion scintigraphy are technically advanced methods for the evaluation of heart diseases. Although cardiovascular magnetic resonance imaging is considered to be painless, patients may experience anxiety due to the closed environment. A prospective randomised intervention study, not registered. The sample (n=148) consisted of 97 patients referred for cardiovascular magnetic resonance imaging, randomised to receive either video information in addition to standard text-information (CMR-video/n=49) or standard text-information alone (CMR-standard/n=48). A third group undergoing myocardial perfusion scintigraphy (n=51) was compared with the cardiovascular magnetic resonance imaging-standard group. Anxiety was evaluated before, immediately after the procedure and 1week later. Five questionnaires were used: Cardiac Anxiety Questionnaire, State-Trait Anxiety Inventory, Hospital Anxiety and Depression scale, MRI Fear Survey Schedule and the MRI-Anxiety Questionnaire. Motion artefacts were evaluated by three observers, blinded to the information given. Data were collected between April 2015-April 2016. The study followed the CONSORT guidelines. The CMR-video group scored lower (better) than the cardiovascular magnetic resonance imaging-standard group in the factor Relaxation (p=.039) but not in the factor Anxiety. Anxiety levels were lower during scintigraphic examinations compared to the CMR-standard group (p<.001). No difference was found regarding motion artefacts between CMR-video and CMR-standard. Patient ability to relax during cardiovascular magnetic resonance imaging increased by adding video information prior the exam, which is important in relation to perceived quality in nursing. No effect was seen on motion artefacts. Video information prior to examinations can be an easy and time effective method to help patients cooperate in imaging procedures.

PERIOPERATIVE myocardial complications after noncardiac surgery affect more than 1 million operations each year and are leading causes of morbidity and mortality, especially among patients undergoing vascular surgery. Myocardial complications such as perioperative myocardial infarction (MI) are common and are the most likely cause for perioperative death in all surgical populations.1In general surgery, the risk for perioperative MI is 0.8% in men older than 50 yr2and varies with the cardiovascular status, comorbidities, and the extent of the procedure, reaching more than 20% among patients undergoing vascular surgery.3As patients become older and sicker and procedures become more aggressive and extensive, physicians must find novel approaches to evaluate and prepare cardiac patients for noncardiac procedures and reduce perioperative myocardial events.The American Heart Association–American College of Cardiology guidelines for cardiac risk evaluation use the patient's history, physical examination, and functional capacity, and taking into account the expected surgery, one may recommend further assessment with noninvasive testing or coronary angiography.4The purpose of this review is to provide an overview of available imaging tools that could potentially be used for perioperative evaluation of cardiac patients before noncardiac surgery, in accord with recent guidelines and with focus on the recent progress made with cardiac computed tomography (CT).Traditionally, preoperative evaluation has relied on the patient's history, physical examination, and functional capacity. After preoperative evaluation, the physician should determine the clinical predictors as major, intermediate, or minor4and evaluate the functional capacity of the patient. New York Heart Association class II heart failure, which equals 4 metabolic equivalents,4has also been found to be an important predictor of perioperative cardiac complications after major noncardiac surgery. The cutoff value of 4 or more metabolic equivalents determines an adequate cardiac functional capacity and reserve and predicts perioperative cardiac events in patients treated with high-risk noncardiac surgery.5Finally, the extent and risk associated with the procedure should be categorized as major, intermediate, or minor. The cardiac risk is the combined incidence of cardiac death and nonfatal MI and is greater than 5% for high-risk procedures, whereas intermediate-risk procedures have less than 5% cardiac risk (1–5%), and low-risk procedures have less than 1% cardiac risk.4Indeed, based on these criteria, the flowchart of Eagle et al. 4suggests when a noninvasive test or coronary angiography may be appropriate to evaluate cardiac function, reserve, or ability to withstand surgical stress. Pertinent to the utility of evaluating cardiac structure and function, new imaging techniques for evaluation of the heart have emerged in the past couple of decades. These techniques may provide more comprehensive information regarding the structure and function of the heart and coronary arteries, and evaluate adequately patients who have mechanical restrictions to perform exercise-induced stress. Such developments in cardiac imaging may eventually provide an opportunity to revise and refine the steps involving the evaluation of cardiac patients.6,7This approach should be evaluated cautiously and should take into consideration recent recommendations and guidelines. For example, the current American Heart Association–American College of Cardiology guidelines recommend the use of perioperative β-blocker therapy as an alternative approach to decrease cardiovascular risk.8β-Blockers initiated at least 1 week before major vascular surgery and continued for 30 days postoperatively reduced significantly the perioperative incidence of nonfatal MI and death from cardiac causes in high-risk patients.9In addition to beta blockers, statins may reduce perioperative mortality in patients undergoing major vascular surgery and may have an additive effect to β-blockers.10,11The strategy of preoperative coronary artery revascularization before elective major vascular surgery to reduce perioperative cardiac morbidity and mortality was investigated in the Coronary Artery Revascularization Prophylaxis trial, a multicenter study involving patients with vascular disease and significant coronary artery disease (CAD), but no unstable angina. Cardiac revascularization did not result in increased survival, either perioperatively or in long-term follow-up, in patients who needed elective vascular surgery.12However, although the Coronary Artery Revascularization Prophylaxis study is a cornerstone trial in the perioperative care of cardiac patients for noncardiac surgery, it had some limitations. In particular, only 9% of the patients scheduled to undergo vascular operations were eligible for the study. The main reasons for exclusion were insufficient cardiac risk, an urgent vascular surgery, previous revascularization without ischemia, severe coexisting illnesses, left main coronary artery stenosis of at least 50%, left ventricular ejection fraction less than 20%, and severe aortic stenosis. The outcome of these patients has not been adequately evaluated.Indeed, taking into consideration the patient's perioperative cardiac risk according to clinical predictors, functional capacity, and the extent of the future surgery, the anesthesiologist must decide whether further cardiac assessment or perioperative medical management are indicated. Patients whose functional capacity is difficult to establish, who underwent previous coronary revascularization, who have unstable or changed cardiac status, or who have severe comorbidities may need further evaluation. Several imaging techniques are available for evaluation of cardiac patients, including cardiac CT, coronary angiography, stress echocardiography, cardiac magnetic resonance imaging (CMRI), and myocardial nuclear studies.Computed tomography is widely available in large medical centers and is routinely used in clinical practice. The basic principle of CT is that a fan-shaped, thin x-ray beam passes through the body at many angles to allow for cross-sectional images. After collimation of the beam (i.e. , achieving a definitive slice thickness using a collimator) to reduce scatter, the photons are recorded on a corresponding detector array, and the transmission data are digitized. A “filtered back projection” reconstruction algorithm, which takes into account the attenuation of the x-ray beam along its path, allows for reconstruction of the grayscale values of each picture element (pixel) with reference to the value for water and air, to depict cross-sectional images. Reconstruction algorithms and multirow detectors applied in current scanners enable three-dimensional volumetric imaging and multiple high-quality reconstructions of various volumes of interest.13Current clinical scanners used for cardiovascular imaging employ either a rotating x-ray source with a circular, stationary detector array (e.g. , helical CT) or electromagnetic deflection of an electron beam to replace mechanical motion (electron beam computed tomography [EBCT]).14Multidetector computed tomography (MDCT) is a helical CT with a large array of detectors, which allow it to acquire a large number of slices simultaneously (4–256) and greatly increase its resolution. However, to obtain quantitative measurements of tissue opacity within a specific cardiac phase (e.g. , for measurement of perfusion), the scanning time should generally be less than 100 ms.15Sufficiently high temporal resolution (the time required to acquire the data for one image) is currently offered only by the EBCT (50 ms/image) and the novel Dual-Source CT (SOMATOM® Defi-nition; Siemens Medical Systems, Forchheim, Germany) (83 ms/image), a recently released model of MDCT that has two x-ray tubes (rather than one) positioned at 90° to each other, thereby doubling temporal resolution.16However, this technique remains to be validated for such measurements.Electron beam computed tomography was developed in the 1980s and hailed as an ultimate cardiac scanner. It allows almost simultaneous data acquisition from up to eight parallel slices (7–8 mm thick) by rapid sweeping of an electron beam along target rings in as little as 50 ms per scan. Because it does not involve moving parts (and therefore decreases the need for cooling), the speed of acquisition (temporal resolution) with EBCT is faster than with MDCT (table 1) and usually does not require slowing the heart rate pharmacologically. The high speed of EBCT is offset by moderate image quality and relatively restricted power for acquisition of a large number of images, which decreased its popularity for comprehensive cardiac studies. Consequently, its availability is limited, resulting in declining use of this technology.Multidetector computed tomography is a relatively ubiquitous and newer scanner that has temporal resolution of 330–400 ms/image, which enables many studies of cardiac anatomy and global function (e.g. , ejection fraction and cardiac output). Spatial resolution can be achieved with 64-slice scanners using isotropic voxels (consistent three-dimensional image quality in any reconstruction plane) of 0.4 × 0.4 × 0.4 mm. The technique involves continuous rotation of the x-ray tube and detectors and simultaneous translation of the patient through the gantry opening, and can acquire multiple simultaneous sections of variable widths using prospective electrocardiographic triggering.17Alternatively, retrospective electrocardiographic gating enables reconstruction of images at any desired time in the cardiac cycle. Data can be used from multiple slices to reconstruct other imaging planes. The images are of best quality when the resting heart rate (HR) is less than 70 beats/min. At faster heart rates, motion artifacts may become more prominent, and therefore HR may need to be pharmacologically decreased before scanning. The temporal resolution of MDCT determines the overall scan time. As gantry rotation speeds increase, the minimum slice thickness decreases, with submillimeter sections throughout the heart acquired during a single breath hold. The overall CT scan time is approximately 12 s, and the mean total time for the examination is less than 13 min with 64-slice technology.The improvement in MDCT technology enabled the assessment of significant luminal stenosis and identification of nonstenotic atherosclerotic plaques. Virtual noninvasive angiography, with three-dimensional reconstruction of coronary anatomy from cardiac CT images, can provide information about coronary luminal obstruction, calcium scoring, and composition of the plaque.18Furthermore, it has the added benefit of offering fine details of the examined vessels (fig. 1).Excellent sensitivity and specificity were found for evaluation of proximal, middle and distal left anterior descending, first diagonal, proximal and distal circumflex, obtuse marginal and proximal mid and distal right coronary artery.19A meta-analysis of diagnostic performance of MDCT compared with invasive coronary angiography showed a sensitivity of 85% and a specificity of 95% for identification of CAD. On average, 87% of segments had diagnostic image quality, with a significant increase from 78% with 4-slice systems to 96% with the more recent 16-slice systems.6Multidetector computed tomography could provide a clinically useful tool in the workup of symptomatic patients before angiography.20A patient with chronic chest pain indicative of CAD could undergo CT angiography, and if low calcium score and no circumferential calcifications (or other test results indicative of ischemia) are found, invasive angiography may not be indicated, because MDCT has excellent negative predictive value to rule out CAD. By this approach, patients with primarily microvessel disease (which may cause angina and abnormal stress test results) may be identified and not required to go through unnecessary fluoroscopic angiography, and aggressive medical therapy would be indicated before surgery.20MDCT also has high diagnostic accuracy in detecting bypass graft stenosis and occlusions in symptomatic patients after coronary artery bypass grafting, which might potentially reduce dramatically the number of unnecessary invasive angiographies performed in these patients.21However, although the MDCT technique has been extensively evaluated for its accuracy for detection of CAD compared with standard coronary angiography,6,19,20,22it has not been applied for the routine evaluation of cardiac patients for noncardiac surgery and, to date, is not a recommended technique for perioperative risk stratification.The main technical limitations of cardiac CT for evaluation of the coronary arteries include the difficulty in handling cardiac motion, arrhythmia, severe calcifications, vessel size less than 1.5 mm, breathing, the presence of stents, and poor enhancement.22Lesions with extensive calcified components and implanted coronary stents compromise the accuracy of MDCT coronary angiography by causing artifacts.23Stent type and diameter influence evaluability of in-stent restenosis by MDCT, but in evaluable stents, sensitivity is still 86% and specificity is 98%.24When HR is reduced below 70 beats/min, image quality is improved, especially in terms of the visualization of the right coronary and left circumflex arteries, which are both significantly prone to motion artifacts (particularly at higher HR) because of their close proximity to the atrium, which is reactivated during the early diastolic phase.25For evaluation of coronary artery calcium volume with MDCT, thin-slice retrospective spiral electrocardiographic–gated scanning is desirable.26Most studies on coronary calcification have been performed using EBCT, which is still considered the “gold standard” (table 1). To image the coronary arteries with EBCT, 30–40 axial images are obtained with 3-mm slice thickness, using single-slice prospective electrocardiographic triggering in the craniocaudal direction along the full length of the heart. Rapid image acquisition at 100 ms allows accurate measurement of calcium deposits in the coronary arteries.27Quantification of coronary artery calcifications was found to be independently associated with cardiac events in a 3-yr follow-up of thousands of initially asymptomatic patients.28Noninvasive characterization and quantification of atherosclerotic plaque burden may also have important implications for the prevention of CAD progression and its complications.18For the past several years, CMRI has been considered the reference standard for assessment of left ventricular (LV) functions. However, cardiac CT is playing an increasingly important role in this evaluation. Both CT and CMRI surpass two-dimensional imaging techniques, such as standard two-dimensional echocardiography, for cardiac quantification because of their ability to generate contiguous short axis cine images, allowing for three-dimensional measurements without the use of geometric assumptions.29Postprocessing tools allow fast and semiautomatic determination of LV function parameters from MDCT data in analogy to known CMRI evaluation approaches.30Studies have demonstrated excellent correlation between cardiac CT and CMRI for LV ejection fraction, end-diastolic volume, end-systolic volume, stroke volume, and myocardial mass.30Furthermore, global and regional LV functions agree well with echocardiography, with correlation coefficients ranging from 0.91 to 0.97 for MDCT and from 0.93 to 0.98 for CMRI.7,30,31Although MDCT is not considered to be the first-line modality for assessment of LV function, it can provide a combined assessment of cardiac morphology and function without the need for additional radiation exposure in patients undergoing MDCT coronary angiography. The acquisition of images is performed according to the R–R interval. For MDCT, the image data are gated with the electrocardiogram to allow reconstruction at various times throughout the cardiac cycle. CT allows quantitative analysis of regional and global systolic function in normal and pathologic conditions by using short axis slices from the base to the apex of the heart. End-diastole and end-systole are defined as maximal and minimal LV volume, and LV ejection fraction is the difference between them.30Diastolic function can also be assessed from the rate of change of the LV volume during diastole.13Also, because temporal resolution for electrocardiographic-gated cardiac CT scans (down to 165 ms) is poorer in comparison to cine CMRI (30–40 ms), CT imaging may miss peak ejection rate or peak filling rate. However, advances in MDCT imaging that will improve temporal resolution may correct this problem.16Regional wall motion abnormalities, myocardial thinning, ventricular aneurysm, and mural thrombi in the infarcted area can all be detected by cardiac CT. Dual-phase contrast CT can detect acute MI characterized by an initial filling defect and late enhancement at the site of the damaged myocardium. Late enhancement may have the potential to distinguish viable from nonviable myocardium, and has significant prognostic value for the recovery of the myocardial wall motion and thickness after ischemia and in response to therapeutic revascularization.32In patients with previous MI, MDCT permits accurate, noninvasive assessment of coronary artery stenosis, LV function, and perfusion, assessed from a single data set.33If the myocardium supplied by the stenotic or occluded vessel is still viable, medical treatment or revascularization can be considered to reduce the risk of perioperative ischemia in the affected myocardium.Experimental studies in animals and humans demonstrated that cardiac CT could be used to assess microvascular function, although this technique is not used clinically, partially because of high radiation exposure. Evaluation of myocardial perfusion at rest and after infusion of vasodilators imposing cardiac challenge can reveal the presence of otherwise undetectable limited myocardial flow reserve, which might have a significant value for detection of borderline or very early alterations in cardiac microvascular function, as well as detecting increases in microvascular permeability that may reflect endothelial dysfunction or ischemic changes.34The use of CT scanning is limited by the need for contrast media and radiation exposure (table 2). The risks associated with contrast media administration include extravasation at the contrast injection site, allergic contrast reaction, and a decline in function, as well as and vascular In patients with administration of contrast media is media has been reduced and is than that used during angiography potential of alternative contrast such as might further decrease the various radiation are with MDCT, and scanning may allow a decrease in radiation of cardiac function and volumes may from to potentially any based on gated that data several cardiac limitations include the need to image quality in patients and the need for and physicians for acquisition and of cardiac CT data of the limitations of MDCT is the need to decrease heart rate at time of which may be in some It further the need for CT technology to acquisition time a greater volume of developed scanners such as the Dual-Source MDCT may this to temporal the number of that can be from axial CT images is limited only by gantry rotation speed and patient heart decreased ejection fraction in is associated with decreased overall postoperatively and with increased incidence of heart failure, but no has been found with main associated with cardiac events is the presence of wall motion has a low predictive is appropriate in patients who American Heart Association–American College of Cardiology clinical guidelines and who would require if no surgery were as well as in with aortic sensitivity of can be by the heart either or pharmacologically. stress test with administration of can provide information about cardiac function, ventricular and function, and a is in to 85% of maximal HR or or as wall motion abnormalities, whereas a in ejection fraction in response to administration is a of more severe CAD. has excellent negative predictive values and sensitivity but moderate specificity and is significantly and long-term cardiac risk using initial cardiac risk assessment and noninvasive testing with may the perioperative and long-term treatment of patients undergoing major vascular comparison with other noninvasive a diagnostic new of may improve cardiac risk compared with standard in patients undergoing noncardiac surgery whose functional capacity be evaluated by stress stress has a relatively low and can assess cardiac structure and function, wall thickness, aortic and the presence of However, has sensitivity for detection of single vessel disease and is to distinguish between microvascular and CAD (table because wall motion can be found in progress has been made during the past in which is used to evaluate a of heart and including CAD. CMRI is based on the of which and to an magnetic CMRI a causes deflection of the from the direction of the main magnetic The is during the of the back their in the magnetic are and data are used for image many image are for assessment of global ventricular and function, ejection fraction, stroke volume, and detection of and acute and chronic LV can be and using analysis of may find use in the clinical magnetic resonance imaging has higher temporal resolution but resolution (the ability to distinguish between two on the image) than cardiac resolution of 1.5 × 1.5 may have an CMRI for detection of because its overall accuracy in the detection of coronary artery stenosis is higher (table 2). The axial image of MDCT allows more vascular as compared with CMRI that may artifacts to sensitivity and specificity of CMRI for detection of coronary artery stenosis are and whereas of coronary segments are is noninvasive and high-quality images of the heart (table 2). However, CMRI examination be applied in patients who have a or and the presence of and stents can CMRI images and to is the of the test than an during which the patient must be CMRI is an important noninvasive tool and does not the use of potential contrast are needed to evaluate its use as a perioperative tool and its correlation with perioperative cardiac use of myocardial computed tomography and tomography has in the past two because of their in useful information about myocardial perfusion and function based on administration of an unstable with has been used for many for preoperative cardiac evaluation. the coronary response associated with testing not the HR and has been used in patients scheduled to undergo vascular surgery. However, has low and negative predictive and was not significantly associated with the incidence of perioperative MI, ischemia, or other the other the of this test was when used in patients undergoing aortic surgery whose risk could not have been on the of clinical value of imaging for preoperative cardiac assessment in high-risk has been demonstrated in patients before and vascular of myocardial with is important in patients with LV function to and the potential of revascularization can be These studies showed excellent correlation of with cardiac both perioperatively within 30 days and at long-term has high sensitivity in coronary disease but limited sensitivity for detecting significant disease and CAD (table 2). It also little functional data other than myocardial perfusion and has relatively low resolution that the quality of scanners have a maximal of approximately is required for before the performance of myocardial perfusion because the coronary effect of or is by with can be used to functional recovery of viable myocardium after coronary artery a is from a a in and with an results in of both the and the with to in the of electromagnetic radiation of two which in from each and are detected by a of radiation detectors at The difference in time the photons each radiation detector is used to detect the source of the of radiation detectors are in the gantry with million of per the imaging is useful in noninvasive quantification of myocardial flow and coronary flow reserve, and is also of detecting early disease in high-risk asymptomatic and the progression or of is to especially in patients and in undergoing remains a significant its high and limited availability currently its use as a technology of coronary angiography is well and involves injection of contrast media into the coronary artery by imaging in multiple planes. is by comparison of diameter at a stenotic site with that of a normal reference can evaluate coronary stenosis, LV function, and (table coronary angiography is an invasive technique and has a major rate of including mortality in of it may involve for the patient and is associated with other such as vascular complications for diagnostic and for complications and complications the risk of coronary angiography with that of MDCT, some risks are common to both procedures, such as an allergic contrast reaction, and exposure to whereas are to contrast media injection causes severe allergic in of risks of a of for MDCT angiography and for coronary angiography. However, the and risks and a overall risk of mortality from coronary that for MDCT angiography use of a power in a vascular with cardiac CT an additional risk of which in of the relatively significant number of negative invasive angiographies performed each the risks to this procedure by using noninvasive can greatly to the morbidity and mortality of coronary angiography. the number of negative angiographies might to significant because the of cardiac may be as as times that of cardiac evaluation of patients with or known cardiac disease for noncardiac surgery has been in the of of and The preoperative evaluation on assessment of clinical predictors, functional capacity, and the extent of the surgical The first-line to reduce perioperative morbidity and mortality include medical therapy with with HR and However, the strategy may also include of noninvasive imaging techniques that provide comprehensive and assessment of cardiac structure and In particular, in patients with restricted physical in assessment of the functional capacity is obtained using imaging with evaluation of the heart may potentially to treatment Cardiac CT is can provide coronary angiography for plaque assessment and scoring, and can assess LV function, and myocardial in a fast and relatively CT can in out significant CAD and determination of the need for invasive angiography. cardiac CT might potentially be the tool for overall evaluation of cardiac patients for noncardiac surgery. at this is no that preoperative cardiac CT scanning is to reduce perioperative studies are needed to assess the role of cardiac CT in the evaluation of patients at risk for cardiovascular disease before noncardiac surgery.

Medical image quality assessment (MIQA) is a vital prerequisite in various medical image analysis applications. Most existing MIQA algorithms are fully supervised that request a large amount of annotated data. However, annotating medical images is time-consuming and labor-intensive. In this paper, we propose an unsupervised anomaly-aware framework with test-time clustering for optical coherence tomography angiography (OCTA) image quality assessment in a setting wherein only a set of high-quality samples are accessible in the training phase. Specifically, a feature-embedding-based low-quality representation module is proposed to quantify the quality of OCTA images and then to discriminate between outstanding quality and non-outstanding quality. Within the non-outstanding quality class, to further distinguish gradable images from ungradable ones, we perform dimension reduction and clustering of multi-scale image features extracted by the trained OCTA quality representation network. Extensive experiments are conducted on one publicly accessible dataset sOCTA-3×3-10k, with superiority of our proposed framework being successfully established.

An optimized CNN-based quality assessment model for screen content image

The presence of motion artifacts in the photoplethysmographic (PPG) signals is one of the major obstacles in the extraction of reliable cardiovascular parameters in real time and continuous monitoring applications. In the current paper we present a comparison between two motion artifacts detection methodologies proposed by Couceiro et al. [1] and Correia T. [2]. The first method is based on the analysis of the variations in the time and period domain characteristics of the PPG signal. The second method analysis the differences between the changes in the Heart Rate measured from the PPG and the ECG signals. Both methodologies are validated in healthy and cardiovascular diseased volunteers, for 11 different motion artifact patterns. The results achieved by the presented methodologies show a better performance of the first algorithm (SE: 83% and SP: 87%) and the great importance of both time and period domain features in the discrimination of motion artifacts from clean PPG pulses.

The purpose of our study was to evaluate the effects of 0.3-second high-resolution CT (HRCT) of the lung using partial reconstruction on cardiac motion artifacts and image noise. Thirty-seven pairs of 0.3-second (partial reconstruction) and 0.75-second (full reconstruction) HRCT images were obtained for the lower lung zone during full-inspiration breath-holding. Imaging parameters other than temporal resolution were identical for each patient. Two radiologists visually graded motion artifacts of the cardiac border, bronchi, pulmonary vessels, and fissure in the left lung on a 4-point scale (with 4 indicating no artifacts). The maximum width of motion along the left cardiac border and the area percentage of motion artifacts in the left lung were calculated. Image noise in the air and lung was also determined. Cardiac motion artifacts and image noises were compared between the two sets of CT images. Visual grades for the cardiac border (4 +/- 0), bronchi (3.8 +/- 0.7), pulmonary vessels (3.6 +/- 0.8), and fissure (3.9 +/- 0.5) were higher for 0.3-second images than for 0.75-second images (1.7 +/- 0.7, 2.0 +/- 1.0, 1.6 +/- 0.7, and 2.4 +/- 0.9, respectively) (p < 0.001). The maximum width of motion along the left cardiac border (0.1 +/- 0.5 mm) and the area percentage of motion artifacts in the left lung (6.7% +/- 18.4%) were smaller for 0.3-second images than for 0.75-second images (4.5 +/- 1.7 mm and 36.2% +/- 20.9%, respectively) (p < 0.001). Image noises in the air (38.0 +/- 9.2) and the lung (86.0 +/- 23.1) were greater for 0.3-second images than for 0.75-second images (35.6 +/- 9.6 and 76.0 +/- 20.3, respectively) (p < 0.01). Compared with 0.75-second HRCT using full reconstruction, 0.3-second HRCT using partial reconstruction substantially reduces cardiac motion artifacts in the lung at the expense of increasing image noise.

Highlights From the Circulation Family of Journals.

Area at risk (AAR) is an important parameter for the assessment of the salvage area after revascularization in acute myocardial infarction (AMI). By combining AAR assessment by T2-weighted imaging and scar quantification by late gadolinium enhancement imaging cardiovascular magnetic resonance (CMR) offers a promising alternative to the “classical” modality of Tc99m-sestamibi single photon emission tomography (SPECT). Current T2 weighted sequences for edema imaging in CMR are limited by low contrast to noise ratios and motion artifacts. During the last years novel CMR imaging techniques for quantification of acute myocardial injury, particularly the T1-mapping and T2-mapping, have attracted rising attention. But no direct comparison between the different sequences in the setting of AMI or a validation against SPECT has been reported so far. We analyzed 14 patients undergoing primary coronary revascularization in AMI in whom both a pre-intervention Tc99m-sestamibi-SPECT and CMR imaging at a median of 3.4 (interquartile range 3.3–3.6) days after the acute event were performed. Size of AAR was measured by three different non-contrast CMR techniques on corresponding short axis slices: T2-weighted, fat-suppressed turbospin echo sequence (TSE), T2-mapping from T2-prepared balanced steady state free precession sequences (T2-MAP) and T1-mapping from modified look locker inversion recovery (MOLLI) sequences. For each CMR sequence, the AAR was quantified by appropriate methods (absolute values for mapping sequences, comparison with remote myocardium for other sequences) and correlated with Tc99m-sestamibi-SPECT. All measurements were performed on a 1.5 Tesla scanner. The size of the AAR assessed by CMR was 28.7 ± 20.9 % of left ventricular myocardial volume (%LV) for TSE, 45.8 ± 16.6 %LV for T2-MAP, and 40.1 ± 14.4 %LV for MOLLI. AAR assessed by SPECT measured 41.6 ± 20.7 %LV. Correlation analysis revealed best correlation with SPECT for T2-MAP at a T2-threshold of 60 ms (ms) (slope = 0.99, Pearson’s r = 0.94), and for MOLLI at T1-threshold of 1,075 ms (slope 0.86, r = 0.91, Pearson’s r = 0.45). For the assessment of AAR in AMI, the novel T2-mapping technique correlates best with SPECT size, T1-mapping with MOLLI and standard T2-weighted imaging showed similar good correlations.

Optimization of activity-driven event detection for long-term ambulatory urodynamics.

With the latest popularization of remote desktop software and online work, screen content images (SCIs) are increasingly being contacted by people. Many image quality assessment (IQA) methods based on deep learning have been proposed for SCIs. However, most IQA methods for SCIs pay less attention to regional characteristics. An asymmetric two-stream network based on region features is proposed for SCI objective quality assessment, in which features are derived from a channel separation feature extraction (CSFE) sub-network and a spatial analysis feature extraction (SAFE) sub-network. In the CSFE sub-network, three channel separation (CS) modules are utilized to strengthen the ability for capturing deepthwise features in the pictorial regions. In the SAFE sub-network, two spatial analysis (SA) module are applied to focus on the structural information in the spatial domain for the textual regions. Two asymmetric sub-networks can efficiently extract the corresponding region features by separately considering the pictorial and textual regions of SCIs. In addition, an adaptive global weighting strategy is introduced to estimate the quality for SCIs. The experimental results on the screen image quality assessment database show that the proposed model is superior to the other state-of-the-art SCI quality assessment methods.

During the process of image acquisition, images are often subject to out-of-focus or defocus blur because of the improper adjustment of the camera's focal length, this image blur will degrade the image quality. However, in the literature, image quality assessment (IQA) methods dedicated to evaluating the quality of images with out-of-focus blur remain few. Therefore, in this paper, we focus our attention on the quality assessment of images that suffer from out-of-focus blur and propose an objective quality assessment method accordingly. Concretely, we construct a dedicated out-of-focus blurred image dataset, which is composed of 150 images subjected to different degrees of out-of-focus blur and the mean opinion scores (MOSs). Then, we propose a specific objective quality metric for the blurred images, which combines image sharpness assessment and saliency-guided pooling strategy. Experimental results demonstrate the proposed metric highly correlates with human judgements of image quality.

Multi-view-plus-depth (MVD) representation has gained significant attention recently as a means to encode 3D scenes, allowing for intermediate views to be synthesized on-the-fly at the display site through depth-image-based-rendering (DIBR). Automatic quality assessment of MVD images/videos is critical for the optimal design of MVD image/video coding and transmission schemes. Most existing image quality assessment (IQA) and video quality assessment (VQA) methods are applicable only after the DIBR process. Such post-DIBR measures are valuable in assessing the overall system performance, but are difficult to be directly employed in the encoder optimization process in MVD image/video coding. Here we make one of the first attempts to develop a perceptual pre-DIBR IQA approach for MVD images by employing an information content weighted approach that balances between local quality measures of texture and depth images. Experiment results show that the proposed approach achieves competitive performance when compared with state-of-the-art IQA algorithms applied post-DIBR.

An investigation of diseases using magnetic resonance (MR) imaging requires automatic image quality assessment methods able to exclude low-quality scans. Such methods can be also employed for an optimization of parameters of imaging systems or evaluation of image processing algorithms. Therefore, in this paper, a novel blind image quality assessment (BIQA) method for the evaluation of MR images is introduced. It is observed that the result of filtering using non-maximum suppression (NMS) strongly depends on the perceptual quality of an input image. Hence, in the method, the image is first processed by the NMS with various levels of acceptable local intensity difference. Then, the quality is efficiently expressed by the entropy of a sequence of extrema numbers obtained with the thresholded NMS. The proposed BIQA approach is compared with ten state-of-the-art techniques on a dataset containing MR images and subjective scores provided by 31 experienced radiologists. The Pearson, Spearman, Kendall correlation coefficients and root mean square error for the method assessing images in the dataset were 0.6741, 0.3540, 0.2428, and 0.5375, respectively. The extensive experimental evaluation of the BIQA methods reveals that the introduced measure outperforms related techniques by a large margin as it correlates better with human scores.

Cardiac magnetic resonance (CMR) images play a growing role in the diagnostic imaging of cardiovascular diseases. Full coverage of the left ventricle (LV), from base to apex, is a basic criterion for CMR image quality and necessary for accurate measurement of cardiac volume and functional assessment. Incomplete coverage of the LV is identified through visual inspection, which is time-consuming and usually done retrospectively in the assessment of large imaging cohorts. This paper proposes a novel automatic method for determining LV coverage from CMR images by using Fisher-discriminative three-dimensional (FD3D) convolutional neural networks (CNNs). In contrast to our previous method employing 2D CNNs, this approach utilizes spatial contextual information in CMR volumes, extracts more representative high-level features and enhances the discriminative capacity of the baseline 2D CNN learning framework, thus achieving superior detection accuracy. A two-stage framework is proposed to identify missing basal and apical slices in measurements of CMR volume. First, the FD3D CNN extracts high-level features from the CMR stacks. These image representations are then used to detect the missing basal and apical slices. Compared to the traditional 3D CNN strategy, the proposed FD3D CNN minimizes within-class scatter and maximizes between-class scatter. We performed extensive experiments to validate the proposed method on more than 5,000 independent volumetric CMR scans from the UK Biobank study, achieving low error rates for missing basal/apical slice detection (4.9%/4.6%). The proposed method can also be adopted for assessing LV coverage for other types of CMR image data.