ABSTRACT
BACKGROUND: Fluoroscopy guided interventions (FGIs) pose a risk of prolonged radiation exposure; personalized patient dosimetry is necessary to improve patient safety during these procedures. However, current FGIs systems do not capture the precise exposure regions of the patient, making it challenging to perform patient-procedure-specific dosimetry. Thus, there is a pressing need to develop approaches to extract and use this information to enable personalized radiation dosimetry for interventional procedures. PURPOSE: To propose a deep learning (DL) approach for the automatic localization of 3D anatomical landmarks on randomly collimated and magnified 2D head fluoroscopy images. MATERIALS AND METHODS: The model was developed with datasets comprising 800 000 pseudo 2D synthetic images (mixture of vessel-enhanced and non-enhancement), each with 55 annotated anatomical landmarks (two are landmarks for eye lenses), generated from 135 retrospectively collected head computed tomography (CT) volumetric data. Before training, dynamic random cropping was performed to mimic the varied field-size collimation in FGI procedures. Gaussian-distributed additive noise was applied to each individual image to enhance the robustness of the DL model in handling image degradation that may occur during clinical image acquisition in a clinical environment. The model was trained with 629 370 synthetic images for approximately 275 000 iterations and evaluated against a synthetic image test set and a clinical fluoroscopy test set. RESULTS: The model shows good performance in estimating in- and out-of-image landmark positions and shows feasibility to instantiate the skull shape. The model successfully detected 96.4% and 92.5% 2D and 3D landmarks, respectively, within a 10 mm error on synthetic test images. It demonstrated an average of 3.6 ± 2.3 mm mean radial error and successfully detected 96.8% 2D landmarks within 10 mm error on clinical fluoroscopy images. CONCLUSION: Our deep-learning model successfully localizes anatomical landmarks and estimates the gross shape of skull structures from collimated 2D projection views. This method may help identify the exposure region required for patient-specific organ dosimetry in FGIs procedures.
ABSTRACT
BACKGROUND/INTRODUCTION: Traumatic brain injury (TBI) remains a leading cause of disability and mortality, with skull fractures being a frequent and serious consequence. Accurate and rapid diagnosis of these fractures is crucial, yet current manual methods via cranial CT scans are time-consuming and prone to error. METHODS: This review paper focuses on the evolution of computer-aided diagnosis (CAD) systems for detecting skull fractures in TBI patients. It critically assesses advancements from feature-based algorithms to modern machine learning and deep learning techniques. We examine current approaches to data acquisition, the use of public datasets, algorithmic strategies, and performance metrics RESULTS: The review highlights the potential of CAD systems to provide quick and reliable diagnostics, particularly outside regular clinical hours and in under-resourced settings. Our discussion encapsulates the challenges inherent in automated skull fracture assessment and suggests directions for future research to enhance diagnostic accuracy and patient care. CONCLUSION: With CAD systems, we stand on the cusp of significantly improving TBI management, underscoring the need for continued innovation in this field.
Subject(s)
Skull Fractures , Tomography, X-Ray Computed , Humans , Skull Fractures/diagnostic imaging , Tomography, X-Ray Computed/methods , Diagnosis, Computer-Assisted/methods , Image Processing, Computer-Assisted/methods , Brain Injuries, Traumatic/diagnostic imaging , Machine Learning , Algorithms , Deep Learning , InventionsABSTRACT
This study aims to assess the accuracy of automatic atlas-based contours for various key anatomical structures in prostate radiotherapy treatment planning. The evaluated structures include the bladder, rectum, prostate, seminal vesicles, femoral heads and penile bulb. CT images from 20 patients who underwent intensity-modulated radiotherapy were randomly chosen to create an atlas library. Atlas contours of the seven anatomical structures were generated using four software packages: ABAS, Eclipse, MIM, and RayStation. These contours were then compared to manual delineations performed by oncologists, which served as the ground truth. Evaluation metrics such as dice similarity coefficient (DSC), mean distance to agreement (MDA), and volume ratio (VR) were calculated to assess the accuracy of the contours. Additionally, the time taken by each software to generate the atlas contour was recorded. The mean DSC values for the bladder exhibited strong agreement (>0.8) with manual delineations for all software except for Eclipse and RayStation. Similarly, the femoral heads showed significant similarity between the atlas contours and ground truth across all software, with mean DSC values exceeding 0.9 and MDA values close to zero. On the other hand, the penile bulb displayed only moderate agreement with the ground truth, with mean DSC values ranging from 0.5 to 0.7 for all software. A similar trend was observed in the prostate atlas contours, except for MIM, which achieved a mean DSC of over 0.8. For the rectum, both ABAS and MIM atlases demonstrated strong agreement with the ground truth, resulting in mean DSC values of more than 0.8. Overall, MIM and ABAS outperformed Eclipse and RayStation in both DSC and MDA. These results indicate that the atlas-based segmentation employed in this study produces acceptable contours for the anatomical structures of interest in prostate radiotherapy treatment planning.
ABSTRACT
BACKGROUND: Increased afterload in aortic stenosis (AS) induces left ventricle (LV) remodeling to preserve a normal ejection fraction. This compensatory response can become maladaptive and manifest with motion abnormality. It is a clinical challenge to identify contractile and relaxation dysfunction during early subclinical stage to prevent irreversible deterioration. PURPOSE: To evaluate the changes of regional wall dynamics in 3D + time domain as remodeling progresses in AS. STUDY TYPE: Retrospective. POPULATION: A total of 31 AS patients with reduced and preserved ejection fraction (14 AS_rEF: 7 male, 66.5 [7.8] years old; 17 AS_pEF: 12 male, 67.0 [6.0] years old) and 15 healthy (6 male, 61.0 [7.0] years old). FIELD STRENGTH/SEQUENCE: 1.5 T Magnetic resonance imaging/steady state free precession and late-gadolinium enhancement sequences. ASSESSMENT: Individual LV models were reconstructed in 3D + time domain and motion metrics including wall thickening (TI), dyssynchrony index (DI), contraction rate (CR), and relaxation rate (RR) were automatically extracted and associated with the presence of scarring and remodeling. STATISTICAL TESTS: Shapiro-Wilk: data normality; Kruskal-Wallis: significant difference (P < 0.05); ICC and CV: variability; Mann-Whitney: effect size. RESULTS: AS_rEF group shows distinct deterioration of cardiac motions compared to AS_pEF and healthy groups (TIAS_rEF : 0.92 [0.85] mm, TIAS_pEF : 5.13 [1.99] mm, TIhealthy : 3.61 [1.09] mm, ES: 0.48-0.83; DIAS_rEF : 17.11 [7.89]%, DIAS_pEF : 6.39 [4.04]%, DIhealthy : 5.71 [1.87]%, ES: 0.32-0.85; CRAS_rEF : 8.69 [6.11] mm/second, CRAS_pEF : 16.48 [6.70] mm/second, CRhealthy : 10.82 [4.57] mm/second, ES: 0.29-0.60; RRAS_rEF : 8.45 [4.84] mm/second; RRAS_pEF : 13.49 [8.56] mm/second, RRhealthy : 9.31 [2.48] mm/second, ES: 0.14-0.43). The difference in the motion metrics between healthy and AS_pEF groups were insignificant (P-value = 0.16-0.72). AS_rEF group was dominated by eccentric hypertrophy (47.1%) with concomitant scarring. Conversely, AS_pEF group was dominated by concentric remodeling and hypertrophy (71.4%), which could demonstrate hyperkinesia with slight wall dyssynchrony than healthy. Dysfunction of LV mechanics corresponded to the presence of myocardial scarring (54.9% in AS), which reverted the compensatory mechanisms initiated and performed by LV remodeling. DATA CONCLUSION: The proposed 3D + time modeling technique may distinguish regional motion abnormalities between AS_pEF, AS_rEF, and healthy cohorts, aiding clinical diagnosis and monitoring of AS progression. Subclinical myocardial dysfunction is evident in early AS despite of normal EF. LEVEL OF EVIDENCE: 4 TECHNICAL EFFICACY: Stage 1.
Subject(s)
Aortic Valve Stenosis , Contrast Media , Humans , Male , Child , Retrospective Studies , Cicatrix , Gadolinium , Magnetic Resonance Imaging , Aortic Valve Stenosis/diagnostic imaging , Hypertrophy , Ventricular Function, Left , Stroke Volume , Ventricular RemodelingABSTRACT
In fluoroscopy-guided interventions (FGIs), obtaining large quantities of labelled data for deep learning (DL) can be difficult. Synthetic labelled data can serve as an alternative, generated via pseudo 2D projections of CT volumetric data. However, contrasted vessels have low visibility in simple 2D projections of contrasted CT data. To overcome this, we propose an alternative method to generate fluoroscopy-like radiographs from contrasted head CT Angiography (CTA) volumetric data. The technique involves segmentation of brain tissue, bone, and contrasted vessels from CTA volumetric data, followed by an algorithm to adjust HU values, and finally, a standard ray-based projection is applied to generate the 2D image. The resulting synthetic images were compared to clinical fluoroscopy images for perceptual similarity and subject contrast measurements. Good perceptual similarity was demonstrated on vessel-enhanced synthetic images as compared to the clinical fluoroscopic images. Statistical tests of equivalence show that enhanced synthetic and clinical images have statistically equivalent mean subject contrast within 25% bounds. Furthermore, validation experiments confirmed that the proposed method for generating synthetic images improved the performance of DL models in certain regression tasks, such as localizing anatomical landmarks in clinical fluoroscopy images. Through enhanced pseudo 2D projection of CTA volume data, synthetic images with similar features to real clinical fluoroscopic images can be generated. The use of synthetic images as an alternative source for DL datasets represents a potential solution to the application of DL in FGIs procedures.
Subject(s)
Deep Learning , Radiology, Interventional , Radiography , Fluoroscopy/methods , AlgorithmsABSTRACT
20% of brain tumor patients present with seizures at the onset of diagnosis, while a further 25-40% develop epileptic seizures as the tumor progresses. Tumor-related epilepsy (TRE) is a condition in which the tumor causes recurring, unprovoked seizures. The occurrence of TRE differs between patients, along with the effectiveness of treatment methods. Therefore, determining the tumor properties that correlate with epilepsy can help guide TRE treatment. This article reviews the MRI sequences and image post-processing algorithms in the study of TRE. It focuses on epilepsy caused by glioma tumors because it is the most common type of malignant brain tumor and it has a high prevalence of epilepsy. In correlational TRE studies, conventional MRI sequences and diffusion-weighted MRI (DWI) are used to extract variables related to the tumor radiological characteristics, called imaging factors. Image post-processing is used to correlate the imaging factors with the incidence of epilepsy. The earlier studies of TRE used univariate and multivariate analysis to study the correlations between specific variables and incidence of epilepsy. Later, studies used voxel-based morphometry and voxel lesion-symptom mapping. Radiomics has been recently used to post-process the images for the study of TRE. This article will discuss the limitation of the existing imaging modalities and post-processing algorithms. It ends with some suggestions and challenges for future TRE studies.
ABSTRACT
This study investigates the feasibility of using texture radiomics features extracted from mammography images to distinguish between benign and malignant breast lesions and to classify benign lesions into different categories and determine the best machine learning (ML) model to perform the tasks. Six hundred and twenty-two breast lesions from 200 retrospective patient data were segmented and analysed. Three hundred fifty radiomics features were extracted using the Standardized Environment for Radiomics Analysis (SERA) library, one of the radiomics implementations endorsed by the Image Biomarker Standardisation Initiative (IBSI). The radiomics features and selected patient characteristics were used to train selected machine learning models to classify the breast lesions. A fivefold cross-validation was used to evaluate the performance of the ML models and the top 10 most important features were identified. The random forest (RF) ensemble gave the highest accuracy (89.3%) and positive predictive value (66%) and likelihood ratio of 13.5 in categorising benign and malignant lesions. For the classification of benign lesions, the RF model again gave the highest likelihood ratio of 3.4 compared to the other models. Morphological and textural radiomics features were identified as the top 10 most important features from the random forest models. Patient age was also identified as one of the significant features in the RF model. We concluded that machine learning models trained against texture-based radiomics features and patient features give reasonable performance in differentiating benign versus malignant breast lesions. Our study also demonstrated that the radiomics-based machine learning models were able to emulate the visual assessment of mammography lesions, typically used by radiologists, leading to a better understanding of how the machine learning model arrive at their decision.
Subject(s)
Breast , Mammography , Humans , Retrospective Studies , Breast/diagnostic imaging , Mammography/methods , Machine Learning , Random ForestABSTRACT
Left ventricular adaptations can be a complex process under the influence of aortic stenosis (AS) and comorbidities. This study proposed and assessed the feasibility of using a motion-corrected personalized 3D + time LV modeling technique to evaluate the adaptive and maladaptive LV response to aid treatment decision-making. A total of 22 AS patients were analyzed and compared against 10 healthy subjects. The 3D + time analysis showed a highly distinct and personalized pattern of remodeling in individual AS patients which is associated with comorbidities and fibrosis. Patients with AS alone showed better wall thickening and synchrony than those comorbid with hypertension. Ischemic heart disease in AS caused impaired wall thickening and synchrony and systolic function. Apart from showing significant correlations to echocardiography and clinical MRI measurements (r: 0.70-0.95; p < 0.01), the proposed technique helped in detecting subclinical and subtle LV dysfunction, providing a better approach to evaluate AS patients for specific treatment, surgical planning, and follow-up recovery.
Subject(s)
Aortic Valve Stenosis , Ventricular Dysfunction, Left , Humans , Ventricular Function, Left/physiology , Heart Ventricles/diagnostic imaging , Magnetic Resonance Imaging , Echocardiography , Ventricular Dysfunction, Left/diagnostic imaging , Ventricular Dysfunction, Left/etiologyABSTRACT
BACKGROUND: In subjects with isolated growth hormone deficiency (IGHD), recombinant human growth hormone (rhGH) is an approved method to achieve potential mid-parental height. However, data reporting rhGH treatment response in terms of brain structure volumes were scarce. We report the volumetric changes of the pituitary gland, basal ganglia, corpus callosum, thalamus, hippocampus and amygdala in these subjects post rhGH treatment. MATERIALS AND METHODS: This was a longitudinal study of eight IGHD subjects (2 males, 6 females) with a mean age of 11.1 ± 0.8 years and age-matched control groups. The pituitary gland, basal ganglia and limbic structures volumes were obtained using 3T MRI voxel-based morphology. The left-hand bone age was assessed using the Tanner-Whitehouse method. Follow-up imaging was performed after an average of 1.8 ± 0.4 years on rhGH. RESULTS: Subjects with IGHD had a smaller mean volume of the pituitary gland, right thalamus, hippocampus, and amygdala than the controls. After rhGH therapy, these volumes normalized to the age-matched controls. Corpus callosum of IGHD subjects had a larger mean volume than the controls and did not show much volume changes in response to rhGH therapy. There were changes towards normalization of bone age deficit of IGHD in response to rhGH therapy. CONCLUSION: The pituitary gland, hippocampus, and amygdala volumes in IGHD subjects were smaller than age-matched controls and showed the most response to rhGH therapy. Semi-automated volumetric assessment of pituitary gland, hippocampus, and amygdala using MRI may provide an objective assessment of response to rhGH therapy.
Subject(s)
Dwarfism, Pituitary , Human Growth Hormone , Male , Female , Child , Humans , Human Growth Hormone/therapeutic use , Growth Hormone , Longitudinal Studies , Pituitary Gland/diagnostic imagingABSTRACT
Percutaneous coronary intervention (PCI) with stent placement is a treatment effective for coronary artery diseases. Intravascular optical coherence tomography (OCT) with high resolution is used clinically to visualize stent deployment and restenosis, facilitating PCI operation and for complication inspection. Automated stent struts segmentation in OCT images is necessary as each pullback of OCT images could contain thousands of stent struts. In this paper, a deep learning framework is proposed and demonstrated for the automated segmentation of two major clinical stent types: metal stents and bioresorbable vascular scaffolds (BVS). U-Net, the current most prominent deep learning network in biomedical segmentation, was implemented for segmentation with cropped input. The architectures of MobileNetV2 and DenseNet121 were also adapted into U-Net for improvement in speed and accuracy. The results suggested that the proposed automated algorithm's segmentation performance approaches the level of independent human obsevers and is feasible for both types of stents despite their distinct appearance. U-Net with DenseNet121 encoder (U-Dense) performed best with Dice's coefficient of 0.86 for BVS segmentation, and precision/recall of 0.92/0.92 for metal stent segmentation under optimal crop window size of 256.
Subject(s)
Deep Learning , Drug-Eluting Stents , Percutaneous Coronary Intervention , Absorbable Implants , Coronary Angiography/methods , Coronary Vessels , Humans , Stents , Tomography, Optical Coherence/methods , Treatment OutcomeABSTRACT
PURPOSE: This systematic review aims to understand the dose estimation approaches and their major challenges. Specifically, we focused on state-of-the-art Monte Carlo (MC) methods in fluoroscopy-guided interventional procedures. METHODS: All relevant studies were identified through keyword searches in electronic databases from inception until September 2020. The searched publications were reviewed, categorised and analysed based on their respective methodology. RESULTS: Hundred and one publications were identified which utilised existing MC-based applications/programs or customised MC simulations. Two outstanding challenges were identified that contribute to uncertainties in the virtual simulation reconstruction. The first challenge involves the use of anatomical models to represent individuals. Currently, phantom libraries best balance the needs of clinical practicality with those of specificity. However, mismatches of anatomical variations including body size and organ shape can create significant discrepancies in dose estimations. The second challenge is that the exact positioning of the patient relative to the beam is generally unknown. Most dose prediction models assume the patient is located centrally on the examination couch, which can lead to significant errors. CONCLUSION: The continuing rise of computing power suggests a near future where MC methods become practical for routine clinical dosimetry. Dynamic, deformable phantoms help to improve patient specificity, but at present are only limited to adjustment of gross body volume. Dynamic internal organ displacement or reshaping is likely the next logical frontier. Image-based alignment is probably the most promising solution to enable this, but it must be automated to be clinically practical.
Subject(s)
Radiometry , Fluoroscopy , Humans , Monte Carlo Method , Phantoms, Imaging , Radiation DosageABSTRACT
BACKGROUND: Charcot-Marie-Tooth (CMT) disease is diagnosed through clinical findings and genetic testing. While there are neurophysiological tools and clinical functional scales in CMT, objective disease biomarkers that can facilitate in monitoring disease progression are limited. PURPOSE: To investigate the utility of diffusion tensor imaging (DTI) in determining the microstructural integrity of sciatic and peroneal nerves and its correlation with the MRI grading of muscle atrophy severity and clinical function in CMT as determined by the CMT neuropathy score (CMTNS). STUDY TYPE: Prospective case-control. SUBJECTS: Nine CMT patients and nine age-matched controls. FIELD STRENGTH/SEQUENCE: 3 T T1 -weighted in-/out-of phase spoiled gradient recalled echo (SPGR) and DTI sequences. ASSESSMENT: Fractional anisotropy (FA), axial diffusivity (AD), radial diffusivity (RD), and mean diffusivity (MD) values for sciatic and peroneal nerves were obtained from DTI. Muscle atrophy was graded according to the Goutallier classification using in-/out-of phase SPGRs. DTI parameters and muscle atrophy grades were compared between CMT and controls, and the relationship between DTI parameters, muscle atrophy grades, and CMTNS were assessed. STATISTICAL TESTS: The Wilcoxon Signed Ranks test was used to compare DTI parameters between CMT and controls. The relationship between DTI parameters, muscle atrophy grades, and CMTNS were analyzed using the Spearman correlation. Receiver operating characteristic (ROC) analyses of DTI parameters that can differentiate CMT from healthy controls were done. RESULTS: There was a significant reduction in FA and increase in RD of both nerves (P < 0.05) in CMT, with significant correlations between FA (negative; P < 0.05) and RD (positive; P < 0.05) with muscle atrophy grade. In the sciatic nerve, there was significant correlation between FA and CMTNS (r = -0.795; P < 0.05). FA and RD could discriminate CMT from controls with high sensitivity (77.8-100%) and specificity (88.9-100%). DATA CONCLUSION: There were significant differences of DTI parameters between CMT and controls, with significant correlations between DTI parameters, muscle atrophy grade, and CMTNS. Level of Evidence 2 Technical Efficacy Stage 2 J. MAGN. RESON. IMAGING 2021;53:437-444.
Subject(s)
Charcot-Marie-Tooth Disease , Diffusion Tensor Imaging , Anisotropy , Charcot-Marie-Tooth Disease/diagnostic imaging , Humans , Magnetic Resonance Imaging , Peripheral Nerves/diagnostic imaging , Prospective StudiesABSTRACT
PURPOSE: We present the implementation of e-learning in the Master of Medical Physics programme at the University of Malaya during a partial lockdown from March to June 2020 due to the COVID-19 pandemic. METHODS: Teaching and Learning (T&L) activities were conducted virtually on e-learning platforms. The students' experience and feedback were evaluated after 15 weeks. RESULTS: We found that while students preferred face-to-face, physical teaching, they were able to adapt to the new norm of e-learning. More than 60% of the students agreed that pre-recorded lectures and viewing videos of practical sessions, plus answering short questions, were beneficial. Certain aspects, such as hands-on practical and clinical experience, could never be replaced. The e-learning and study-from-home environment accorded a lot of flexibility. However, students also found it challenging to focus because of distractions, lack of engagement and mental stress. Technical problems, such as poor Internet connectivity and limited data plans, also compounded the problem. CONCLUSION: We expect e-learning to prevail in future. Hybrid learning strategies, which includes face-to-face classes and e-learning, will become common, at least in the medical physics programme of the University of Malaya even after the pandemic.
Subject(s)
COVID-19/epidemiology , COVID-19/prevention & control , Computer-Assisted Instruction/methods , Education, Distance/methods , Educational Technology/methods , Humans , Internet , Learning , Malaysia , Program Development , Program Evaluation/methods , Students, Medical , Teaching , UniversitiesABSTRACT
Differential diagnosis of hypertensive heart disease (HHD) and hypertrophic cardiomyopathy (HCM) is clinically challenging but important for treatment management. This study aims to phenotype HHD and HCM in 3D + time domain by using a multiparametric motion-corrected personalized modeling algorithm and cardiac magnetic resonance (CMR). 44 CMR data, including 12 healthy, 16 HHD and 16 HCM cases, were examined. Multiple CMR phenotype data consisting of geometric and dynamic variables were extracted globally and regionally from the models over a full cardiac cycle for comparison against healthy models and clinical reports. Statistical classifications were used to identify the distinctive characteristics and disease subtypes with overlapping functional data, providing insights into the challenges for differential diagnosis of both types of disease. While HCM is characterized by localized extreme hypertrophy of the LV, wall thickening/contraction/strain was found to be normal and in sync, though it was occasionally exaggerated at normotrophic/less severely hypertrophic regions during systole to preserve the overall ejection fraction (EF) and systolic functionality. Additionally, we observed that hypertrophy in HHD could also be localized, although at less extreme conditions (i.e. more concentric). While fibrosis occurs mostly in those HCM cases with aortic obstruction, only minority of HHD patients were found affected by fibrosis. We demonstrate that subgroups of HHD (i.e. preserved and reduced EF: HHDpEF & HHDrEF) have different 3D + time CMR characteristics. While HHDpEF has cardiac functions in normal range, dilation and heart failure are indicated in HHDrEF as reflected by low LV wall thickening/contraction/strain and synchrony, as well as much reduced EF.
Subject(s)
Cardiomyopathy, Hypertrophic , Heart Diseases , Hypertension , Cardiomyopathy, Hypertrophic/diagnostic imaging , Humans , Magnetic Resonance Imaging , Magnetic Resonance Imaging, CineABSTRACT
BACKGROUND: Intrathecal and intravenous chemotherapy, specifically methotrexate, might contribute to neural microstructural damage. OBJECTIVE: To assess, by diffusion tensor imaging, microstructural integrity of white matter in paediatric patients with acute lymphoblastic leukaemia (ALL) following intrathecal and intravenous chemotherapy. MATERIALS AND METHODS: Eleven children diagnosed with de novo ALL underwent MRI scans of the brain with diffusion tensor imaging (DTI) prior to commencement of chemotherapy and at 12 months after diagnosis, using a 3-tesla (T) MRI scanner. We investigated the changes in DTI parameters in white matter tracts before and after chemotherapy using tract-based spatial statistics overlaid on the International Consortium of Brain Mapping DTI-81 atlas. All of the children underwent formal neurodevelopmental assessment at the two study time points. RESULTS: Whole-brain DTI analysis showed significant changes between the two time points, affecting several white matter tracts. The tracts demonstrated longitudinal changes of decreasing mean and radial diffusivity. The neurodevelopment of the children was near compatible for age at the end of ALL treatment. CONCLUSION: The quantification of white matter tracts changes in children undergoing chemotherapy showed improving longitudinal values in DTI metrics (stable fractional anisotropy, decreasing mean and radial diffusivity), which are incompatible with deterioration of microstructural integrity in these children.
Subject(s)
Antimetabolites, Antineoplastic/adverse effects , Diffusion Tensor Imaging , Methotrexate/adverse effects , Precursor Cell Lymphoblastic Leukemia-Lymphoma/drug therapy , White Matter/drug effects , White Matter/ultrastructure , Anisotropy , Brain Mapping/methods , Child , Child, Preschool , Female , Humans , Image Interpretation, Computer-Assisted , MaleABSTRACT
BACKGROUND AND OBJECTIVES: Continuous monitoring of physiological parameters such as photoplethysmography (PPG) has attracted increased interest due to advances in wearable sensors. However, PPG recordings are susceptible to various artifacts, and thus reducing the reliability of PPG-driven parameters, such as oxygen saturation, heart rate, blood pressure and respiration. This paper proposes a one-dimensional convolution neural network (1-D-CNN) to classify five-second PPG segments into clean or artifact-affected segments, avoiding data-dependent pulse segmentation techniques and heavy manual feature engineering. METHODS: Continuous raw PPG waveforms were blindly allocated into segments with an equal length (5s) without leveraging any pulse location information and were normalized with Z-score normalization methods. A 1-D-CNN was designed to automatically learn the intrinsic features of the PPG waveform, and perform the required classification. Several training hyperparameters (initial learning rate and gradient threshold) were varied to investigate the effect of these parameters on the performance of the network. Subsequently, this proposed network was trained and validated with 30 subjects, and then tested with eight subjects, with our local dataset. Moreover, two independent datasets downloaded from the PhysioNet MIMIC II database were used to evaluate the robustness of the proposed network. RESULTS: A 13 layer 1-D-CNN model was designed. Within our local study dataset evaluation, the proposed network achieved a testing accuracy of 94.9%. The classification accuracy of two independent datasets also achieved satisfactory accuracy of 93.8% and 86.7% respectively. Our model achieved a comparable performance with most reported works, with the potential to show good generalization as the proposed network was evaluated with multiple cohorts (overall accuracy of 94.5%). CONCLUSION: This paper demonstrated the feasibility and effectiveness of applying blind signal processing and deep learning techniques to PPG motion artifact detection, whereby manual feature thresholding was avoided and yet a high generalization ability was achieved.
Subject(s)
Artifacts , Photoplethysmography , Algorithms , Heart Rate , Humans , Motion , Neural Networks, Computer , Reproducibility of Results , Signal Processing, Computer-AssistedABSTRACT
OBJECTIVES: To measure the clinical, structural and functional changes of an individualised structured cognitive rehabilitation in mild traumatic brain injury (mTBI) population. SETTING: A single centre study, Malaysia. PARTICIPANTS: Adults aged between 18 and 60 years with mTBI as a result of road traffic accident, with no previous history of head trauma, minimum of 9 years education and abnormal cognition at 3 months will be included. The exclusion criteria include pre-existing chronic illness or neurological/psychiatric condition, long-term medication that affects cognitive/psychological status, clinical evidence of substance intoxication at the time of injury and major polytrauma. Based on multiple estimated calculations, the minimum intended sample size is 50 participants (Cohen's d effect size=0.35; alpha level of 0.05; 85% power to detect statistical significance; 40% attrition rate). INTERVENTIONS: Intervention group will receive individualised structured cognitive rehabilitation. Control group will receive the best patient-centred care for attention disorders. Therapy frequency for both groups will be 1 hour per week for 12 weeks. OUTCOME MEASURES: Primary: Neuropsychological Assessment Battery-Screening Module (S-NAB) scores. Secondary: Diffusion Tensor Imaging (DTI) parameters and Goal Attainment Scaling score (GAS). RESULTS: Results will include descriptive statistics of population demographics, CogniPlus cognitive program and metacognitive strategies. The effect of intervention will be the effect size of S-NAB scores and mean GAS T scores. DTI parameters will be compared between groups via repeated measure analysis. Correlation analysis of outcome measures will be calculated using Pearson's correlation coefficient. CONCLUSION: This is a complex clinical intervention with multiple outcome measures to provide a comprehensive evidence-based treatment model. ETHICS AND DISSEMINATION: The study protocol was approved by the Medical Research Ethics Committee UMMC (MREC ID NO: 2016928-4293). The findings of the trial will be disseminated through peer-reviewed journals and scientific conferences. TRIAL REGISTRATION NUMBER: NCT03237676.
Subject(s)
Accidental Injuries/complications , Attention Deficit Disorder with Hyperactivity/rehabilitation , Brain Concussion/rehabilitation , Cognitive Behavioral Therapy/instrumentation , Attention Deficit Disorder with Hyperactivity/diagnosis , Brain Concussion/diagnosis , Cognition , Diffusion Tensor Imaging , Humans , Malaysia , Neuropsychological Tests , Randomized Controlled Trials as Topic , Recovery of Function , Surveys and QuestionnairesABSTRACT
PURPOSE: To evaluate a 2D-4D registration-cum-segmentation framework for the delineation of left ventricle (LV) in late gadolinium enhanced (LGE) MRI and for the localization of infarcts in patient-specific 3D LV models. METHODS: A 3-step framework was proposed, consisting of: (1) 3D LV model reconstruction from motion-corrected 4D cine-MRI; (2) Registration of 2D LGE-MRI with 4D cine-MRI; (3) LV contour extraction from the intersection of LGE slices with the LV model. The framework was evaluated against cardiac MRI data from 27 patients scanned within 6 months after acute myocardial infarction. We compared the use of local Pearson's correlation (LPC) and normalized mutual information (NMI) as similarity measures for the registration. The use of 2 and 6 long-axis (LA) cine-MRI scans was also compared. The accuracy of the framework was evaluated using manual segmentation, and the interobserver variability of the scar volume derived from the segmented LV was determined using Bland-Altman analysis. RESULTS: LPC outperformed NMI as a similarity measure for the proposed framework using 6 LA scans, with Hausdorrf distance (HD) of 1.19 ± 0.53 mm versus 1.51 ± 2.01 mm (endocardial) and 1.21 ± 0.48 mm versus 1.46 ± 1.78 mm (epicardial), respectively. Segmentation using 2 LA scans was comparable to 6 LA scans with a HD of 1.23 ± 0.70 mm (endocardial) and 1.25 ± 0.74 mm (epicardial). The framework yielded a lower interobserver variability in scar volumes compared with manual segmentation. CONCLUSION: The framework showed high accuracy and robustness in delineating LV in LGE-MRI and allowed for bidirectional mapping of information between LGE- and cine-MRI scans, crucial in personalized model studies for treatment planning.
Subject(s)
Gadolinium/chemistry , Heart Ventricles/diagnostic imaging , Imaging, Three-Dimensional , Magnetic Resonance Imaging , Myocardial Infarction/diagnostic imaging , Algorithms , Computer Simulation , Humans , Image Processing, Computer-Assisted , Motion , Observer Variation , Prognosis , Radiotherapy Planning, Computer-Assisted , Reproducibility of ResultsABSTRACT
RATIONALE AND OBJECTIVES: Magnetic resonance spectroscopy is a noninvasive imaging technique that allows for reliable assessment of microscopic changes in brain cytoarchitecture, neuronal injuries, and neurochemical changes resultant from traumatic insults. We aimed to evaluate the acute alteration of neurometabolites in complicated and uncomplicated mild traumatic brain injury (mTBI) patients in comparison to control subjects using proton magnetic resonance spectroscopy (1H magnetic resonance spectroscopy). MATERIAL AND METHODS: Forty-eight subjects (23 complicated mTBI [cmTBI] patients, 12 uncomplicated mTBI [umTBI] patients, and 13 controls) underwent magnetic resonance imaging scan with additional single voxel spectroscopy sequence. Magnetic resonance imaging scans for patients were done at an average of 10 hours (standard deviation 4.26) post injury. The single voxel spectroscopy adjacent to side of injury and noninjury regions were analysed to obtain absolute concentrations and ratio relative to creatine of the neurometabolites. One-way analysis of variance was performed to compare neurometabolite concentrations of the three groups, and a correlation study was done between the neurometabolite concentration and Glasgow Coma Scale. RESULTS: Significant difference was found in ratio of N-acetylaspartate to creatine (NAA/Cr + PCr) (χ2(2) = 0.22, P < .05) between the groups. The sum of NAA and N-acetylaspartylglutamate (NAAG) also shows significant differences in both the absolute concentration (NAA + NAAG) and ratio to creatine (NAA + NAAG/Cr + PCr) between groups (χ2(2) = 4.03, P < .05and (χ2(2) = 0.79, P < .05)). NAA values were lower in cmTBI and umTBI compared to control group. A moderate weak positive correlation were found between Glasgow Coma Scale with NAA/Cr + PCr (ρ = 0.36, P < .05 and NAA + NAAG/Cr + PCr (ρ = 0.45, P < .05)), whereas a moderate correlation was seen with NAA + NAAG (ρ = 0.38, P < .05). CONCLUSION: Neurometabolite alterations were already apparent at onset of both complicated and uncomplicated traumatic brain injury. The ratio of NAA and NAAG has potential to serve as a biomarker reflecting injury severity in a quantifiable manner as it discriminates between the complicated and uncomplicated cases of mTBI.