Search

Segmentation of the fetus from 2-dimensional (2D) magnetic resonance imaging (MRI) can aid radiologists with clinical decision making for disease diagnosis. Machine learning can facilitate this process of automatic segmentation, making diagnosis more accurate and user independent. We propose a deep learning (DL) framework for 2D fetal MRI segmentation using a Cross Attention Squeeze Excitation Network (CASE-Net) for research and clinical applications. CASE-Net is an end-to-end segmentation architecture with relevant modules that are evidence based. The goal of CASE-Net is to emphasize localization of contextual information that is relevant in biomedical segmentation, by combining attention mechanisms with squeeze-and-excitation (SE) blocks. This is a retrospective study with 34 patients. Our experiments have shown that our proposed CASE-Net achieved the highest segmentation Dice score of 87.36%, outperforming other competitive segmentation architectures.

Purpose: Development of a warp-based automated brain segmentation approach of 3D fluid-attenuated inversion recovery (FLAIR) images and comparison to 3D T1-based segmentation., Methods: 3D FLAIR and 3D T1-weighted sequences of 30 healthy subjects (mean age 29.9 ± 8.3 years, 8 female) were acquired on the same 3T MR scanner. Warp-based segmentation was applied for volumetry of total gray matter (GM), white matter (WM), and 116 atlas regions. Segmentation results of both sequences were compared using Pearson correlation (r)., Results: Correlation of GM segmentation results based on FLAIR and T1 was overall good for cortical structures (mean r across all cortical structures = 0.76). Comparatively weaker results were found in the occipital lobe (r = 0.77), central region (mean r = 0.58), basal ganglia (mean r = 0.59), thalamus (r = 0.30), and cerebellum (r = 0.73). FLAIR segmentation underestimated volume of the central region compared to T1, but showed a better anatomic concordance with the occipital lobe on visual review and subcortical structures, when also compared to manual segmentation. Visual analysis of FLAIR-based WM segmentation revealed frequent misclassification of regions of high signal intensity as GM., Conclusion: Warp-based FLAIR segmentation yields comparable results to T1 segmentation for most cortical GM structures and may provide anatomically more congruent segmentation of subcortical GM structures. Selected cortical regions, especially the central region and total WM, seem to be underestimated on FLAIR segmentation.

Objectives: The aim of this study was to investigate whether radiomic analysis with random survival forests (RSFs) can predict overall survival from T1-weighted contrast-enhanced baseline magnetic resonance imaging (MRI) scans in a cohort of glioblastoma multiforme (GBM) patients with uniform treatment., Materials and Methods: This retrospective study was approved by the institutional review board and informed consent was waived. The MRI scans from 66 patients with newly diagnosed GBM from a previous prospective study were analyzed. Tumors were segmented manually on contrast-enhanced 3-dimensional T1-weighted images. Using these segmentations, P = 208 quantitative image features characterizing tumor shape, signal intensity, and texture were calculated in an automated fashion. On this data set, an RSF was trained using 10-fold cross validation to establish a link between image features and overall survival, and the individual risk for each patient was predicted. The mean concordance index was assessed as a measure of prediction accuracy. Association of individual risk with overall survival was assessed using Kaplan-Meier analysis and a univariate proportional hazards model., Results: Mean overall survival was 14 months (range, 0.8-85 months). Mean concordance index of the 10-fold cross-validated RSF was 0.67. Kaplan-Meier analysis clearly distinguished 2 patient groups with high and low predicted individual risk (P = 5.5 × 10). Low predicted individual mortality was found to be a favorable prognostic factor for overall survival in a univariate Cox proportional hazards model (hazards ratio, 1.038; 95% confidence interval, 1.015-1.062; P = 0.0059)., Conclusions: This study demonstrates that baseline MRI in GBM patients contains prognostic information, which can be accessed by radiomic analysis using RSFs.

Objective: The aim of this study was to test the diagnostic value of wavelet-based angiographic reconstruction of CT perfusion data (waveletCTA) to detect cerebral venous sinus thrombosis (CVST) in patients who underwent whole-brain CT perfusion imaging (WB-CTP)., Materials and Methods: Datasets were retrospectively selected from an initial cohort of 2863 consecutive patients who had undergone multiparametric CT including WB-CTP. WaveletCTA was reconstructed from WB-CTP: the angiographic signal was generated by voxel-based wavelet transform of time attenuation curves (TACs) from WB-CTP raw data. In a preliminary clinical evaluation, waveletCTA was analyzed by 2 readers with respect to presence and location of CVST. Venous CT and MR angiography (venCTA/venMRA) served as reference standard. Diagnostic confidence for CVST detection and the quality of depiction for venous sections were evaluated on 5-point Likert scales. Thrombus extent was assessed by length measurements. The mean CT attenuation and waveletCTA signal of the thrombus and of flowing blood were quantified., Results: Sixteen patients were included: 10 patients with venCTA-/venMRA-confirmed CVST and 6 patients with arterial single-phase CT angiography (artCTA)-suspected but follow-up-excluded CVST. The reconstruction of waveletCTA was successful in all patients. Among the patients with confirmed CVST, waveletCTA correctly demonstrated presence, location, and extent of the thrombosis in 10/10 cases. In 6 patients with artCTA-suspected but follow-up-excluded CVST, waveletCTA correctly ruled out CVST in 5 patients. Reading waveletCTA in addition to artCTA significantly increased the diagnostic confidence concerning CVST compared with reading artCTA alone (4.4 vs 3.6, P = 0.044). The mean flowing blood-to-thrombus ratio was highest in waveletCTA, followed by venCTA and artCTA (146.2 vs 5.9 vs 2.6, each with P < 0.001). In waveletCTA, the venous sections were depicted better compared with artCTA (4.2 vs 2.6, P < 0.001), and equally well compared with venCTA/venMRA (4.2 vs 4.1, P = 0.374)., Conclusions: WaveletCTA was technically feasible in CVST patients and reliably identified CVST in a preliminary clinical evaluation. WaveletCTA might serve as an additional reconstruction to rule out or incidentally detect CVST in patients who undergo WB-CTP.

Purpose: The purpose of this study was to propose and evaluate a new wavelet-based technique for classification of arterial and venous vessels using time-resolved cerebral CT perfusion data sets., Methods: Fourteen consecutive patients (mean age 73 yr, range 17-97) with suspected stroke but no pathology in follow-up MRI were included. A CT perfusion scan with 32 dynamic phases was performed during intravenous bolus contrast-agent application. After rigid-body motion correction, a Paul wavelet (order 1) was used to calculate voxelwise the wavelet power spectrum (WPS) of each attenuation-time course. The angiographic intensity A was defined as the maximum of the WPS, located at the coordinates T (time axis) and W (scale/width axis) within the WPS. Using these three parameters (A, T, W) separately as well as combined by (1) Fisher's linear discriminant analysis (FLDA), (2) logistic regression (LogR) analysis, or (3) support vector machine (SVM) analysis, their potential to classify 18 different arterial and venous vessel segments per subject was evaluated., Results: The best vessel classification was obtained using all three parameters A and T and W [area under the curve (AUC): 0.953 with FLDA and 0.957 with LogR or SVM]. In direct comparison, the wavelet-derived parameters provided performance at least equal to conventional attenuation-time-course parameters. The maximum AUC obtained from the proposed wavelet parameters was slightly (although not statistically significantly) higher than the maximum AUC (0.945) obtained from the conventional parameters., Conclusions: A new method to classify arterial and venous cerebral vessels with high statistical accuracy was introduced based on the time-domain wavelet transform of dynamic CT perfusion data in combination with linear or nonlinear multidimensional classification techniques.

Objectives: We aimed to volumetrically quantify endolymph and perilymph spaces of the inner ear in order to establish a methodological basis for further investigations into the pathophysiology and therapeutic monitoring of Menière's disease., Methods: Sixteen patients (eight females, aged 38-71 years) with definite unilateral Menière's disease were included in this study. Magnetic resonance (MR) cisternography with a T2-SPACE sequence was combined with a Real reconstruction inversion recovery (Real-IR) sequence for delineation of inner ear fluid spaces. Machine learning and automated local thresholding segmentation algorithms were applied for three-dimensional (3D) reconstruction and volumetric quantification of endolymphatic hydrops. Test-retest reliability was assessed by the intra-class coefficient; correlation of cochlear endolymph volume ratio with hearing function was assessed by the Pearson correlation coefficient., Results: Endolymph volume ratios could be reliably measured in all patients, with a mean (range) value of 15% (2-25) for the cochlea and 28% (12-40) for the vestibulum. Test-retest reliability was excellent, with an intra-class coefficient of 0.99. Cochlear endolymphatic hydrops was significantly correlated with hearing loss (r = 0.747, p = 0.001)., Conclusions: MR imaging after local contrast application and image processing, including machine learning and automated local thresholding, enable the volumetric quantification of endolymphatic hydrops. This allows for a quantitative assessment of the effect of therapeutic interventions on endolymphatic hydrops., Key Points: • Endolymphatic hydrops is the pathological hallmark of Menière's disease. • Endolymphatic hydrops can be visualized by locally enhanced ultra-high-resolution MR imaging. • Computer-aided image processing enables quantification of endolymphatic hydrops. • Endolymphatic hydrops correlates with hearing loss in patients with Menière's disease. • Therapeutic trials in Menière's disease can be monitored with this quantitative approach.

Aim: In children with bilateral spastic cerebral palsy (CP), periventricular leukomalacia (PVL) is commonly identified on magnetic resonance imaging. We characterized this white matter condition by examining callosal microstructure, interhemispheric inhibitory competence (IIC), and mirror movements., Method: We examined seven children (age range 11y 9mo-17y 9mo, median age 15y 10mo, four females, three males) with bilateral spastic CP/PVL (Gross Motor Function Classification System level I or II, Manual Ability Classification System level I) and 12 age-matched controls (age range 11y 7mo-17y 1mo, median age 15y 6mo, seven females, five males). Fractional anisotropy of the transcallosal motor fibres (TCMF) and the corticospinal tract (CST) of both sides were calculated. The parameters of IIC (transcranial magnetic stimulation) and mirror movements were measured using a standardized clinical examination and a computer-based hand motor test., Results: Fractional anisotropy was lower in children with bilateral spastic CP/PVL regarding the TCMF, but not the left or right CST. Resting motor threshold was elevated in children with bilateral spastic CP/PVL whereas measures of IIC tended to be lower. Mirror movements were markedly elevated in bilateral spastic CP/PVL., Interpretation: This study provides new information on different aspects of motor function in children with bilateral spastic CP/PVL. Decreased fractional anisotropy of TCMF is consistent with impairment of hand motor function in children with bilateral spastic CP/PVL. The previously overlooked microstructure of the TCMF may serve as a potential indicator for hand motor function in patients with bilateral spastic CP/PVL., (© The Authors. Journal compilation © Mac Keith Press 2010.)

Purpose: As a Mardi Gras joke, the original jam or custard fillings of German Mardi Gras donuts are frequently replaced with mustard which cannot be identified on the outside of the donut. The aim of our study was to evaluate the impact of modern CT and MRI techniques on the diagnostic evaluation of donuts filled with mustard, jam or custard., Materials and Methods: 4 commercially available donuts were included in the study. One was filled with custard (PK) and one with jam (MK). Two donuts were specifically prepared and filled with Bavarian mustard (SK1) or extra-hot (SK2) mustard. 3 Tesla MRI was performed with T 2- and T 1-weighted STIR, diffusion and susceptibility-weighted (SWI) sequences. In addition, the donuts underwent dual-energy CT., Results: PK was able to be easily differentiated from the other donuts due to its hyperintensity in the STIR sequences and hypointensity in the T 1-weighted sequences. MK was able to be differentiated from S 1K and S 2K on the basis of its diffusion properties. S 1K demonstrated a pronounced heterogeneity of its matrix, especially in SWI. In CT, PK showed a pronounced hypoattenuation with negative Hounsfield units in contrast to the strongly hyperdense MK, S 1K and S 2K. S 1K and S 2K demonstrated X-ray attenuation considerably dependent on the X-ray energy., Conclusion: Donuts filled with jam, custard or mustard can be readily differentiated by modern MRI and CT techniques. Therefore, eating a mustard-filled donut can be reliably avoided.

The corpus callosum is formed between the 7th and the 20th gestational week. If this process is disrupted, partial or complete callosal agenesis may ensue. As large parts of the supra- and infratentorial brain are created during this critical period, associated anomalies need always to be searched for when callosal agenesis is present. Associations with neuro-genetic syndromes also exist. The corpus callosum is generally formed from front to back ("front-to-back rule"). Therefore, a partial callosal agenesis usually involves the posterior portion of the corpus callosum, while a secondary lesion of the corpus callosum does not follow this rule. Holoprosencephalies are a notable exception to this rule, as the frontal part of the corpus callosum is absent in spite of their classification as congenital malformations. They represent a disturbance of the differentiation and cleavage of the prosencephalon with a disruption of the separation of the cerebral hemispheres. Holoprosencephalies can be due to genetic causes, but also to intrauterine infections or other teratogenic causes. The holoprosencephalies are subdivided into alobar, semilobar and lobar holoprosencephalies. This article aims to describe the most important features of callosal agenesis and holoprosencephalies highlighting the respective imaging characteristics.

Disorders of supratentorial cortical development are usually divided into disorders of neuronal proliferation, neuronal migration and cortical organization. Based upon molecular biologic discoveries, a modified classification has recently been proposed. The category of malformations of abnormal neuronal and glial proliferation and apoptosis now includes microlissencephalies, megalencephalies, hemimegalencephalies and cortical dysplasias with balloon cells. Malformations due to abnormal neuronal migration now subsume the lissencephaly spectrum including the subcortical band heterotopias, the cobblestone complex and the group of heterotopias. Malformations due to abnormal cortical organization include the spectrum of polymicrogyria and schizencephaly as well as cortical dysplasias without balloon cells. High-resolution magnetic resonance imaging (MRI) has led to an increasing awareness of these malformations. This article aims to illustrate the classification, MRI presentation and relevant clinical features of the most commonly encountered disorders of cortical development.

Purpose: The newly developed 16-slice multi detector row computed tomography (MDCT) allows for isotropic imaging at rotation times under 500 ms. We aim to describe our first experiences with this novel technology in the diagnostic imaging of the craniocervical vascular system., Methods: The craniocervical vascular system of 8 consecutive patients was examined with a standardized protocol with the new 16-slice MDCT. The volume covered included the carotid arteries from their origin at the aortic arch and the intracranial arterial and venous vessels in their entirety. Homogeneity of contrast, contrast-to-parenchyma ratios, image quality and radiation doses were assessed., Results: A homogeneous vessel-to-parenchyma ratio was reached in the carotids as well as in the intracranial arterial and venous vessels. Image quality was rated as excellent in all patients. The estimated mean effective radiation dose amounted to 2.31 mSv +/-0.57, the mean dose length product to 608 mGy*cm +/-151., Conclusion: The newly developed 16-slice MDCT technology appears to be a promising tool in the diagnostic assessment of the cervicocranial vessels.

Purpose: Evaluation of the accuracy of a new electromagnetic target system for interventional CT-guided procedures with virtual navigation in a previously acquired helical CT., Material and Methods: The new target system CT-Guide 1010 (Ultraguide, Tirat Hacarmel, Israel) for CT-guided interventions was adapted to the video signal of the Somatom Plus 4 and Volume Zoom (Siemens, Erlangen, Germany). A helical CT-dataset including skin-based sensor cubes was transferred to the integrated navigation system inside the scanner room. 50 image-guided interventions and biopsies were performed outside the gantry using virtual navigation to reach the lesion. The accuracy of the procedures was evaluated using documentation of the needle tip with CT-fluoroscopy, results of histology, and follow-up., Results: The deviation between planned and documented needle tip was 2.2 +/- 2.1 mm in 50 procedures. Time between the end of planning-CT and needle positioning using the system was 13 minutes. There were no complications due to the use of the system., Conclusion: The CT-Guide allows for virtual real-time navigation with high accuracy. Advantages are the free needle angulation without gantry tilt, use of optimal CT perfusion phase for virtual navigation, and reduction of radiation exposure to the patient and interventionalist.