Synthetic MR Imaging-Based WM Signal Suppression Identifies Neonatal Brainstem Pathways in Vivo.

BACKGROUND AND PURPOSE: Multidynamic multiecho sequence-based imaging enables investigators to reconstruct multiple MR imaging contrasts on the basis of a single scan. This study investigated the feasibility of synthetic MRI-based WM signal suppression (syWMSS), a synthetic inversion recovery approach in which a short TI suppresses myelin-related signals, for the identification of early myelinating brainstem pathways. MATERIALS AND METHODS: Thirty-one cases of neonatal MR imaging, which included multidynamic multiecho data and conventionally acquired T1- and T2-weighted sequences, were analyzed. The multidynamic multiecho postprocessing software SyMRI was used to generate syWMSS data (TR/TE/TI = 3000/5/410 ms). Two raters discriminated early myelinating brainstem pathways (decussation of the superior cerebellar peduncle, medial lemniscus, central tegmental tract, and medial longitudinal fascicle [the latter 3 assessed at the level of the pons]) on syWMSS data and reference standard contrasts. RESULTS: On the basis of syWMSS data, the decussation of the superior cerebellar peduncle (31/31); left/right medial lemniscus (31/31; 30/31); left/right central tegmental tract (19/31; 20/31); and left/right medial longitudinal fascicle (30/31) were reliably identified by both raters. On the basis of T1-weighted contrasts, the decussation of the superior cerebellar peduncle (14/31); left/right medial lemniscus (22/31; 16/31); left/right central tegmental tract (1/31); and left/right medial longitudinal fascicle (9/31; 8/31) were reliably identified by both raters. On the basis of T2-weighted contrasts, the decussation of the superior cerebellar peduncle (28/31); left/right medial lemniscus (16/31; 12/31); left/right central tegmental tract (23/31; 18/31); and left/right medial longitudinal fascicle (15/31; 14/31) were reliably identified by both raters. CONCLUSIONS: syWMSS data provide a feasible imaging technique with which to study early myelinating brainstem pathways. MR imaging approaches that use myelin signal suppression contribute to a more sensitive assessment of myelination patterns at early stages of cerebral development.

Different from the Beginning: WM Maturity of Female and Male Extremely Preterm Neonates-A Quantitative MRI Study.

BACKGROUND AND PURPOSE: Former preterm born males are at higher risk for neurodevelopmental disabilities compared with female infants born at the same gestational age. This retrospective study investigated sex-related differences in the maturity of early myelinating brain regions in infants born <28 weeks' gestational age using diffusion tensor- and relaxometry-based MR imaging. MATERIALS AND METHODS: Quantitative MR imaging sequence acquisitions were analyzed in a sample of 35 extremely preterm neonates imaged at term-equivalent ages. Quantitative MR imaging metrics (fractional anisotropy; ADC [10-3mm2/s]; and T1-/T2-relaxation times [ms]) of the medulla oblongata, pontine tegmentum, midbrain, and the right/left posterior limbs of the internal capsule were determined on diffusion tensor- and multidynamic, multiecho sequence-based imaging data. ANCOVA and a paired t test were used to compare female and male infants and to detect hemispheric developmental asymmetries. RESULTS: Seventeen female (mean gestational age at birth: 26 + 0 [SD, 1 + 4] weeks+days) and 18 male (mean gestational age at birth: 26 + 1 [SD, 1 + 3] weeks+days) infants were enrolled in this study. Significant differences were observed in the T2-relaxation time (P = .014) of the pontine tegmentum, T1-relaxation time (P = .011)/T2-relaxation time (P = .024) of the midbrain, and T1-relaxation time (P = .032) of the left posterior limb of the internal capsule. In both sexes, fractional anisotropy (P [♀] < .001/P [♂] < .001) and ADC (P [♀] = .017/P [♂] = .028) differed significantly between the right and left posterior limbs of the internal capsule. CONCLUSIONS: The combined use of various quantitative MR imaging metrics detects sex-related and interhemispheric differences of WM maturity. The brainstem and the left posterior limb of the internal capsule of male preterm neonates are more immature compared with those of female infants at term-equivalent ages. Sex differences in WM maturation need further attention for the personalization of neonatal brain imaging.

Mapping Human Fetal Brain Maturation In Vivo Using Quantitative MRI.

BACKGROUND AND PURPOSE: On the basis of a single multidynamic multiecho sequence acquisition, SyMRI generates a variety of quantitative image data that can characterize tissue-specific properties. The aim of this retrospective study was to evaluate the feasibility of SyMRI for the qualitative and quantitative assessment of fetal brain maturation. MATERIALS AND METHODS: In 52 fetuses, multidynamic multiecho sequence acquisitions were available. SyMRI was used to perform multidynamic multiecho-based postprocessing. Fetal brain maturity was scored qualitatively on the basis of SyMRI-generated MR imaging data. The results were compared with conventionally acquired T1-weighted/T2-weighted contrasts as a standard of reference. Myelin-related changes in T1-/T2-relaxation time/relaxation rate, proton density, and MR imaging signal intensity of the developing fetal brain stem were measured. A Pearson correlation analysis was used to detect correlations between the following: 1) the gestational age at MR imaging and the fetal brain maturity score, and 2) the gestational age at MR imaging and the quantitative measurements. RESULTS: SyMRI provided images of sufficient quality in 12/52 (23.08%) (range, 23 + 6-34 + 0) fetal multidynamic multiecho sequence acquisitions. The fetal brain maturity score positively correlated with gestational age at MR imaging (SyMRI: r = 0.915, P < .001/standard of reference: r = 0.966, P < .001). Myelination-related changes in the T2 relaxation time/T2 relaxation rate of the medulla oblongata significantly correlated with gestational age at MR imaging (T2-relaxation time: r = -0.739, P = .006/T2-relaxation rate: r = 0.790, P = .002). CONCLUSIONS: Fetal motion limits the applicability of multidynamic multiecho-based postprocessing. However, SyMRI-generated image data of sufficient quality enable the qualitative assessment of maturity-related changes of the fetal brain. In addition, quantitative T2 relaxation time/T2 relaxation rate mapping characterizes myelin-related changes of the brain stem prenatally. This approach, if successful, opens novel possibilities for the evaluation of structural and biochemical aspects of fetal brain maturation.

Comparison of the colon with T1 breath-hold vs T1 free-breathing-A retrospective fetal MRI study.

OBJECTIVES: Fetal magnetic resonance imaging (MRI) plays an increasingly important role in the prenatal diagnosis of gastrointestinal abnormalities. During gestation, the bowel develops T1-weighted hyperintensity due to meconium formation. Currently used T1-weighted sequences are performed in maternal breath-hold (BH) technique, which may take up to 20 s. The free-breathing (FB) T1-weighted 3D radial VIBE (volumetric interpolated breath-hold examination) sequence requires no breath-hold, improving patient comfort. This study aimed to address how well the FB acquisition technique can visualize large bowel structures compared to the routinely performed breath-hold sequence. METHODS: Forty-seven fetal MRI studies between 21 and 36 weeks of gestation without abdominal pathologies on prenatal MRI and ultrasound were included. All fetal scans were performed using a Philips Ingenia 1.5 T MRI. Coronal T1-weighted BH and FB sequences without fat suppression were compared. The following acquisition parameters were used (T1, FB): resolution 1.137 mm, 1.004 mm; matrix size 288 × 288, 448 × 448; FOV 328 mm, 450 mm; TR 81-132 ms, 3.47 ms; TE 4.6 ms, 1.47 ms. Due to the necessity of the breath-hold the duration of the sequence could not exceed 20 s (mean duration of the T1-weighted BH sequence 15.17 s, and mean duration of the FB sequence 26.42 s). In all examined fetuses the following structures were evaluated with respect to their visibility (0-not visible, 1-partially visible, 2-clearly visible): rectum, sigmoid, descending, transverse and ascending colon, cecum. Furthermore, motion artifacts were assessed (0-none, 1-intermediate, 2-severe motion artifacts), and the signal intensity (SI) ratio between maternal fat and fetal rectum SI was calculated. RESULTS: No significant differences in the visibility of sigmoid and colon between BH and FB were detected, only the cecum could be seen slightly better (in 29.8 % of cases) using BH technique. Motion artifacts were similar between BH and FB. There was a non-significant SI difference (p = 0.68) in the rectum, with a higher SI in the BH sequence. CONCLUSIONS: The FB acquisition technique compared to T1 using BH is equal regarding visibility of bowel structures and artifacts. Due to non-inferiority to the BH technique, the FB sequence is a good alternative in cases where BH cannot be performed. As the FB sequence further allows for thinner slices with a good signal, even small bowel loops may be visualized.

[Artificial intelligence in lung imaging]. / Künstliche Intelligenz in der Bildgebung der Lunge.

CLINICAL/METHODICAL ISSUE: Artificial intelligence (AI) has the potential to improve diagnostic accuracy and management in patients with lung disease through automated detection, quantification, classification, and prediction of disease progression. STANDARD RADIOLOGICAL METHODS: Owing to unspecific symptoms, few well-defined CT disease patterns, and varying prognosis, interstitial lungs disease represents a focus of AI-based research. METHODICAL INNOVATIONS: Supervised and unsupervised machine learning can identify CT disease patterns using features which may allow the analysis of associations with specific diseases and outcomes. PERFORMANCE: Machine learning on the one hand improves computer-aided detection of pulmonary nodules. On the other hand it enables further characterization of pulmonary nodules, which may improve resource effectiveness regarding lung cancer screening programs. ACHIEVEMENTS: There are several challenges regarding AI-based CT data analysis. Besides the need for powerful algorithms, expert annotations and extensive training data sets that reflect physiologic and pathologic variability are required for effective machine learning. Comparability and reproducibility of AI research deserve consideration due to a lack of standardization in this emerging field. PRACTICAL RECOMMENDATIONS: This review article presents the state of the art and the challenges concerning AI in lung imaging with special consideration of interstitial lung disease, and detection and consideration of pulmonary nodules.

Three-dimensional reconstruction of defects in congenital diaphragmatic hernia: a fetal MRI study.

OBJECTIVE: To assess the clinical feasibility and validity of fetal magnetic resonance imaging (MRI)-based three-dimensional (3D) reconstruction to locate, classify and quantify diaphragmatic defects in congenital diaphragmatic hernia (CDH). METHODS: This retrospective study included 46 cases of CDH which underwent a total of 69 fetal MRI scans (65 in-vivo and four postmortem) at the Medical University of Vienna during the period 1 January 2002 to 1 January 2017. Scans were performed between 16 and 38 gestational weeks using steady-state free precession, T2-weighted and T1-weighted sequences. MRI data were retrieved from the hospital database and manual segmentation of the diaphragm was performed with the open-source software, ITK-SNAP. The resulting 3D models of the fetal diaphragm and its defect(s) were validated by postmortem MRI segmentation and/or comparison of 3D model-based classification of the defect with a reference classification based on autopsy and/or surgery reports. Surface areas of the intact diaphragm and of the defect were measured and used to calculate defect-diaphragmatic ratios (DDR). The need for prosthetic patch repair and, in cases with repeated in-vivo fetal MRI scans, diaphragm growth dynamics, were analyzed based on DDR. RESULTS: Fetal MRI-based manual segmentation of the diaphragm in CDH was feasible for all 65 (100%) of the in-vivo fetal MRI scans. Based on the 3D diaphragmatic models, one bilateral and 45 unilateral defects (n = 47) were further classified as posterolateral (23/47, 48.9%), lateral (7/47, 14.9%) or hemidiaphragmatic (17/47, 36.2%) defects, and none (0%) was classified as anterolateral. This classification of defect location was correct in all 37 (100%) of the cases in which this information could be verified. Nineteen cases had a follow-up fetal MRI scan; in five (26.3%) of these, the initial CDH classification was altered by the results of the second scan. Thirty-three fetuses underwent postnatal diaphragmatic surgical repair; 20 fetuses (all of those with DDR ≥ 54 and 88% of those with DDR > 30) received a diaphragmatic patch, while the other 13 underwent primary surgical repair. Individual DDRs at initial and at follow-up in-vivo fetal MRI correlated significantly (P < 0.001). CONCLUSIONS: MRI-based 3D reconstruction of the fetal diaphragm in CDH has been validated to visualize, locate, classify and quantify the defect. Planning of postnatal surgery may be optimized by MRI-based prediction of the necessity for patch placement and the ability to personalize patch design based on 3D-printable templates. Copyright © 2019 ISUOG. Published by John Wiley & Sons Ltd.

Machine learning: from radiomics to discovery and routine.

Machine learning is rapidly gaining importance in radiology. It allows for the exploitation of patterns in imaging data and in patient records for a more accurate and precise quantification, diagnosis, and prognosis. Here, we outline the basics of machine learning relevant for radiology, and review the current state of the art, the limitations, and the challenges faced as these techniques become an important building block of precision medicine. Furthermore, we discuss the roles machine learning can play in clinical routine and research and predict how it might change the field of radiology.