Deep learning-based segmentation of whole-body fetal MRI and fetal weight estimation: assessing performance, repeatability, and reproducibility.

OBJECTIVES: To develop a deep-learning method for whole-body fetal segmentation based on MRI; to assess the method's repeatability, reproducibility, and accuracy; to create an MRI-based normal fetal weight growth chart; and to assess the sensitivity to detect fetuses with growth restriction (FGR). METHODS: Retrospective data of 348 fetuses with gestational age (GA) of 19-39 weeks were included: 249 normal appropriate for GA (AGA), 19 FGR, and 80 Other (having various imaging abnormalities). A fetal whole-body segmentation model with a quality estimation module was developed and evaluated in 169 cases. The method was evaluated for its repeatability (repeated scans within the same scanner, n = 22), reproducibility (different scanners, n = 6), and accuracy (compared with birth weight, n = 7). A normal MRI-based growth chart was derived. RESULTS: The method achieved a Dice = 0.973, absolute volume difference ratio (VDR) = 1.8% and VDR mean difference = 0.75% ([Formula: see text]: - 3.95%, 5.46), and high agreement with the gold standard. The method achieved a repeatability coefficient = 4.01%, ICC = 0.99, high reproducibility with a mean difference = 2.21% ([Formula: see text]: - 1.92%, 6.35%), and high accuracy with a mean difference between estimated fetal weight (EFW) and birth weight of - 0.39% ([Formula: see text]: - 8.23%, 7.45%). A normal growth chart (n = 246) was consistent with four ultrasound charts. EFW based on MRI correctly predicted birth-weight percentiles for all 18 fetuses ≤ 10thpercentile and for 14 out of 17 FGR fetuses below the 3rd percentile. Six fetuses referred to MRI as AGA were found to be < 3rd percentile. CONCLUSIONS: The proposed method for automatic MRI-based EFW demonstrated high performance and sensitivity to identify FGR fetuses. CLINICAL RELEVANCE STATEMENT: Results from this study support the use of the automatic fetal weight estimation method based on MRI for the assessment of fetal development and to detect fetuses at risk for growth restriction. KEY POINTS: • An AI-based segmentation method with a quality assessment module for fetal weight estimation based on MRI was developed, achieving high repeatability, reproducibility, and accuracy. • An MRI-based fetal weight growth chart constructed from a large cohort of normal and appropriate gestational-age fetuses is proposed. • The method showed a high sensitivity for the diagnosis of small fetuses suspected of growth restriction.

Automatic Quantification of Normal Brain Gyrification Patterns and Changes in Fetuses with Polymicrogyria and Lissencephaly Based on MRI.

BACKGROUND AND PURPOSE: The current imaging assessment of fetal brain gyrification is performed qualitatively and subjectively using sonography and MR imaging. A few previous studies have suggested methods for quantification of fetal gyrification based on 3D reconstructed MR imaging, which requires unique data and is time-consuming. In this study, we aimed to develop an automatic pipeline for gyrification assessment based on routinely acquired fetal 2D MR imaging data, to quantify normal changes with gestation, and to measure differences in fetuses with lissencephaly and polymicrogyria compared with controls. MATERIALS AND METHODS: We included coronal T2-weighted MR imaging data of 162 fetuses retrospectively collected from 2 clinical sites: 134 controls, 12 with lissencephaly, 13 with polymicrogyria, and 3 with suspected lissencephaly based on sonography, yet with normal MR imaging diagnoses. Following brain segmentation, 5 gyrification parameters were calculated separately for each hemisphere on the basis of the area and ratio between the contours of the cerebrum and its convex hull. Seven machine learning classifiers were evaluated to differentiate control fetuses and fetuses with lissencephaly or polymicrogyria. RESULTS: In control fetuses, all parameters changed significantly with gestational age (P < .05). Compared with controls, fetuses with lissencephaly showed significant reductions in all gyrification parameters (P ≤ .02). Similarly, significant reductions were detected for fetuses with polymicrogyria in several parameters (P ≤ .001). The 3 suspected fetuses showed normal gyrification values, supporting the MR imaging diagnosis. An XGBoost-linear algorithm achieved the best results for classification between fetuses with lissencephaly and control fetuses (n = 32), with an area under the curve of 0.90 and a recall of 0.83. Similarly, a random forest classifier showed the best performance for classification of fetuses with polymicrogyria and control fetuses (n = 33), with an area under the curve of 0.84 and a recall of 0.62. CONCLUSIONS: This study presents a pipeline for automatic quantification of fetal brain gyrification and provides normal developmental curves from a large cohort. Our method significantly differentiated fetuses with lissencephaly and polymicrogyria, demonstrating lower gyrification values. The method can aid radiologic assessment, highlight fetuses at risk, and may improve early identification of fetuses with cortical malformations.

Fetal MRI-Based Body and Adiposity Quantification for Small for Gestational Age Perinatal Risk Stratification.

BACKGROUND: Small for gestational age (SGA) fetuses are at risk for perinatal adverse outcomes. Fetal body composition reflects the fetal nutrition status and hold promise as potential prognostic indicator. MRI quantification of fetal anthropometrics may enhance SGA risk stratification. HYPOTHESIS: Smaller, leaner fetuses are malnourished and will experience unfavorable outcomes. STUDY TYPE: Prospective. POPULATION: 40 SGA fetuses, 26 (61.9%) females: 10/40 (25%) had obstetric interventions due to non-reassuring fetal status (NRFS), and 17/40 (42.5%) experienced adverse neonatal events (CANO). Participants underwent MRI between gestational ages 30 + 2 and 37 + 2. FIELD STRENGTH/SEQUENCE: 3-T, True Fast Imaging with Steady State Free Precession (TruFISP) and T1 -weighted two-point Dixon (T1 W Dixon) sequences. ASSESSMENT: Total body volume (TBV), fat signal fraction (FSF), and the fat-to-body volumes ratio (FBVR) were extracted from TruFISP and T1 W Dixon images, and computed from automatic fetal body and subcutaneous fat segmentations by deep learning. Subjects were followed until hospital discharge, and obstetric interventions and neonatal adverse events were recorded. STATISTICAL TESTS: Univariate and multivariate logistic regressions for the association between TBV, FBVR, and FSF and interventions for NRFS and CANO. Fisher's exact test was used to measure the association between sonographic FGR criteria and perinatal outcomes. Sensitivity, specificity, positive and negative predictive values, and accuracy were calculated. A P-value <0.05 was considered statistically significant. RESULTS: FBVR (odds ratio [OR] 0.39, 95% confidence interval [CI] 0.2-0.76) and FSF (OR 0.95, CI 0.91-0.99) were linked with NRFS interventions. Furthermore, TBV (OR 0.69, CI 0.56-0.86) and FSF (OR 0.96, CI 0.93-0.99) were linked to CANO. The FBVR sensitivity/specificity for obstetric interventions was 85.7%/87.5%, and the TBV sensitivity/specificity for CANO was 82.35%/86.4%. The sonographic criteria sensitivity/specificity for obstetric interventions was 100%/33.3% and insignificant for CANO (P = 0.145). DATA CONCLUSION: Reduced TBV and FBVR may be associated with higher rates of obstetric interventions for NRFS and CANO. EVIDENCE LEVEL: 2 TECHNICAL EFFICACY: Stage 5.

Fetal brain tissue annotation and segmentation challenge results.

In-utero fetal MRI is emerging as an important tool in the diagnosis and analysis of the developing human brain. Automatic segmentation of the developing fetal brain is a vital step in the quantitative analysis of prenatal neurodevelopment both in the research and clinical context. However, manual segmentation of cerebral structures is time-consuming and prone to error and inter-observer variability. Therefore, we organized the Fetal Tissue Annotation (FeTA) Challenge in 2021 in order to encourage the development of automatic segmentation algorithms on an international level. The challenge utilized FeTA Dataset, an open dataset of fetal brain MRI reconstructions segmented into seven different tissues (external cerebrospinal fluid, gray matter, white matter, ventricles, cerebellum, brainstem, deep gray matter). 20 international teams participated in this challenge, submitting a total of 21 algorithms for evaluation. In this paper, we provide a detailed analysis of the results from both a technical and clinical perspective. All participants relied on deep learning methods, mainly U-Nets, with some variability present in the network architecture, optimization, and image pre- and post-processing. The majority of teams used existing medical imaging deep learning frameworks. The main differences between the submissions were the fine tuning done during training, and the specific pre- and post-processing steps performed. The challenge results showed that almost all submissions performed similarly. Four of the top five teams used ensemble learning methods. However, one team's algorithm performed significantly superior to the other submissions, and consisted of an asymmetrical U-Net network architecture. This paper provides a first of its kind benchmark for future automatic multi-tissue segmentation algorithms for the developing human brain in utero.

Reduced adipose tissue in growth-restricted fetuses using quantitative analysis of magnetic resonance images.

OBJECTIVES: Fat-water MRI can be used to quantify tissues' lipid content. We aimed to quantify fetal third trimester normal whole-body subcutaneous lipid deposition and explore differences between appropriate for gestational age (AGA), fetal growth restriction (FGR), and small for gestational age fetuses (SGAs). METHODS: We prospectively recruited women with FGR and SGA-complicated pregnancies and retrospectively recruited the AGA cohort (sonographic estimated fetal weight [EFW] ≥ 10th centile). FGR was defined using the accepted Delphi criteria, and fetuses with an EFW < 10th centile that did not meet the Delphi criteria were defined as SGA. Fat-water and anatomical images were acquired in 3 T MRI scanners. The entire fetal subcutaneous fat was semi-automatically segmented. Three adiposity parameters were calculated: fat signal fraction (FSF) and two novel parameters, i.e., fat-to-body volume ratio (FBVR) and estimated total lipid content (ETLC = FSF*FBVR). Normal lipid deposition with gestation and differences between groups were assessed. RESULTS: Thirty-seven AGA, 18 FGR, and 9 SGA pregnancies were included. All three adiposity parameters increased between 30 and 39 weeks (p < 0.001). All three adiposity parameters were significantly lower in FGR compared with AGA (p ≤ 0.001). Only ETLC and FSF were significantly lower in SGA compared with AGA using regression analysis (p = 0.018-0.036, respectively). Compared with SGA, FGR had a significantly lower FBVR (p = 0.011) with no significant differences in FSF and ETLC (p ≥ 0.053). CONCLUSIONS: Whole-body subcutaneous lipid accretion increased throughout the third trimester. Reduced lipid deposition is predominant in FGR and may be used to differentiate FGR from SGA, assess FGR severity, and study other malnourishment pathologies. CLINICAL RELEVANCE STATEMENT: Fetuses with growth restriction have reduced lipid deposition than appropriately developing fetuses measured using MRI. Reduced fat accretion is linked with worse outcomes and may be used for growth restriction risk stratification. KEY POINTS: • Fat-water MRI can be used to assess the fetal nutritional status quantitatively. • Lipid deposition increased throughout the third trimester in AGA fetuses. • FGR and SGA have reduced lipid deposition compared with AGA fetuses, more predominant in FGR.

Improved differentiation between hypo/hypertelorism and normal fetuses based on MRI using automatic ocular biometric measurements, ocular ratios, and machine learning multi-parametric classification.

OBJECTIVES: To differentiate hypo-/hypertelorism (abnormal) from normal fetuses using automatic biometric measurements and machine learning (ML) classification based on MRI. METHODS: MRI data of normal (n = 244) and abnormal (n = 52) fetuses of 22-40 weeks' gestational age (GA), scanned between March 2008 and June 2020 on 1.5/3T systems with various T2-weighted sequences and image resolutions, were included. A fully automatic method including deep learning and geometric algorithms was developed to measure the binocular (BOD), inter-ocular (IOD), ocular (OD) diameters, and ocular volume (OV). Two new parameters, BOD-ratio and IOD-ratio, were defined as the ratio between BOD/IOD relative to the sum of both globes' OD, respectively. Eight ML classifiers were evaluated to detect abnormalities using measured and computed parameters. RESULTS: The automatic method yielded a mean difference of BOD = 0.70 mm, IOD = 0.81 mm, OD = 1.00 mm, and a 3D-Dice score of OV = 93.7%. In normal fetuses, all four measurements increased with GA. Constant values were detected for BOD-ratio = 1.56 ± 0.05 and IOD-ratio = 0.60 ± 0.05 across all GA and when calculated from previously published reference data of both MRI and ultrasound. A random forest classifier yielded the best results on an independent test set (n = 58): AUC-ROC = 0.941 and F1-Score = 0.711 in comparison to AUC-ROC = 0.650 and F1-Score = 0.385 achieved based on the accepted criteria that define hypo/hypertelorism based on IOD (< 5th or > 95th percentiles). Using the explainable ML method, the two computed ratios were found as the most contributing parameters. CONCLUSIONS: The developed fully automatic method demonstrates high performance on varied clinical imaging data. The new BOD and IOD ratios and ML multi-parametric classifier are suggested to improve the differentiation of hypo-/hypertelorism from normal fetuses. KEY POINTS: • A fully automatic method for computing fetal ocular biometry from MRI is proposed, achieving high performance, comparable to that of an expert fetal neuro-radiologist. • Two new parameters, IOD-ratio and BOD-ratio, are proposed for routine clinical use in ultrasound and MRI. These two ratios are constant across gestational age in normal fetuses, consistent across studies, and differentiate between fetuses with and without hypo/hypertelorism. • Multi-parametric machine learning classification based on automatic measurements and the two new ratios improves the identification of fetal ocular anomalies beyond the accepted criteria (<5th or >95th IOD percentiles).

The rediscovered motor-related area 55b emerges as a core hub of music perception.

Passive listening to music, without sound production or evident movement, is long known to activate motor control regions. Nevertheless, the exact neuroanatomical correlates of the auditory-motor association and its underlying neural mechanisms have not been fully determined. Here, based on a NeuroSynth meta-analysis and three original fMRI paradigms of music perception, we show that the long-ignored pre-motor region, area 55b, an anatomically unique and functionally intriguing region, is a core hub of music perception. Moreover, results of a brain-behavior correlation analysis implicate neural entrainment as the underlying mechanism of area 55b's contribution to music perception. In view of the current results and prior literature, area 55b is proposed as a keystone of sensorimotor integration, a fundamental brain machinery underlying simple to hierarchically complex behaviors. Refining the neuroanatomical and physiological understanding of sensorimotor integration is expected to have a major impact on various fields, from brain disorders to artificial general intelligence.

Predicting EGFR mutation status by a deep learning approach in patients with non-small cell lung cancer brain metastases.

PURPOSE: Non-small cell lung cancer (NSCLC) tends to metastasize to the brain. Between 10 and 60% of NSCLCs harbor an activating mutation in the epidermal growth-factor receptor (EGFR), which may be targeted with selective EGFR inhibitors. However, due to a high discordance rate between the molecular profile of the primary tumor and the brain metastases (BMs), identifying an individual patient's EGFR status of the BMs necessitates tissue diagnosis via an invasive surgical procedure. We employed a deep learning (DL) method with the aim of noninvasive detection of the EGFR mutation status in NSCLC BM. METHODS: We retrospectively collected clinical, radiological, and pathological-molecular data of all the NSCLC patients who had been diagnosed with BMs and underwent resection of their BM during 2009-2019. The study population was then divided into two groups based upon EGFR mutational status. We further employed a DL technique to classify the two groups according to their preoperative magnetic resonance imaging features. Augmentation techniques, transfer learning approach, and post-processing of the predicted results were applied to overcome the relatively small cohort. Finally, we established the accuracy of our model in predicting EGFR mutation status of BM of NSCLC. RESULTS: Fifty-nine patients were included in the study, 16 patients harbored EGFR mutations. Our model predicted mutational status with mean accuracy of 89.8%, sensitivity of 68.7%, specificity of 97.7%, and a receiver operating characteristic curve value of 0.91 across the 5 validation datasets. CONCLUSION: DL-based noninvasive molecular characterization is feasible, has high accuracy and should be further validated in large prospective cohorts.

Neural Responsivity to Reward Versus Punishment Shortly After Trauma Predicts Long-Term Development of Posttraumatic Stress Symptoms.

BACKGROUND: Processing negatively and positively valenced stimuli involves multiple brain regions including the amygdala and ventral striatum (VS). Posttraumatic stress disorder (PTSD) is often associated with hyperresponsivity to negatively valenced stimuli, yet recent evidence also points to deficient positive valence functioning. It is yet unclear what the relative contribution is of such opposing valence processing shortly after trauma to the development of chronic PTSD. METHODS: Neurobehavioral indicators of motivational positive versus negative valence sensitivities were longitudinally assessed in 171 adults (87 females, age = 34.19 ± 11.47 years) at 1, 6, and 14 months following trauma exposure (time point 1 [TP1], TP2, and TP3, respectively). Using a gambling functional magnetic resonance imaging paradigm, amygdala and VS functionality (activity and functional connectivity with the prefrontal cortex) in response to rewards versus punishments were assessed with relation to PTSD severity at different time points. The effect of valence processing was depicted behaviorally by the amount of risk taken to maximize reward. RESULTS: PTSD severity at TP1 was associated with greater neural functionality in the amygdala (but not in the VS) toward punishments versus rewards, and with fewer risky choices. PTSD severity at TP3 was associated with decreased neural functionality in both the VS and the amygdala toward rewards versus punishments at TP1 (but not with risky behavior). Explainable machine learning revealed the primacy of VS-biased processing, over the amygdala, in predicting PTSD severity at TP3. CONCLUSIONS: These results highlight the importance of biased neural responsivity to positive relative to negative motivational outcomes in PTSD development. Novel therapeutic strategies early after trauma may thus target both valence fronts.

Ex-Vivo MRI of the Normal Human Placenta: Structural-Functional Interplay and the Association With Birth Weight.

BACKGROUND: Advanced magnetic resonance imaging (MRI) methods are increasingly being used to assess the human placenta. Yet, the structure-function interplay in normal placentas and their associations with pregnancy risks are not fully understood. PURPOSE: To characterize the normal human placental structure (volume and umbilical cord centricity index (CI)) and function (perfusion) ex-vivo using MRI, to assess their association with birth weight (BW), and identify imaging-markers for placentas at risk for dysfunction. STUDY TYPE: Prospective. POPULATION: Twenty normal term ex-vivo placentas. FIELD STRENGTH/SEQUENCE: 3 T/ T1 and T2 weighted (T1 W, T2 W) turbo spin-echo, three-dimensional susceptibility-weighted image, and time-resolved angiography with interleaved stochastic trajectories (TWIST), during passage of a contrast agent using MRI compatible perfusion system that mimics placental flow. ASSESSMENT: Placental volume and CI were manually extracted from the T1 W images by a fetal-placental MRI scientist (D.L., 7 years of experience). Perfusion maps including bolus arrival-time and full-width at half maximum were calculated from the TWIST data. Mean values, entropy, and asymmetries were calculated from each perfusion map, relating to both the whole placenta and volumes of interest (VOIs) within the umbilical cord and its daughter blood vessels. STATISTICAL TESTS: Pearson correlations with correction for multiple comparisons using false discovery rate were performed between structural and functional parameters, and with BW, with P < 0.05 considered significant. RESULTS: All placentas were successfully perfused and scanned. Significant correlations were found between whole placenta and VOIs perfusion parameters (mean R = 0.76 ± 0.06, range = 0.67-0.89), which were also significantly correlated with CI (mean R = 0.72 ± 0.05, range = 0.65-0.79). BW was correlated with placental volume (R = 0.62), but not with CI (P = 0.40). BW was also correlated with local perfusion asymmetry (R = -0.71). DATA CONCLUSION: Results demonstrate a gradient of placental function, associated with CI and suggest several ex-vivo imaging-markers that might indicate an increased risk for placental dysfunction. LEVEL OF EVIDENCE: 1 TECHNICAL EFFICACY: Stage 1.

Automatic linear measurements of the fetal brain on MRI with deep neural networks.

PURPOSE: Timely, accurate and reliable assessment of fetal brain development is essential to reduce short and long-term risks to fetus and mother. Fetal MRI is increasingly used for fetal brain assessment. Three key biometric linear measurements important for fetal brain evaluation are cerebral biparietal diameter (CBD), bone biparietal diameter (BBD), and trans-cerebellum diameter (TCD), obtained manually by expert radiologists on reference slices, which is time consuming and prone to human error. The aim of this study was to develop a fully automatic method computing the CBD, BBD and TCD measurements from fetal brain MRI. METHODS: The input is fetal brain MRI volumes which may include the fetal body and the mother's abdomen. The outputs are the measurement values and reference slices on which the measurements were computed. The method, which follows the manual measurements principle, consists of five stages: (1) computation of a region of interest that includes the fetal brain with an anisotropic 3D U-Net classifier; (2) reference slice selection with a convolutional neural network; (3) slice-wise fetal brain structures segmentation with a multi-class U-Net classifier; (4) computation of the fetal brain midsagittal line and fetal brain orientation, and; (5) computation of the measurements. RESULTS: Experimental results on 214 volumes for CBD, BBD and TCD measurements yielded a mean [Formula: see text] difference of 1.55 mm, 1.45 mm and 1.23 mm, respectively, and a Bland-Altman 95% confidence interval ([Formula: see text] of 3.92 mm, 3.98 mm and 2.25 mm, respectively. These results are similar to the manual inter-observer variability, and are consistent across gestational ages and brain conditions. CONCLUSIONS: The proposed automatic method for computing biometric linear measurements of the fetal brain from MR imaging achieves human-level performance. It has the potential of being a useful method for the assessment of fetal brain biometry in normal and pathological cases, and of improving routine clinical practice.