Development and Validation of Deep Learning-Based Automated Detection of Cervical Lymphadenopathy in Patients with Lymphoma for Treatment Response Assessment: A Bi-institutional Feasibility Study.

The purpose is to train and evaluate a deep learning (DL) model for the accurate detection and segmentation of abnormal cervical lymph nodes (LN) on head and neck contrast-enhanced CT scans in patients diagnosed with lymphoma and evaluate the clinical utility of the DL model in response assessment. This retrospective study included patients who underwent CT for abnormal cervical LN and lymphoma assessment between January 2021 and July 2022. Patients were grouped into the development (n = 76), internal test 1 (n = 27), internal test 2 (n = 87), and external test (n = 26) cohorts. A 3D SegResNet model was used to train the CT images. The volume change rates of cervical LN across longitudinal CT scans were compared among patients with different treatment outcomes (stable, response, and progression). Dice similarity coefficient (DSC) and the Bland-Altman plot were used to assess the model's segmentation performance and reliability, respectively. No significant differences in baseline clinical characteristics were found across cohorts (age, P = 0.55; sex, P = 0.13; diagnoses, P = 0.06). The mean DSC was 0.39 ± 0.2 with a precision and recall of 60.9% and 57.0%, respectively. Most LN volumes were within the limits of agreement on the Bland-Altman plot. The volume change rates among the three groups differed significantly (progression (n = 74), 342.2%; response (n = 8), - 79.2%; stable (n = 5), - 8.1%; all P < 0.01). Our proposed DL segmentation model showed modest performance in quantifying the cervical LN burden on CT in patients with lymphoma. Longitudinal changes in cervical LN volume, as predicted by the DL model, were useful for treatment response assessment.

Diagnostic Performance Using a Combination of MRI Findings for Evaluating Cognitive Decline.

Purpose: We investigated potentially promising imaging findings and their combinations in the evaluation of cognitive decline. Materials and Methods: This retrospective study included 138 patients with subjective cognitive impairments, who underwent brain MRI. We classified the same group of patients into Alzheimer's disease (AD) and non-AD groups, based on the neuropsychiatric evaluation. We analyzed imaging findings, including white matter hyperintensity (WMH) and cerebral microbleeds (CMBs), using the Kruskal-Wallis test for group comparison, and receiver operating characteristic (ROC) curve analysis for assessing the diagnostic performance of imaging findings. Results: CMBs in the lobar or deep locations demonstrated higher prevalence in the patients with AD compared to those in the non-AD group. The presence of lobar CMBs combined with periventricular WMH (area under the ROC curve [AUC] = 0.702 [95% confidence interval: 0.599-0.806], p < 0.001) showed the highest performance in differentiation of AD from non-AD group. Conclusion: Combinations of imaging findings can serve as useful additive diagnostic tools in the assessment of cognitive decline.

Cerebrovascular Malformation Mimicking Recurrent Lymphoma on Dual Time-Point 18F-FDOPA PET.

ABSTRACT: Although 18F-FDG is the dominant radiotracer for PET imaging of hematological malignancies, radiolabeled amino acids have also been investigated to improve image quality in areas of high 18F-FDG uptake such as the central nervous system. We present a case of a 57-year-old woman who underwent an 18F-FDOPA scan for primary CNS lymphoma, which demonstrated an unexpected false-positive uptake in the right frontal lobe, due to a developmental venous anomaly.

Comparison of the Therapeutic Efficacy and Technical Outcomes between Conventional Fixed Electrodes and Adjustable Electrodes in the Radiofrequency Ablation of Benign Thyroid Nodules.

OBJECTIVE: This study aimed to compare therapeutic efficacy and technical outcomes between adjustable electrode (AE) and conventional fixed electrode (FE) for radiofrequency ablation (RFA) of benign thyroid nodules. MATERIALS AND METHODS: Between 2013 and 2021, RFA was performed on histologically proven benign thyroid nodules. For the AE method, AE length ≥ 1 cm with higher power and < 1 cm with lower power were utilized for ablating feeding vessels and nodules, especially those near anatomical structures, respectively. The therapeutic efficacy (volume reduction rate [VRR], complication rate, and regrowth rate) and technical outcomes (total energy delivery, ablated volume/energy, RFA time, and ablated volume/time) of FE and AE were compared. Continuous parameters were compared using a two-sample t-test or Mann-Whitney U test, and categorical parameters were compared using a chi-squared test or Fisher's exact test. RESULTS: A total of 182 nodules (FE: 92 vs. AE: 90) in 173 patients (mean age ± standard deviation, 47.0 ± 14.7 years; female, 90.8% [157/173]; median follow-up, 726 days [interquartile range, 441-1075 days]) were analyzed. The therapeutic efficacy was comparable, whereas technical outcomes were more favorable for AE. Both electrodes demonstrated comparable overall median VRR (FE: 92.4% vs. AE: 84.9%, P = 0.240) without immediate major complications. Overall regrowth rates were comparable between the two groups (FE: 2.2% [2/90] vs. AE: 1.1% [1/90], P > 0.99). AE demonstrated a shorter median RFA time (FE: 811 vs. AE: 627 seconds, P = 0.009). Both delivered comparable median energy (FE: 42.8 vs. AE: 29.2 kJ, P = 0.069), but AE demonstrated higher median ablated volume/energy and median ablated volume/time (FE: 0.2 vs. AE: 0.3 cc/kJ, P < 0.001; and FE: 0.7 vs. AE: 1.0 cc/min, P < 0.001, respectively). CONCLUSION: Therapeutic efficacy between FE and AE was comparable. AE demonstrated better technical outcomes than FE in terms of RFA time, ablated volume/energy, and ablated volume/time.

A highly efficient strategy for simultaneous analysis of eleven monomeric hydroxylated-stilbenes in grape wines and juices by a salting-out assisted liquid-liquid (SALLE) extraction combined with a high-performance liquid chromatography-tandem mass spectrometry (HPLC-MS/MS).

Wine and grape juice are known to have hydroxylated stilbenes, a group of phytoalexins, health-promoting compounds. The determination of stilbene species in wine and grape juices remains a challenging task. Here, we propose an efficient strategy for the simultaneous determination of eleven hydroxylated stilbenes in grape wines and grape juice by a SALLE in conjunction with an HPLC-ESI-MS/MS. SALLE and HPLC-MS/MS conditions were optimized. The proposed method was validated and found to be a simple, sensitive, and reliable measure for the determination of the stilbene species in wine and grape juice. The SALLE was fast and efficient, taking only 10 min. The HPLC-MS/MS was able to quickly separate and quantitate the eleven stilbenes (9 min running). This method was successfully applied to determine the contents of stilbene species in commercial wine and grape juice in Korea. This method is advantageous in reduced sample preparation time, low labor, and cost efficiency.

Diagnostic Performance of Ultrasound-Based Risk Stratification Systems for Thyroid Nodules: A Systematic Review and Meta-Analysis.

BACKGRUOUND: This study investigated the diagnostic performance of biopsy criteria in four society ultrasonography risk stratification systems (RSSs) for thyroid nodules, including the 2021 Korean (K)-Thyroid Imaging Reporting and Data System (TIRADS). METHODS: The Ovid-MEDLINE, Embase, Cochrane, and KoreaMed databases were searched and a manual search was conducted to identify original articles investigating the diagnostic performance of biopsy criteria for thyroid nodules (≥1 cm) in four widely used society RSSs. RESULTS: Eleven articles were included. The pooled sensitivity and specificity were 82% (95% confidence interval [CI], 74% to 87%) and 60% (95% CI, 52% to 67%) for the American College of Radiology (ACR)-TIRADS, 89% (95% CI, 85% to 93%) and 34% (95% CI, 26% to 42%) for the American Thyroid Association (ATA) system, 88% (95% CI, 81% to 92%) and 42% (95% CI, 22% to 67%) for the European (EU)-TIRADS, and 96% (95% CI, 94% to 97%) and 21% (95% CI, 17% to 25%) for the 2016 K-TIRADS. The sensitivity and specificity were 76% (95% CI, 74% to 79%) and 50% (95% CI, 49% to 52%) for the 2021 K-TIRADS1.5 (1.5-cm size cut-off for intermediate-suspicion nodules). The pooled unnecessary biopsy rates of the ACR-TIRADS, ATA system, EU-TIRADS, and 2016 K-TIRADS were 41% (95% CI, 32% to 49%), 65% (95% CI, 56% to 74%), 68% (95% CI, 60% to 75%), and 79% (95% CI, 74% to 83%), respectively. The unnecessary biopsy rate was 50% (95% CI, 47% to 53%) for the 2021 K-TIRADS1.5. CONCLUSION: The unnecessary biopsy rate of the 2021 K-TIRADS1.5 was substantially lower than that of the 2016 K-TIRADS and comparable to that of the ACR-TIRADS. The 2021 K-TIRADS may help reduce potential harm due to unnecessary biopsies.

Standardized Imaging and Reporting for Thyroid Ultrasound: Korean Society of Thyroid Radiology Consensus Statement and Recommendation.

Ultrasonography (US) is a primary imaging modality for diagnosing nodular thyroid disease and has an essential role in identifying the most appropriate management strategy for patients with nodular thyroid disease. Standardized imaging techniques and reporting formats for thyroid US are necessary. For this purpose, the Korean Society of Thyroid Radiology (KSThR) organized a task force in June 2021 and developed recommendations for standardized imaging technique and reporting format, based on the 2021 KSThR consensus statement and recommendations for US-based diagnosis and management of thyroid nodules. The goal was to achieve an expert consensus applicable to clinical practice.

Correlation between radiologic depth of invasion and pathologic depth of invasion in oral cavity squamous cell carcinoma: A systematic review and meta-analysis.

OBJECTIVES: To comprehensively assess the correlation between radiologic depth of invasion (rDOI) and pathologic depth of invasion (pDOI) in oral cavity squamous cell carcinoma (OSCC) by meta-analysis. MATERIALS AND METHODS: PubMed and Embase databases were searched to find pertinent articles reporting rDOI of OSCC. Studies evaluating the correlations and mean differences (MDs) between rDOI and pDOI were included. The rDOI was measured based on ultrasound (US), computed tomography (CT), or magnetic resonance imaging (MRI). The correlation coefficients and MDs between rDOI and pDOI were meta-analytically pooled. Between-study heterogeneity was assessed using Higgins' inconsistency index (I2). Subgroup analysis was performed based on imaging modality. RESULTS: Twenty-three studies with 1787 patients were included. The pooled correlation coefficient and MD were 0.86 (95 % confidence interval [CI], 0.82-0.90; I2 = 66.9 %) and 1.84 mm (95 % CI, 1.02-2.65 mm; I2 = 88.2 %), respectively. In subgroup analysis, MRI showed the largest MD (n = 12, 2.61 mm), followed by US (n = 2, -0.41 mm) and CT (n = 2, 0.12 mm). US showed the highest correlation coefficient (n = 3, 0.91), followed by MRI (n = 12, 0.85) and CT (n = 3, 0.82). CONCLUSION: rDOI measured by US, CT, and MRI demonstrated excellent correlations with pDOI.

Diagnosis of nasal bone fractures on plain radiographs via convolutional neural networks.

This study aimed to assess the performance of deep learning (DL) algorithms in the diagnosis of nasal bone fractures on radiographs and compare it with that of experienced radiologists. In this retrospective study, 6713 patients whose nasal radiographs were examined for suspected nasal bone fractures between January 2009 and October 2020 were assessed. Our dataset was randomly split into training (n = 4325), validation (n = 481), and internal test (n = 1250) sets; a separate external dataset (n = 102) was used. The area under the receiver operating characteristic curve (AUC), sensitivity, and specificity of the DL algorithm and the two radiologists were compared. The AUCs of the DL algorithm for the internal and external test sets were 0.85 (95% CI, 0.83-0.86) and 0.86 (95% CI, 0.78-0.93), respectively, and those of the two radiologists for the external test set were 0.80 (95% CI, 0.73-0.87) and 0.75 (95% CI, 0.68-0.82). The DL algorithm therefore significantly exceeded radiologist 2 (P = 0.021) but did not significantly differ from radiologist 1 (P = 0.142). The sensitivity and specificity of the DL algorithm were 83.1% (95% CI, 71.2-93.2%) and 83.7% (95% CI, 69.8-93.0%), respectively. Our DL algorithm performs comparably to experienced radiologists in diagnosing nasal bone fractures on radiographs.

MRI-Based Demonstration of the Normal Glymphatic System in a Human Population: A Systematic Review.

Background: The glymphatic system has been described as one that facilitates the exchange between the cerebrospinal fluid (CSF) and interstitial fluid, and many recent studies have demonstrated glymphatic flow based on magnetic resonance imaging (MRI). We aim to systematically review the studies demonstrating a normal glymphatic flow in a human population using MRI and to propose a detailed glymphatic imaging protocol. Methods: We searched the MEDLINE and EMBASE databases to identify studies with human participants involving MRI-based demonstrations of the normal glymphatic flow. We extracted data on the imaging sequence, imaging protocol, and the targeted anatomical structures on each study. Results: According to contrast-enhanced MRI studies, peak enhancement was sequentially detected first in the CSF space, followed by the brain parenchyma, the meningeal lymphatic vessel (MLV), and, finally, the cervical lymph nodes, corresponding with glymphatic flow and explaining the drainage into the MLV. Non-contrast flow-sensitive MRI studies revealed similar glymphatic inflow from the CSF space to the brain parenchyma and efflux of exchanged fluid from the brain parenchyma to the MLV. Conclusion: We may recommend T1-weighted contrast-enhanced MRI for visualizing glymphatic flow. Our result can increase understanding of the glymphatic system and may lay the groundwork for establishing central nervous system fluid dynamic theories and developing standardized imaging protocols.

Deep convolutional neural network for classification of thyroid nodules on ultrasound: Comparison of the diagnostic performance with that of radiologists.

PURPOSE: This study aimed to train and validate deep learning (DL) models for differentiating malignant from benign thyroid nodules on US images and compare their performance with that of radiologists. METHODS: Images of thyroid nodules in patients who underwent US-guided fine-needle aspiration biopsy at our institution between January 2010 and March 2020 were retrospectively reviewed. Four radiologists independently classified the images. Images of thyroid nodules were trained using three different image classification DL models (VGG16, VGG19, and ResNet). The diagnostic performances of the DL models were calculated for the internal and external datasets and compared with the diagnoses of the four radiologists. Pairwise comparisons of the AUCs between the radiologists and DL models were made using bootstrap-based tests. RESULTS: In total, 15,409 images from 7,321 patients (mean age, 60 ± 13 years; malignant nodules, 20.7%) were randomly grouped into training (n = 12,327) and validation (n = 3,082) sets. Independent internal (n = 432; 197 patients) and external (n = 168; 59 patients) test sets were also acquired. The DL models demonstrated a higher diagnostic performance than the radiologists in the internal test set (AUC, 0.83 - 0.86 vs. 0.71 - 0.76, P < 0.05), but not in the external test set. The VGG16 model demonstrated the highest diagnostic performance in internal (AUC, 0.86; sensitivity, 91.8%; specificity, 73.2%) and external (AUC: 0.83; sensitivity: 78.6%; specificity: 76.8%) test sets. However, no statistical differences were found in the AUCs among the DL models. CONCLUSIONS: The DL models demonstrated comparable diagnostic performance to radiologists in distinguishing benign from malignant thyroid nodules on US images and may play a potential role in augmenting radiologists' diagnosis of thyroid nodules.

Radiofrequency ablation of recurrent thyroid cancers: anatomy-based management.

Although differentiated thyroid cancer has an excellent prognosis and low mortality, its recurrence rate has been reported to be very high. While surgery is recommended as the standard treatment for recurrent thyroid cancer, reoperation may increase the incidence of complications due to fibrosis and anatomical distortion. In patients with high surgical risk or those who refuse repeated surgery, ultrasonography-guided ablation techniques including radiofrequency ablation, ethanol ablation, and laser ablation are suggested as alternative treatments. In this manuscript, we introduce an anatomy-based management concept for radiofrequency ablation for recurrent thyroid cancer, and present a review of the relevant literature, including indications, basic techniques, clinical outcomes, and complications.

Preoperative Magnetic Resonance Image and Computerized Tomography Findings Predictive of Facial Nerve Invasion in Patients with Parotid Cancer without Preoperative Facial Weakness-A Retrospective Observational Study.

(1) Background: Facial nerve resection with reconstruction helps achieve optimal outcomes in the treatment of facial nerve invasion (FNI) of parotid cancer. Preoperative imaging is crucial to predict facial nerve reconstruction. The radiological findings of CT or MRI may predict FNI in the parotid cancer even without facial paralysis. Methods: We retrospectively reviewed the records of 151 patients without facial nerve paralysis before surgery who had undergone tumor resection. Previously untreated parotid cancers were included. (2) Results: The median follow-up duration was 62 months (range: 24-120 months). The FNI (+) group (n = 30) showed a significantly worse 5-year overall survival compared with the FNI (-) group (75.5 vs. 93.9%; hazard ratio = 4.19; 95% confidence interval: 1.74-10.08; p = 0.001). The tumor margin, tumor size, presence in the anterolateral parotid region (area 3), retromandibular vein involvement, distance from the stylomastoid foramen to the upper tumor margin, and a high tumor grade were significant factors related to FNI in the univariate analysis. A spiculated tumor margin, the tumor size (2.2 cm), and presence in area 3 were factors predicting FNI in the logistic regression model (p = 0.020, 0.005, and 0.050, respectively; odds ratio: 4.02, 6.40, and 8.16, respectively). (3) Conclusions: The tumor size (≥2.2 cm), spiculated margin, and presence in area 3 as presented in CT and MRI may help clinicians preoperatively predict FNI in patients with parotid cancer and establish an appropriate surgical plan.

Effect of olfactory bulb atrophy on the success of olfactory training.

PURPOSE: To evaluate the success of olfactory training in patients with olfactory loss and olfactory bulb (OB) atrophy detected on magnetic resonance imaging (MRI) and other characteristics. METHODS: This study included 48 patients with olfactory loss who underwent a nasal endoscopic examination and MRI before olfactory training. The Korean Version of the Sniffin' Sticks Test was performed before and after training. The olfactory training success was defined as an improvement of more than 6 points in the Threshold-Discrimination-Identification (TDI) score. Patient characteristics and OB atrophy pre-training were compared between successful and unsuccessful groups. RESULTS: The etiology of olfactory loss included respiratory viral infection in 30 (62.5%), trauma in 10 (20.8%), and idiopathic loss in 8 (16.7%) patients. Twenty-three (47.9%) of 48 patients exhibited successful olfactory training. Etiology, age, gender, and symptom duration were not different between unsuccessful and successful groups. Pre-training discrimination, identification, and TDI scores were significantly different between unsuccessful and successful groups (P < 0.05). Success rate of patients with bilateral OB atrophy was significantly lower than that of patients with unilateral OB atrophy and normal morphology (P = 0.006). OB height was significantly lower in the unsuccessful group than in the successful group (P < 0.05). Bilateral OB atrophy was an independent risk factor for failure of olfactory training according to the multivariate analysis. CONCLUSION: Olfactory loss patients with bilateral OB atrophy may not be able to improve olfactory function after olfactory training.

Diffusion-weighted MRI for predicting treatment response in patients with nasopharyngeal carcinoma: a systematic review and meta-analysis.

Early prediction of treatment response in nasopharyngeal carcinoma is clinically relevant for optimizing treatment strategies. This meta-analysis was performed to evaluate whether apparent diffusion coefficient (ADC) from diffusion-weighted imaging (DWI) can predict treatment response of patients with nasopharyngeal carcinoma. A systematic search of PubMed-MEDLINE and Embase was performed to identify relevant original articles until July 22, 2021. We included studies which performed DWI for predicting locoregional treatment response in nasopharyngeal carcinoma treated with neoadjuvant chemotherapy, definitive chemoradiation, or radiation therapy. Hazard ratios were meta-analytically pooled using a random-effects model for the pooled estimates of overall survival, local relapse-free survival, distant metastasis-free survival and their 95% CIs. ADC showed a pooled sensitivity of 87% (95% CI 72-94%) and specificity of 70% (95% CI 56-80%) for predicting treatment response. Significant between-study heterogeneity was observed for both pooled sensitivity (I2 = 68.5%) and specificity (I2 = 92.2%) (P < 0.01). The pooled hazard ratios of low pretreatment ADC for assessing overall survival, local relapse-free survival, and distant metastasis-free survival were 1.42 (95% CI 1.09-1.85), 2.31 (95% CI 1.42-3.74), and 1.35 (95% CI 1.05-1.74), respectively. In patients with nasopharyngeal carcinoma, pretreatment ADC demonstrated good predictive performance for treatment response.

Semi-Supervised Learning in Medical MRI Segmentation: Brain Tissue with White Matter Hyperintensity Segmentation Using FLAIR MRI.

White-matter hyperintensity (WMH) is a primary biomarker for small-vessel cerebrovascular disease, Alzheimer's disease (AD), and others. The association of WMH with brain structural changes has also recently been reported. Although fluid-attenuated inversion recovery (FLAIR) magnetic resonance imaging (MRI) provide valuable information about WMH, FLAIR does not provide other normal tissue information. The multi-modal analysis of FLAIR and T1-weighted (T1w) MRI is thus desirable for WMH-related brain aging studies. In clinical settings, however, FLAIR is often the only available modality. In this study, we thus propose a semi-supervised learning method for full brain segmentation using FLAIR. The results of our proposed method were compared with the reference labels, which were obtained by FreeSurfer segmentation on T1w MRI. The relative volume difference between the two sets of results shows that our proposed method has high reliability. We further evaluated our proposed WMH segmentation by comparing the Dice similarity coefficients of the reference and the results of our proposed method. We believe our semi-supervised learning method has a great potential for use for other MRI sequences and will encourage others to perform brain tissue segmentation using MRI modalities other than T1w.

2020 Imaging Guidelines for Thyroid Nodules and Differentiated Thyroid Cancer: Korean Society of Thyroid Radiology.

Imaging plays a key role in the diagnosis and characterization of thyroid diseases, and the information provided by imaging studies is essential for management planning. A referral guideline for imaging studies may help physicians make reasonable decisions and minimize the number of unnecessary examinations. The Korean Society of Thyroid Radiology (KSThR) developed imaging guidelines for thyroid nodules and differentiated thyroid cancer using an adaptation process through a collaboration between the National Evidence-based Healthcare Collaborating Agency and the working group of KSThR, which is composed of radiologists specializing in thyroid imaging. When evidence is either insufficient or equivocal, expert opinion may supplement the available evidence for recommending imaging. Therefore, we suggest rating the appropriateness of imaging for specific clinical situations in this guideline.

Development of Random Forest Algorithm Based Prediction Model of Alzheimer's Disease Using Neurodegeneration Pattern.

OBJECTIVE: Alzheimer's disease (AD) is the most common type of dementia and the prevalence rapidly increased as the elderly population increased worldwide. In the contemporary model of AD, it is regarded as a disease continuum involving preclinical stage to severe dementia. For accurate diagnosis and disease monitoring, objective index reflecting structural change of brain is needed to correctly assess a patient's severity of neurodegeneration independent from the patient's clinical symptoms. The main aim of this paper is to develop a random forest (RF) algorithm-based prediction model of AD using structural magnetic resonance imaging (MRI). METHODS: We evaluated diagnostic accuracy and performance of our RF based prediction model using newly developed brain segmentation method compared with the Freesurfer's which is a commonly used segmentation software. RESULTS: Our RF model showed high diagnostic accuracy for differentiating healthy controls from AD and mild cognitive impairment (MCI) using structural MRI, patient characteristics, and cognitive function (HC vs. AD 93.5%, AUC 0.99; HC vs. MCI 80.8%, AUC 0.88). Moreover, segmentation processing time of our algorithm (<5 minutes) was much shorter than of Freesurfer's (6-8 hours). CONCLUSION: Our RF model might be an effective automatic brain segmentation tool which can be easily applied in real clinical practice.

Deep Learning-Based Segmentation to Establish East Asian Normative Volumes Using Multisite Structural MRI.

Normative brain magnetic resonance imaging (MRI) is essential to interpret the state of an individual's brain health. However, a normative study is often expensive for small research groups. Although several attempts have been made to establish brain MRI norms, the focus has been limited to certain age ranges. This study aimed to establish East Asian normative brain data using multi-site MRI and determine the robustness of these data for clinical research. Normative MRI was gathered covering a wide range of cognitively normal East Asian populations (age: 18-96 years) from two open sources and three research sites. Eight sub-regional volumes were extracted in the left and right hemispheres using an in-house deep learning-based tool. Repeated measure consistency and multicenter reliability were determined using intraclass correlation coefficients and compared to a widely used tool, FreeSurfer. Our results showed highly consistent outcomes with high reliability across sites. Our method outperformed FreeSurfer in repeated measure consistency for most structures and multicenter reliability for all structures. The normative MRI we constructed was able to identify sub-regional differences in mild cognitive impairments and dementia after covariate adjustments. Our investigation suggests it is possible to provide a sound normative reference for neurodegenerative or aging research.

Split-Attention U-Net: A Fully Convolutional Network for Robust Multi-Label Segmentation from Brain MRI.

Multi-label brain segmentation from brain magnetic resonance imaging (MRI) provides valuable structural information for most neurological analyses. Due to the complexity of the brain segmentation algorithm, it could delay the delivery of neuroimaging findings. Therefore, we introduce Split-Attention U-Net (SAU-Net), a convolutional neural network with skip pathways and a split-attention module that segments brain MRI scans. The proposed architecture employs split-attention blocks, skip pathways with pyramid levels, and evolving normalization layers. For efficient training, we performed pre-training and fine-tuning with the original and manually modified FreeSurfer labels, respectively. This learning strategy enables involvement of heterogeneous neuroimaging data in the training without the need for many manual annotations. Using nine evaluation datasets, we demonstrated that SAU-Net achieved better segmentation accuracy with better reliability that surpasses those of state-of-the-art methods. We believe that SAU-Net has excellent potential due to its robustness to neuroanatomical variability that would enable almost instantaneous access to accurate neuroimaging biomarkers and its swift processing runtime compared to other methods investigated.