Reduction in limb-movement complexity at term-equivalent age is associated with motor developmental delay in very-preterm or very-low-birth-weight infants.

Reduced complexity during the writhing period can be crucial in the spontaneous movements of high-risk infants for neurologic impairment. This study aimed to verify the association between quantified complexity of upper and lower-limb movements at term-equivalent age and motor development in very-preterm or very-low-birth-weight infants. Video images of spontaneous movements at term-equivalent age were collected from very-preterm or very-low-birth-weight infants. A pretrained pose-estimation model and sample entropy (SE) quantified the complexity of the upper- and lower-limb movements. Motor development was evaluated at 9 months of corrected age using Bayley Scales of Infant and Toddler Development, Third Edition. The SE measures were compared between infants with and without motor developmental delay (MDD). Among 90 infants, 11 exhibited MDD. SE measures at most of the upper and lower limbs were significantly reduced in infants with MDD compared to those without MDD (p < 0.05). Composite scores in the motor domain showed significant positive correlations with SE measures at most upper and lower limbs (p < 0.05). The results show that limb-movement complexity at term-equivalent age is reduced in infants with MDD at 9 months of corrected age. SE of limb movements can be a potentially useful kinematic parameter to detect high-risk infants for MDD.

Comparison of Four Diagnostic Guidelines for Hepatocellular Carcinoma Using Gadoxetic Acid-enhanced Liver MRI.

Background Noninvasive diagnostic guidelines for hepatocellular carcinoma (HCC) vary across different global geographic areas, especially regarding criteria about gadoxetic acid-enhanced MRI. Purpose To compare the diagnostic performance of four different international HCC diagnosis guidelines and readers' judgment in diagnosing HCC using gadoxetic acid-enhanced MRI in patients at high risk for HCC. Materials and Methods This retrospective study included patients who had not undergone treatment, were at risk for HCC, and who underwent gadoxetic acid-enhanced MRI from January 2015 to June 2018 from 11 tertiary hospitals in South Korea. Four radiologists independently reviewed focal liver lesions (FLLs) according to four guidelines: American Association for the Study of Liver Diseases (AASLD)/Liver Imaging Reporting and Data System (LI-RADS), Korean Liver Cancer Association-National Cancer Center (KLCA-NCC), European Association for the Study of the Liver (EASL), and Asian Pacific Association for the Study of the Liver (APASL). Reader judgment (HCC or not HCC) was also recorded. Malignant FLLs were confirmed at pathology, and histologic and clinical follow-up data were used for benign FLLs. The guidelines' diagnostic performance was compared using generalized estimating equations. Additionally, the diagnostic odds ratio was assessed. Results A total of 2445 FLLs (median size, 27.4 mm) were analyzed in 2237 patients (mean age, 59 years ± 11 [SD]; 1666 male patients); 69.3% (1694 of 2445) were HCCs. KLCA-NCC showed the highest accuracy (80.0%; 95% CI: 78.7, 81.2; P = .001), with high sensitivity in Eastern guidelines (APASL, 89.1% [95% CI: 87.8, 90.3]; KLCA-NCC, 78.2% [95% CI: 76.6, 79.7]) and high specificity in Western guidelines (AASLD/LI-RADS, 89.6% [95% CI: 87.8, 91.2]; EASL, 88.1% [95% CI: 86.2, 89.9]) (P = .001). The diagnostic odds ratios were 20.7 (95% CI: 17.0, 25.3) for AASLD/LI-RADS, 18.9 (95% CI: 15.8, 22.6) for KLCA-NCC, 16.8 (95% CI: 13.8, 20.4) for EASL, and 8.9 (95% CI: 7.4, 10.7) for APASL. The readers' judgment demonstrated higher accuracy than that of the guidelines (accuracy, 86.0%; 95% CI: 84.9, 86.9; P = .001). Conclusion Among four different international HCC diagnosis guidelines, Eastern guidelines demonstrated higher sensitivity, whereas Western guidelines displayed higher specificity. KLCA-NCC achieved the highest accuracy, and AASLD/LI-RADS exhibited the highest diagnostic odds ratio. © RSNA, 2024 Supplemental material is available for this article.

Reducing medication errors on emergency department discharge: Evaluation of a collaborative pharmacist-medical officer discharge prescription planning model in a tertiary hospital emergency short stay unit.

OBJECTIVE: To implement and evaluate the impact of a collaborative pharmacist-medical officer model of planning discharge prescriptions, Partnered Pharmacist Discharge Prescription Planning (PPDPP) on the safe use of medicines on discharge in an ED short stay unit (SSU). METHODS: A prospective pre- and post-intervention study measured the proportion of medication errors on discharge prescriptions from the SSU using the Five Rights (5Rs) method. Pharmacists assessed discharge prescriptions generated by the medical officers (MO) during the pre-intervention phase (standard practice). During the PPDPP phase, pharmacists planned electronic prescriptions in consultation with MO and completed prescriptions were independently assessed by another pharmacist. RESULTS: There were 163 and 147 prescriptions collected during the pre- and post-intervention phases, respectively. There was a significant difference in the proportion of discharge prescriptions that met all 5Rs between the standard practice (47.2%) and PPDPP phase (91.8%) (P < 0.001). There was no statistical difference seen in the mean time taken from discharge decision to prescriptions given to patients or patients leaving the SSU between the two phases. There was a non-statically significant trend towards a decrease in time taken for patients to obtain prescriptions by 11% (P = 0.16) and for actual departure time by 6% (P = 0.46). Additionally, the proportion of opioids prescribed as one of the high-risk medication classes reduced from 23.8% to 16.2% (P = 0.023) with the PPDPP model. CONCLUSION: The PPDPP model improved medications safety on discharge from the ED SSU. The PPDPP did not impact patient flow parameters as measured in this study.

[Impact of Anthropometric Indices of Obesity on the Risk of Incident Hypertension in Adults with Prehypertension: A Secondary Analysis of a Cohort Study].

PURPOSE: This study aimed to investigate the impact of anthropometric indices of obesity (body mass index [BMI], waist circumference, waist hip ratio, and body fat percentage) on the incidence of hypertension in adults with prehypertension. METHODS: A longitudinal study design using secondary data form the Korean Genome and Epidemiology Study was employed. The study included 1,838 adults with prehypertension tracked every two years from 2001 to 2018. Statistical analyses, including frequency assessments, number of cases per 1,000 person-years, log-rank tests, Kaplan-Meier curves, and Cox's proportional hazards regression, were conducted using SPSS version 25. RESULTS: Over the observation period (15,783.6 person-years), 1,136 individuals developed hypertension. The incidence of hypertension was significantly higher in the obesity groups defined by BMI (hazard ratio [HR] = 1.33), waist circumference (HR = 1.34), waist hip ratio (HR = 1.29), and body fat percentage (HR = 1.31) compared to the non-obese group. These findings indicate an increased risk of hypertension associated with obesity as measured by these indices. CONCLUSION: The study underscores the importance of avoiding obesity to prevent hypertension in individuals with prehypertension. Specifically, BMI, waist circumference, waist hip circumference, and body fat percentage were identified as significant risk factors for hypertension. The results suggest the need for individualized weight control interventions, emphasizing the role of health professionals in addressing the heightened hypertension risk in this population.

Self-supervised multi-modal training from uncurated images and reports enables monitoring AI in radiology.

The escalating demand for artificial intelligence (AI) systems that can monitor and supervise human errors and abnormalities in healthcare presents unique challenges. Recent advances in vision-language models reveal the challenges of monitoring AI by understanding both visual and textual concepts and their semantic correspondences. However, there has been limited success in the application of vision-language models in the medical domain. Current vision-language models and learning strategies for photographic images and captions call for a web-scale data corpus of image and text pairs which is not often feasible in the medical domain. To address this, we present a model named medical cross-attention vision-language model (Medical X-VL), which leverages key components to be tailored for the medical domain. The model is based on the following components: self-supervised unimodal models in medical domain and a fusion encoder to bridge them, momentum distillation, sentencewise contrastive learning for medical reports, and sentence similarity-adjusted hard negative mining. We experimentally demonstrated that our model enables various zero-shot tasks for monitoring AI, ranging from the zero-shot classification to zero-shot error correction. Our model outperformed current state-of-the-art models in two medical image datasets, suggesting a novel clinical application of our monitoring AI model to alleviate human errors. Our method demonstrates a more specialized capacity for fine-grained understanding, which presents a distinct advantage particularly applicable to the medical domain.

Prevalence and predisposing factors of post-stroke complex regional pain syndrome: Retrospective case-control study.

INTRODUCTION: Poststroke complex regional pain syndrome (CRPS) is an important complication in stroke survivors. The identification of factors associated with post-stroke CRPS is important for preventive measures and early diagnosis. METHODS: A total of 141 first-ever stroke survivors in the subacute stage were retrospectively analyzed. Demographic data, diagnosis time, duration of hospitalization, location of brain lesion, etiology, comorbidities, and blood test findings were investigated. Clinical data included Medical Research Council (MRC) grade, Fugl-Meyer assessment (FMA), National Institute for Health Stroke Scale (NIHSS), Berg Balance Scale (BBS). RESULTS: Among 141 patients with subacute stroke, 22 were diagnosed with CRPS, with a prevalence of 15.6 %. The mean time to diagnosis was 38.6 (±16.5) days. The prevalence according to the degree of paralysis was 33.3 % in MRC grades 0 and 1, 8.6 % in grade 2, and 0 % in grade 3 or higher. The incidence rates within 1 month after stroke were 1.42 % and 22.47 % between 1 and 3 months after stroke, respectively. The independent risk factors for CRPS were hospitalization duration and FMA, NIHSS, and BBS scores. The sensitivity and specificity of the NIHSS score for predicting post-stroke CRPS were 86.4 % and 59.7 %, respectively, with an optimal cutoff value of 7.5. CONCLUSIONS: CRPS of the affected upper limb in stroke patients is associated with stroke severity, including paralysis, and the incidence increases over time during the subacute phase. Additionally, having sufficient strength to move through a full range of motion against gravity had a protective effect against CRPS.

Recurred Uterine Cervical Cancer Initially Manifesting as Hemorrhagic Cyst: A Case Report Recurred Uterine Cervical Cancer Initially Manifesting.

BACKGROUND: Recurrence of uterine cervical cancer is common and often shows a dismal prognosis. Local recurrence usually manifests as solid soft tissue lesions and has rarely been reported to have cystic lesions. CASE PRESENTATION: Herein, we report a case of recurrent uterine cervical cancer with initial manifestation as a hemorrhagic cyst, assessed using strain sonoelastography, CT, and MRI. CONCLUSION: Although cystic recurrence is uncommon, newly detected simple or complex cystic lesions should be closely monitored.

Renal protection CT protocol using low-dose and low-concentration iodine contrast medium in at-risk patients of HCC and with chronic kidney disease: a randomized controlled non-inferiority trial.

BACKGROUND: Although efforts have been made to reduce the dose of Contrast Medium (CM) to improve patient safety, there are ongoing concerns regarding its potential effects on image quality and diagnostic performance. Moreover, research is lacking to establish a lower limit for safe and effective CM dose reduction. To determine whether the image quality of contrast-enhanced liver computed tomography (CT) using a reduced amount of iodinated CM was similar to that of standard liver CT. METHODS: We enrolled participants at risk for hepatocellular carcinoma with decreased estimated glomerular filtration rates (< 60 mL/min/1.73m2). Participants were randomly assigned to the standard group or the renal protection protocol (RPP) group. In the standard group, images were reconstructed using hybrid iterative reconstruction (iDose), while in the RPP group, low monoenergetic (50-keV) images and deep learning (DL)-based iodine-boosting reconstruction were used. Four radiologists independently assessed image quality and lesion conspicuity. RESULTS: Fifty-two participants were assigned to the standard (n = 25) or RPP (n = 27) groups. The iodine load was significantly lower in the RPP group than in the standard group (301.5 ± 1.71 vs. 524 ± 7.37 mgI/kg, P < 0.001). The 50-keV and DL-based iodine-boosting images from the RPP group exhibited higher image contrast than those from the standard group during arterial (3.60 ± 0.65, 3.75 ± 0.60, and 3.09 ± 0.43, respectively) and portal venous phases (4.01 ± 0.49, 3.86 ± 0.42, and 3.21 ± 0.31, respectively) (P < 0.05 for all). Overall image quality was superior in the RPP group (P < 0.05 for all). No significant difference in lesion conspicuity was observed (P > 0.017). CONCLUSIONS: The reduction in image contrast and overall image quality caused by decreased CM can be restored using either low monoenergetic imaging or DL-based iodine-boosting reconstruction. TRIAL REGISTRATION: clinicaltrials.gov, NCT04024514, Registered July 18, 2019, prospectively registered, https://classic. CLINICALTRIALS: gov/ct2/show/NCT04024514 .

Deep Learning-Enabled Detection of Pneumoperitoneum in Supine and Erect Abdominal Radiography: Modeling Using Transfer Learning and Semi-Supervised Learning.

OBJECTIVE: Detection of pneumoperitoneum using abdominal radiography, particularly in the supine position, is often challenging. This study aimed to develop and externally validate a deep learning model for the detection of pneumoperitoneum using supine and erect abdominal radiography. MATERIALS AND METHODS: A model that can utilize "pneumoperitoneum" and "non-pneumoperitoneum" classes was developed through knowledge distillation. To train the proposed model with limited training data and weak labels, it was trained using a recently proposed semi-supervised learning method called distillation for self-supervised and self-train learning (DISTL), which leverages the Vision Transformer. The proposed model was first pre-trained with chest radiographs to utilize common knowledge between modalities, fine-tuned, and self-trained on labeled and unlabeled abdominal radiographs. The proposed model was trained using data from supine and erect abdominal radiographs. In total, 191212 chest radiographs (CheXpert data) were used for pre-training, and 5518 labeled and 16671 unlabeled abdominal radiographs were used for fine-tuning and self-supervised learning, respectively. The proposed model was internally validated on 389 abdominal radiographs and externally validated on 475 and 798 abdominal radiographs from the two institutions. We evaluated the performance in diagnosing pneumoperitoneum using the area under the receiver operating characteristic curve (AUC) and compared it with that of radiologists. RESULTS: In the internal validation, the proposed model had an AUC, sensitivity, and specificity of 0.881, 85.4%, and 73.3% and 0.968, 91.1, and 95.0 for supine and erect positions, respectively. In the external validation at the two institutions, the AUCs were 0.835 and 0.852 for the supine position and 0.909 and 0.944 for the erect position. In the reader study, the readers' performances improved with the assistance of the proposed model. CONCLUSION: The proposed model trained with the DISTL method can accurately detect pneumoperitoneum on abdominal radiography in both the supine and erect positions.

Erratum: Relationship Between MicroRNA Signature and Arterial Stiffness in Patients With Ischemic Stroke.

This corrects the article on p. 28 in vol. 19, PMID: 36606643.

Comparison between Conventional Breath-hold and Respiratory-triggered Magnetic Resonance Cholangiopancreatography with and without Compressed Sensing: Cross-sectional Study.

AIM: The application of compressed sensing (CS) has enabled breath-hold 3D-MRCP with a shorter acquisition time in clinical practice. INTRODUCTION: To compare the image quality of breath-hold (BH) and respiratory-triggered (RT) 3D-MRCP with or without CS application in the same study population. METHODS: In this retrospective study, from February to July 2020, a total of 98 consecutive patients underwent four different acquisition types of 3D-MRCP.; 1) BH MRCP with the generalized autocalibrating partially parallel acquisition (GRAPPA) (BH-GRAPPA), 2) RT-GRAPPA-MRCP, 3) RT-CS-MRCP and 4) BH-CS-MRCP. Relative contrast of common bile duct, 5-scale visibility score of biliary pancreatic ducts, 3-scale artifact score and 5-scale overall image quality score were evaluated by two abdominal radiologists. RESULTS: Relative contrast value was significantly higher in BH-CS or RT-CS than in RT-GRAPPA (0.90 ± 0.057 and 0.89 ± 0.079, respectively, vs. 0.82 ± 0.071, p < 0.01) or BH-GRAPPA (vs. 0.77 ± 0.080, p < 0.01). The area affected by artifact was significantly lower in BH-CS among 4 MRCPs (p < 0.08). Overall image quality score in BH-CS was significantly higher than BH-GRAPPA (3.40 vs. 2.71, p < 0.01). There were no significant differences between RT-GRAPPA and BH-CS (vs. 3.13, p = 0.67) in overall image quality. CONCLUSION: In this study, our results revealed BH-CS had higher relative contrast and comparable or superior image quality among four MRCP sequences.

Transient elastography with controlled attenuation parameter versus two-dimensional shear wave elastography with attenuation imaging for the evaluation of hepatic steatosis and fibrosis in NAFLD.

PURPOSE: This study compared the controlled attenuation parameter (CAP) to attenuation imaging (ATI) in the diagnosis of steatosis and transient elastography (TE) to two-dimensional shear wave elastography (2D-SWE) for the diagnosis of fibrosis in a prospectively constructed nonalcoholic fatty liver disease (NAFLD) patient cohort. METHODS: Participants who underwent TE with CAP were included from a previously constructed NAFLD cohort with multiparametric ultrasound data. The degree of hepatic steatosis and stage of liver fibrosis were assessed. Diagnostic performance was evaluated using the area under the receiver operating characteristic curve (AUROC) for the grades of steatosis (S1-3) and fibrosis (F0-F4). RESULTS: There were 105 participants. The distribution of hepatic steatosis grades (S0-S3) and liver fibrosis stages (F0-F4) was as follows: S0, n=34; S1, n=41; S2, n=22; S3, n=8; F0, n=63; F1, n=25; F2, n=5; F3, n=7; and F4, n=5. No significant difference was found between CAP and ATI in detecting ≥S1 (AUROC: 0.93 vs. 0.93, P=0.956) or ≥S2 (0.94 vs. 0.94, P=0.769). However, the AUROC of ATI in detecting ≥S3 was significantly higher than that of CAP (0.94 vs. 0.87, P=0.047). Regarding the detection of liver fibrosis, no significant difference was found between TE and 2D-SWE. The AUROCs of TE and 2D-SWE were as follows: ≥F1, 0.94 vs. 0.89 (P=0.107); ≥F2, 0.89 vs. 0.90 (P=0.644); ≥F3, 0.91 vs. 0.90 (P=0.703); and ≥F4, 0.88 vs. 0.92 (P=0.209). CONCLUSION: 2D-SWE and TE showed comparable diagnostic performance in assessing liver fibrosis, and ATI provided significantly better performance in detecting ≥S3 steatosis than CAP.

Efficacy and safety of mucous fistula refeeding in preterm infants: an exploratory randomized controlled trial.

BACKGROUND: This study aimed to evaluate whether mucous fistula refeeding (MFR) is safe and beneficial for the growth and intestinal adaptation of preterm infants with enterostomies. METHODS: This exploratory randomized controlled trial enrolled infants born before 35 weeks' gestation with enterostomy. If the stomal output was ≥ 40 mL/kg/day, infants were assigned to the high-output MFR group and received MFR. If the stoma output was < 40 mL/kg/day, infants were randomized to the normal-output MFR group or the control group. Growth, serum citrulline levels, and bowel diameter in loopograms were compared. The safety of MFR was evaluated. RESULTS: Twenty infants were included. The growth rate increased considerably, and the colon diameter was significantly larger after MFR. However, the citrulline levels did not significantly differ between the normal-output MFR and the control group. One case of bowel perforation occurred during the manual reduction for stoma prolapse. Although the association with MFR was unclear, two cases of culture-proven sepsis during MFR were noted. CONCLUSIONS: MFR benefits the growth and intestinal adaptation of preterm infants with enterostomy and can be safely implemented with a standardized protocol. However, infectious complications need to be investigated further. TRIAL REGISTRATION: clinicaltrials.gov NCT02812095, retrospectively registered on June 6, 2016.

Changes in microRNAs during Storage and Processing of Breast Milk.

Human breast milk (HBM) is the ideal source of nutrients for infants and is rich in microRNA (miRNA). In recent years, expressed breast milk feeding rather than direct breastfeeding has become increasingly prevalent for various reasons. Expressed HBM requires storage and processing, which can cause various changes in the ingredients. We investigated how the miRNAs in HBM change due to processes often used in real life. HBM samples collected from 10 participants were each divided into seven groups according to the storage temperature, thawing method, and storage period. In addition, we analyzed the miRNA changes in each group. The number of microRNAs that showed significant expression was not large compared to the thousands of miRNAs contained in breast milk. Therefore, it is difficult to suggest that the various storage and thawing processes have a great influence on the overall expression of miRNA. However, a short-term refrigeration storage method revealed little change in nutrients compared to other storage and thawing methods. Taking all factors into consideration, short-term refrigeration is recommended to minimize changes in the composition or function of breast milk.

Relationship Between MicroRNA Signature and Arterial Stiffness in Patients With Ischemic Stroke.

BACKGROUND AND PURPOSE: We investigated whether circulating microRNAs (miRNAs) is associated with arterial stiffness in patients with acute ischemic stroke. METHODS: We recruited patients with acute ischemic stroke who were admitted to a university hospital stroke center and underwent carotid-femoral pulse wave velocity (cfPWV) measurement using SphygmoCor (AtCor Medical, Sydney, Australia) and brachial-ankle PWV using a volume-plethysmography device (VP-1000, Omron Colin, Komaki, Japan). Circulating miRNAs were measured in venous blood samples stored in EDTA. We selected five miRNAs (miR-17, miR-93, miR-450, miR-629, and let-7i) related to atherosclerosis based on a literature review. Pearson's correlation analysis was applied to the correlations between miRNAs and arterial stiffness parameters. Finally, multivariable linear regression analysis was performed to identify the independent factors for cfPWV. RESULTS: This study included 70 patients (age=71.1±10.3 years [mean±SD], 29 females). The expression levels of miR-93 (r=-0.27, p=0.049) and let-7i (r=-0.27, p=0.039) were inversely correlated with cfPWV. Multivariable linear regression analysis including age, hypertension, and estimated glomerular filtration rate showed that let-7i was independently related with cfPWV (standardized coefficient=-0.262, p=0.036). Correlation analysis indicated that let-7i was positively associated with visceral muscle Hounsfield units on computed tomography (r=0.264, p=0.043). CONCLUSIONS: The expression level of let-7i was independently related to arterial stiffness in patients with cerebral infarction, suggesting that it plays a pathophysiological role in atherosclerosis.

A Huge Gastric Angiolipoma Presenting with Acute Upper Gastrointestinal Hemorrhage: A Case Report.

BACKGROUND: Angiolipoma is a benign neoplasm mainly composed of adipose tissue and proliferating blood vessels and is relatively rare in the gastrointestinal tract. And among them, gastric angiolipomas are extremely rare and tend to be small. CASE PRESENTATION: We report the clinical and imaging features of a patient with a huge angiolipoma in the stomach and an episode of hematemesis and melena, caused by the ulceration of the gastric mucosa overlying the gastric subepithelial angiolipoma revealed by the endoscopic evaluation. The patient was anemic, and the anemia resolved after local surgical resection of the tumor. We also reviewed the imaging and histological features of the presenting gastric angiolipoma. CONCLUSION: Radiologists should be aware of this rare benign gastric tumor that may present with gastrointestinal hemorrhage.

Incidental Diagnosis of Ovarian Torsion Due to a Krukenberg Tumor Originating from Gastric Cancer: A Case Report.

BACKGROUND: Benign lesions cause ovarian torsion more commonly than malignant ovarian tumors. Further, Krukenberg tumors have rarely been reported to cause ovarian torsion. CASE PRESENTATION: Herein, we present the case of a patient with an incidentally discovered ovarian mass, which was finally diagnosed as a Krukenberg tumor accompanying ovarian torsion with primary stomach cancer. We further review the clinical, imaging, and histological features of Krukenberg tumors. CONCLUSION: Radiologists should be aware of Krukenberg tumors that may present with ovarian torsion.

Tunable image quality control of 3-D ultrasound using switchable CycleGAN.

In contrast to 2-D ultrasound (US) for uniaxial plane imaging, a 3-D US imaging system can visualize a volume along three axial planes. This allows for a full view of the anatomy, which is useful for gynecological (GYN) and obstetrical (OB) applications. Unfortunately, the 3-D US has an inherent limitation in resolution compared to the 2-D US. In the case of 3-D US with a 3-D mechanical probe, for example, the image quality is comparable along the beam direction, but significant deterioration in image quality is often observed in the other two axial image planes. To address this, here we propose a novel unsupervised deep learning approach to improve 3-D US image quality. In particular, using unmatched high-quality 2-D US images as a reference, we trained a recently proposed switchable CycleGAN architecture so that every mapping plane in 3-D US can learn the image quality of 2-D US images. Thanks to the switchable architecture, our network can also provide real-time control of image enhancement level based on user preference, which is ideal for a user-centric scanner setup. Extensive experiments with clinical evaluation confirm that our method offers significantly improved image quality as well user-friendly flexibility.

MR Image Denoising and Super-Resolution Using Regularized Reverse Diffusion.

Patient scans from MRI often suffer from noise, which hampers the diagnostic capability of such images. As a method to mitigate such artifacts, denoising is largely studied both within the medical imaging community and beyond the community as a general subject. However, recent deep neural network-based approaches mostly rely on the minimum mean squared error (MMSE) estimates, which tend to produce a blurred output. Moreover, such models suffer when deployed in real-world situations: out-of-distribution data, and complex noise distributions that deviate from the usual parametric noise models. In this work, we propose a new denoising method based on score-based reverse diffusion sampling, which overcomes all the aforementioned drawbacks. Our network, trained only with coronal knee scans, excels even on out-of-distribution in vivo liver MRI data, contaminated with a complex mixture of noise. Even more, we propose a method to enhance the resolution of the denoised image with the same network. With extensive experiments, we show that our method establishes state-of-the-art performance while having desirable properties which prior MMSE denoisers did not have: flexibly choosing the extent of denoising, and quantifying uncertainty.