Noise2Recon: Enabling SNR-robust MRI reconstruction with semi-supervised and self-supervised learning

PURPOSE

To develop a method for building MRI reconstruction neural networks robust to changes in signal-to-noise ratio (SNR) and trainable with a limited number of fully sampled scans.

METHODS

We propose Noise2Recon, a consistency training method for SNR-robust accelerated MRI reconstruction that can use both fully sampled (labeled) and undersampled (unlabeled) scans. Noise2Recon uses unlabeled data by enforcing consistency between model reconstructions of undersampled scans and their noise-augmented counterparts. Noise2Recon was compared to compressed sensing and both supervised and self-supervised deep learning baselines. Experiments were conducted using retrospectively accelerated data from the mridata three-dimensional fast-spin-echo knee and two-dimensional fastMRI brain datasets. All methods were evaluated in label-limited settings and among out-of-distribution (OOD) shifts, including changes in SNR, acceleration factors, and datasets. An extensive ablation study was conducted to characterize the sensitivity of Noise2Recon to hyperparameter choices.

RESULTS

In label-limited settings, Noise2Recon achieved better structural similarity, peak signal-to-noise ratio, and normalized-RMS error than all baselines and matched performance of supervised models, which were trained with14 × $$ 14\times $$ more fully sampled scans. Noise2Recon outperformed all baselines, including state-of-the-art fine-tuning and augmentation techniques, among low-SNR scans and when generalizing to OOD acceleration factors. Augmentation extent and loss weighting hyperparameters had negligible impact on Noise2Recon compared to supervised methods, which may indicate increased training stability.

CONCLUSION

Noise2Recon is a label-efficient reconstruction method that is robust to distribution shifts, such as changes in SNR, acceleration factors, and others, with limited or no fully sampled training data.

The GIANT trial (ECOG-ACRIN EA2186) methods paper: A randomized phase II study of gemcitabine and nab-paclitaxel compared with 5-fluorouracil, leucovorin, and liposomal irinotecan in older patients with treatment-naïve metastatic pancreatic cancer - defining a new treatment option for older vulnerable patients

INTRODUCTION

Pancreatic cancer is the fourth leading cause of cancer-related death in the US with an increasing incidence in older adults (OA) over age 70. There are currently no treatment guidelines for OA with metastatic pancreatic cancer (mPCA) and selecting a chemotherapy regimen for these patients is subjective, based largely on chronologic age and performance status (PS). Geriatric screening tools provide a more objective and accurate evaluation of a patient's overall health but have not yet been validated in patient selection for mPCA treatment. This study aims to elucidate the optimal chemotherapy treatment of vulnerable OA with mPCA and understand the geriatric factors that affect outcomes in this population.

METHODS/DESIGN

The GIANT (ECOG-ACRIN EA2186) study is multicenter, randomized phase II trial enrolling patients over age 70 with newly diagnosed mPCA. This study utilizes a screening geriatric assessment (GA) which characterizes patients as fit, vulnerable, or frail. Patients with mild abnormalities in functional status and/or cognition, moderate comorbidities, or over age 80 are considered vulnerable. Enrolled patients are randomized to one of two dose-reduced treatment regimens (gemcitabine/nab-paclitaxel every other week, or dose-reduced 5-fluoruracil (5FU)/ liposomal irinotecan (nal-IRI) every other week). GA and quality of life (QoL) evaluations are completed prior to treatment initiation and at each disease evaluation. Overall survival (OS) is the primary endpoint, with secondary endpoints including progression free survival (PFS) and objective response rate (ORR). Enrolled patients will be stratified by age (70-74 vs ≥75) and ECOG PS (0-1 vs 2). Additional endpoints of interest for OA include evaluation of risk factors identified through GA, QoL evaluation, and toxicities of interest for older adults. Correlative studies include assessment of pro-inflammatory biomarkers of aging in the blood (IL-6, CRP) and imaging evaluation of sarcopenia as predictors of treatment tolerance.

DISCUSSION

The GIANT study is the first randomized, prospective national trial evaluating vulnerable OA with mPCA aimed at developing a tailored treatment approach for this patient population. This trial has the potential to establish a new way of objectively selecting vulnerable OA with mPCA for modified treatment and to establish a new standard of care in this growing patient population.

TRIAL REGISTRATION

This trial is registered with ClinicalTrial.gov Identifier NCT04233866.

Artificial Intelligence System for Automatic Quantitative Analysis and Radiology Reporting of Leg Length Radiographs

Leg length discrepancies are common orthopedic problems with the potential for poor functional outcomes. These are frequently assessed using bilateral leg length radiographs. The objective was to determine whether an artificial intelligence (AI)-based image analysis system can accurately interpret long leg length radiographic images. We built an end-to-end system to analyze leg length radiographs and generate reports like radiologists, which involves measurement of lengths (femur, tibia, entire leg) and angles (mechanical axis and pelvic tilt), describes presence and location of orthopedic hardware, and reports laterality discrepancies. After IRB approval, a dataset of 1,726 extremities (863 images) from consecutive examinations at a tertiary referral center was retrospectively acquired and partitioned into train/validation and test sets. The training set was annotated and used to train a fasterRCNN-ResNet101 object detection convolutional neural network. A second-stage classifier using a EfficientNet-D0 model was trained to recognize the presence or absence of hardware within extracted joint image patches. The system was deployed in a custom web application that generated a preliminary radiology report. Performance of the system was evaluated using a holdout 220 image test set, annotated by 3 musculoskeletal fellowship trained radiologists. At the object detection level, the system demonstrated a recall of 0.98 and precision of 0.96 in detecting anatomic landmarks. Correlation coefficients between radiologist and AI-generated measurements for femur, tibia, and whole-leg lengths were > 0.99, with mean error of < 1%. Correlation coefficients for mechanical axis angle and pelvic tilt were 0.98 and 0.86, respectively, with mean absolute error of < 1°. AI hardware detection demonstrated an accuracy of 99.8%. Automatic quantitative and qualitative analysis of leg length radiographs using deep learning is feasible and holds potential in improving radiologist workflow.

A randomized phase II study of gemcitabine and nab-paclitaxel compared with 5-fluorouracil, leucovorin, and liposomal irinotecan in older patients with treatment-naïve metastatic pancreatic cancer (GIANT): ECOG-ACRIN EA2186—Trials in progress

TPS4185

Background: Evidence-based data is lacking to guide the care of older adults with newly diagnosed metastatic pancreatic cancer (mPCA). As a result, treatment approach and the selection of chemotherapy regimens are often extrapolated from data from younger patients. Furthermore, vulnerable older adults are often treated with dose adjusted regimens with limited data to support this practice. EA2186 is a phase II randomized controlled trial, and the first prospective study aiming to define the optimal treatment approach of vulnerable older adults with newly diagnosed mPCA. Methods: Patients aged 70 years and over with histologically confirmed pancreatic adenocarcinoma, evidence of metastatic disease, ECOG PS 0-2 and adequate organ function, who are considered vulnerable are eligible for this trial (accrual target 184). This study utilizes a screening geriatric assessment which characterize patients as fit, vulnerable or frail by evaluating functional status, cognition and co-morbidities. Vulnerable patients according to this screening assessment are those with mild abnormalities in functional status, comorbidities and/or cognition, or older than 80 years of age. Those patients will be randomized to receive either modified Gemcitabine/Nab-Paclitaxel or dose-reduced 5-Fluorouracil Leucovorin and Liposomal Irinotecan every 2 weeks. A comprehensive geriatric assessment (GA) and quality of life (QOL) evaluation are completed prior to initiation of therapy for all randomized patients. Follow up will continue until disease progression or withdrawal, with repeated GA and QOL assessments at each disease evaluation. Overall survival is the primary objective, with secondary objectives including progression free survival, and response rate. Enrolled patients will be stratified by age 70-74 vs ≥75, and ECOG PS 0-1 vs 2. Additional endpoints of interest for older adults include: evaluation of risk factors identified through GA, and capturing toxicities of interest for this patient population (i.e. hospitalization, deterioration in PS, and falls). Correlative studies include assessment of pro-inflammatory biomarkers or aging in the blood (IL-6 and CRP) as well as imaging evaluation of sarcopenia and body composition as predictors of treatment tolerance. Clinical trial information: NCT04233866.

Body composition measurements and overall survival in patients with resectable pancreatic adenocarcinoma receiving neoadjuvant chemotherapy: Analysis from SWOG S1505

4131

Background: Sarcopenia and sarcopenic obesity have been associated with overall survival (OS) in patients (pts) with borderline resectable and advanced pancreatic ductal adenocarcinoma (PDA), but little is known about the effect of body composition on OS in pts with resectable PDA. We examined the relationship between skeletal muscle and adipose tissue measurements on baseline computed tomography (CT) and OS of pts with resectable PDA in a secondary analysis of SWOG S1505 (NCT02562716). Methods: SWOG S1505 enrolled pts with resectable PDA who were randomized to receive neoadjuvant FOLFIRINOX or gemcitabine-nab paclitaxel, followed by surgical resection. Baseline axial CT images at the L3 level were analyzed with externally validated software and measurements were recorded for skeletal muscle area (SMA), density (SMD) and index (SMI); visceral adipose tissue area (VATA) and density (VATD); and subcutaneous adipose tissue area (SATA) and density (SATD). Sarcopenia was defined as SMI < 52 cm2/m2 for men and < 39 cm2/m2 for women; sarcopenic obesity was defined as sarcopenia and a body mass index (BMI) >30 kg/m2. The relationships between CT metrics and OS were analyzed using Cox regression models, with 95% CI. Statistical significance was defined as p < 0.05. Results: Of 98 pts with available baseline abdominal CT, 8 were excluded for scan quality, resulting in 90 evaluable cases: 51 men (57%), 39 women (43%); mean age, 63.2 years, SD 8.5; mean BMI, 29.3 kg/m2, SD 6.4; 80 (89%) White, 6 (7%) Black, and 4 (4%) unknown. Sarcopenia was present in 32 (36%) and sarcopenic obesity in 10 (11%) patients. Univariable analyses for the variables of interest indicated VATA (HR 1.24; 0.97-1.60; p = 0.09) and SMD (HR 0.75; 0.57-0.98; p = 0.04) were associated with OS. Analyses adjusted for sex, race, age, BMI, performance score, contrast use, sarcopenia, and sarcopenic obesity showed VATA was associated with OS (HR 1.58; 1.0-2.51; p = 0.05). No significant difference in median OS was observed between pts with vs. without sarcopenia (OS 23.6 [19.3-NA] vs. 27.9 months [18.6-NA], respectively). Pts with vs. without sarcopenic obesity had lower median OS: 18.6 (14.7-NA) vs. 25.1 (10.5-46.0) months, respectively, but this difference was not statistically significant (HR 1.90, 95%CI 0.81-4.47, p = 0.14). Conclusions: This is one of the first studies to systematically evaluate body composition parameters in a prospective trial of patients with resectable PDA who received neoadjuvant chemotherapy. We found that visceral fat (VATA) is a prognostic marker in this population, but that sarcopenia may not be predictive in early PDA. Further studies to define the impact of longitudinal changes in body composition on individual outcomes may provide greater precision in predicting OS for subsets of pts with pancreatic cancer. Clinical trial information: NCT02562716.

Sarcopenia: Beyond Muscle Atrophy and into the New Frontiers of Opportunistic Imaging, Precision Medicine, and Machine Learning

As populations continue to age worldwide, the impact of sarcopenia on public health will continue to grow. The clinically relevant and increasingly common diagnosis of sarcopenia is at the confluence of three tectonic shifts in medicine: opportunistic imaging, precision medicine, and machine learning. This review focuses on the state-of-the-art imaging of sarcopenia and provides context for such imaging by discussing the epidemiology, pathophysiology, consequences, and future directions in the field of sarcopenia.

Musculotendinous Disorders in the Upper Extremity: Part 2. MRI of the Elbow, Forearm, Wrist, and Hand

Muscle and tendon injuries about the elbow and more distal aspect of the upper arm are commonly evaluated and characterized with magnetic resonance imaging. This article reviews our experience with muscle and tendon injury as well as denervation of muscles. We focus on abnormalities of the biceps and the triceps tendons about the elbow as well as abnormalities of the flexor and extensor tendons about the elbow, forearm, wrist, and hand. We also discuss and illustrate our experience with lacerations, crush injuries, and muscle hernias in the forearm.

Radiographic Outcomes of Lisfranc Injuries Treated with a Suture Button Device

Category:

Trauma

Introduction/Purpose:

Disagreement exists as to the optimal treatment of Lisfranc injuries of the midfoot. Some authors suggest that primary arthrodesis should be the treatment of choice, whereas others are proponents of open reduction and internal fixation (ORIF). There has been increased interest in using a suture button device in lieu of traditional screw fixation, which in general must be removed. Biomechanical studies comparing screw fixation with suture button devices have had conflicting results. The aim of this study was evaluate patients with Lisfranc injuries treated with a suture button radiographically over time.

Methods:

43 patients with a Lisfranc injury treated surgically with a suture button device (Tightrope, Arthrex, Naples, FL) by two fellowship-trained orthopaedic foot and ankle surgeons were identified via a database query for the CPT code 28615 (open treatment of tarsometatarsal joint dislocation). The distances between the first and second metatarsal (M1-M2) and between the medial cuneiform and the second metatarsal (C1-M2) were measured on anterior-posterior (AP) radiographs using standardized, readily-identifiable landmarks. All measurements were performed independently by two musculoskeletal radiologists and were averaged. Accuracy of reduction was assessed by comparing measurements on weightbearing AP radiographs of the uninjured foot with the operatively-treated foot at 6 weeks postoperatively, also weightbearing. Maintenance of reduction over time was assessed by comparing measurements at 6 weeks postoperatively with measurements taken at the time of last follow-up. Measurements were also compared between the injured foot preoperatively and at time of last follow-up. T-tests using SPSS software were performed for statistical analysis in comparing the means of the different measurements.

Results:

Average radiographic follow-up was 35.8 weeks (6.1-178.9). Accuracy of reduction was less than 1 mm for both the M1- M2 and C1-M2 measurements when comparing the uninjured foot with the surgically-treated foot at 6 weeks postoperatively; this difference was significant in the C1-M2 measurement (difference = 0.77 mm, p=0.032), but not in the M1-M2 measurement (difference = 0.44 mm, p=0.190). There was no significant difference in measurements between 6 weeks postoperatively and at final follow-up (difference in M1-M2=0.22 mm, p=0.435; difference in C1-M2=0.27 mm, p=0.352). There was significant improvement in the measured distances of the injured foot preoperatively to that measured at final follow-up in the C1-M2 measurements (difference in C1-M2=-0.82mm, p=0.021), while the difference in the M1-M2 measurements approached significance (difference in M1-M2= -0.56 mm, p=0.067).

Conclusion:

To our knowledge, this is the first and largest series assessing radiographic outcomes of Lisfranc injuries in patients treated surgically with a suture button device. After open reduction of Lisfranc injuries, the suture-button device used in this study appears to adequately maintain this reduction over time while patients are weightbearing and back to full activity. We therefore conclude that based on our results, the suture button appears to present an effective alternative to traditional screw fixation for treatment of Lisfranc injuries.

Structural differences and functional similarities between two sugar maple (Acer saccharum) stands

The spatially inexplicit or functional multilayer models used to predict canopy transpiration or photosynthesis are based on the assumption that closed stands show less functional variability than structural variability, because foliage tends to arrange itself in space to optimize the capture of light. To validate this assumption, we compared the structural and functional properties, and the measured and modeled transpiration fluxes of two sugar maple (Acer saccharum Marsh.) stands of comparable leaf mass but differing in height and diameter distributions. One stand was characterized by a well-developed single-layer canopy, whereas the other stand had a multilayered canopy and a stem diameter distribution of the classical inverse-J shape. Stand differences in height and diameter distribution, and canopy gap fraction, were highly significant. There were minor but significant differences in leaf mass and leaf mass per unit leaf area (LMA) distributions. We found no differences in tree-level relationships between basal area and either transpiration flux or sapwood area. We compared measurements of stand transpiration with transpiration estimates obtained from a multilayer gas exchange model, in which only the nonspatial inputs, leaf area index and LMA frequency distribution described stand structure. For both stands, modeled values of daily transpiration closely followed measured values (r(2) = 0.94). These results support use of the nonspatially explicit approach to estimating canopy gas exchange, especially if the intent is to scale-up to larger portions of the landscape.

Use of continuous ambulatory infusions of concentrated subcutaneous (s.q.) hydromorphone versus intravenous (i.v.) morphine: cost implications for palliative care

Health care practitioners are increasingly under pressure to curtail spending while trying to deliver excellent patient care. These issues are also affecting palliative care, particularly now that palliative care programs are expanding. A comparison of cost-effectiveness and feasibility of using continuous subcutaneous (s.q.) ambulatory infusion of hydromorphone versus intravenous (i.v.) ambulatory morphine is illustrated in this study. With the high doses of morphine required in chronic cancer pain, the use of subcutaneous morphine is not feasible due to the volume of solution required to be delivered. Hydromorphone can be prepared in concentrated solutions enabling it to be delivered by the subcutaneous route. Morphine stability data are available. However, hydromorphone stability has only been verified for seven days; thus, stability data were needed post-seven days. Concentrations of 10 mg/ml, 20 mg/ml, 50 mg/ml, and 100 mg/ml, in 0.9 percent normal saline or dextrose 5 percent water, were analyzed via high-performance liquid chromatography (HPLC) at seven and 28 days. Cost comparisons of supplies and associated costs with subcutaneous versus intravenous solutions were obtained. Hydromorphone was found to be stable for 28 days in both dilutants. Cost analysis of a hydromorphone 28-day supply resulted in substantial savings over the equivalent costs of morphine infusions.

Superior labral anteroposterior lesions: MR arthrography with arm traction

OBJECTIVE

The purpose of this study was to investigate the efficacy of arm traction combined with MR arthrography in the evaluation of superior labral anteroposterior (SLAP) lesions.

MATERIALS AND METHODS

Cadaveric shoulders were studied with a 1.5-T MR imaging unit with the arm externally rotated. Fifteen milliliters of a gadolinium-containing contrast agent were injected into the glenohumeral joint. Twenty-four sets of images of cadaveric joints were evaluated independently by two observers. These sets consisted of MR arthrographic images obtained with traction (applied to the wrist using 1- to 3-kg weights) and without traction in five shoulders in which SLAP lesions had been excluded arthroscopically or by cadaveric sectioning; and MR arthrographic images obtained with and without traction in seven shoulders in which various types of SLAP lesions had been created arthroscopically and later confirmed by cadaveric sectioning.

RESULTS

Analysis of the data indicated that MR arthrography in combination with arm traction and external rotation improved diagnostic accuracy with regard to identification and categorization of SLAP lesions when compared with studies made without traction.

CONCLUSION

The combination of MR arthrography and arm traction with the shoulder in external rotation provides a more effective approach for detection of SLAP lesions than does similar MR arthrography performed without arm traction.

Changes in ABA and gene expression in cold-acclimated sugar maple

To determine if cold acclimation of sugar maple (Acer saccharum Marsh.) is associated with specific changes in gene expression under natural hardening conditions, we compared bud and root translatable mRNAs of potted maple seedlings after cold acclimation under natural conditions and following spring dehardening. Cold-hardened roots and buds were sampled in January when tissues reached their maximum hardiness. Freezing tolerance, expressed as the lethal temperature for 50% of the tissues (LT(50)), was estimated at -17 degrees C for roots, and at lower than -36 degrees C for buds. Approximately ten transcripts were specifically synthesized in cold-acclimated buds, or were more abundant in cold-acclimated buds than in unhardened buds. Cold hardening was also associated with changes in translation. At least five translation products were more abundant in cold-acclimated buds and roots compared with unhardened tissues. Abscisic acid (ABA) concentration increased approximately tenfold in the xylem sap following winter acclimation, and the maximum concentration was reached just before maximal acclimation. We discuss the potential involvement of ABA in the observed modification of gene expression during cold hardening.

Effects of soil freezing and drought stress on abscisic acid content of sugar maple sap and leaves

In 1991 and 1992, mature maple trees (Acer saccharum Marsh.) were freeze-stressed or drought-stressed by preventing precipitation (snow or rain) from reaching the forest floor under selected trees. Lack of snow cover caused a decrease in soil temperature to well below 0 degrees C from December to April and a lowering of the soil water content to 10%. The abscisic acid (ABA) concentration in the spring sap of deep-soil frost-stressed trees was significantly higher than in control or drought-stressed trees. The increase in ABA concentration in the xylem sap in the spring of 1991 and 1992 preceded symptoms of canopy decline and a decrease in leaf area that were observed during the summers of 1991 and 1992. These results suggest a role for ABA in root-to-shoot communication in response to environmental stress. The largest differences in ABA concentration induced by the treatments was found in sap collected at the end of sap flow. The increase in ABA concentration in spring sap at the end of the sap flow could be used as an early indicator of stress suffered by trees during the winter. Not only did the increase in ABA concentration occur before any visible symptoms of tree decline appeared, but the trees that showed the most evident decline had the highest ABA concentrations in the spring sap. Leaf ABA concentration was not a good indicator of induced stress.