A publicly available deep learning model and dataset for segmentation of breast, fibroglandular tissue, and vessels in breast MRI.

Breast density, or the amount of fibroglandular tissue (FGT) relative to the overall breast volume, increases the risk of developing breast cancer. Although previous studies have utilized deep learning to assess breast density, the limited public availability of data and quantitative tools hinders the development of better assessment tools. Our objective was to (1) create and share a large dataset of pixel-wise annotations according to well-defined criteria, and (2) develop, evaluate, and share an automated segmentation method for breast, FGT, and blood vessels using convolutional neural networks. We used the Duke Breast Cancer MRI dataset to randomly select 100 MRI studies and manually annotated the breast, FGT, and blood vessels for each study. Model performance was evaluated using the Dice similarity coefficient (DSC). The model achieved DSC values of 0.92 for breast, 0.86 for FGT, and 0.65 for blood vessels on the test set. The correlation between our model's predicted breast density and the manually generated masks was 0.95. The correlation between the predicted breast density and qualitative radiologist assessment was 0.75. Our automated models can accurately segment breast, FGT, and blood vessels using pre-contrast breast MRI data. The data and the models were made publicly available.

Automated neonatal nnU-Net brain MRI extractor trained on a large multi-institutional dataset.

Brain extraction, or skull-stripping, is an essential data preprocessing step for machine learning approaches to brain MRI analysis. Currently, there are limited extraction algorithms for the neonatal brain. We aim to adapt an established deep learning algorithm for the automatic segmentation of neonatal brains from MRI, trained on a large multi-institutional dataset for improved generalizability across image acquisition parameters. Our model, ANUBEX (automated neonatal nnU-Net brain MRI extractor), was designed using nnU-Net and was trained on a subset of participants (N = 433) enrolled in the High-dose Erythropoietin for Asphyxia and Encephalopathy (HEAL) study. We compared the performance of our model to five publicly available models (BET, BSE, CABINET, iBEATv2, ROBEX) across conventional and machine learning methods, tested on two public datasets (NIH and dHCP). We found that our model had a significantly higher Dice score on the aggregate of both data sets and comparable or significantly higher Dice scores on the NIH (low-resolution) and dHCP (high-resolution) datasets independently. ANUBEX performs similarly when trained on sequence-agnostic or motion-degraded MRI, but slightly worse on preterm brains. In conclusion, we created an automatic deep learning-based neonatal brain extraction algorithm that demonstrates accurate performance with both high- and low-resolution MRIs with fast computation time.

MRI-based Deep Learning Assessment of Amyloid, Tau, and Neurodegeneration Biomarker Status across the Alzheimer Disease Spectrum.

Background PET can be used for amyloid-tau-neurodegeneration (ATN) classification in Alzheimer disease, but incurs considerable cost and exposure to ionizing radiation. MRI currently has limited use in characterizing ATN status. Deep learning techniques can detect complex patterns in MRI data and have potential for noninvasive characterization of ATN status. Purpose To use deep learning to predict PET-determined ATN biomarker status using MRI and readily available diagnostic data. Materials and Methods MRI and PET data were retrospectively collected from the Alzheimer's Disease Imaging Initiative. PET scans were paired with MRI scans acquired within 30 days, from August 2005 to September 2020. Pairs were randomly split into subsets as follows: 70% for training, 10% for validation, and 20% for final testing. A bimodal Gaussian mixture model was used to threshold PET scans into positive and negative labels. MRI data were fed into a convolutional neural network to generate imaging features. These features were combined in a logistic regression model with patient demographics, APOE gene status, cognitive scores, hippocampal volumes, and clinical diagnoses to classify each ATN biomarker component as positive or negative. Area under the receiver operating characteristic curve (AUC) analysis was used for model evaluation. Feature importance was derived from model coefficients and gradients. Results There were 2099 amyloid (mean patient age, 75 years ± 10 [SD]; 1110 male), 557 tau (mean patient age, 75 years ± 7; 280 male), and 2768 FDG PET (mean patient age, 75 years ± 7; 1645 male) and MRI pairs. Model AUCs for the test set were as follows: amyloid, 0.79 (95% CI: 0.74, 0.83); tau, 0.73 (95% CI: 0.58, 0.86); and neurodegeneration, 0.86 (95% CI: 0.83, 0.89). Within the networks, high gradients were present in key temporal, parietal, frontal, and occipital cortical regions. Model coefficients for cognitive scores, hippocampal volumes, and APOE status were highest. Conclusion A deep learning algorithm predicted each component of PET-determined ATN status with acceptable to excellent efficacy using MRI and other available diagnostic data. © RSNA, 2023 Supplemental material is available for this article.

Cooperativity of silanol defect chemistry in zeolites.

Deboronation treatment of zeolite B-SSZ-55 can generate vacancy defects consisting of four silanol groups (silanol nests). However, 1H solid-state NMR spectroscopy indicates the prevalence of two silanol groups (silanol dyads) instead of four silanol groups. Such silanol dyads must be formed by the silanol condensation of two silanol groups at the silanol nests. Yet, the exact mechanism of this condensation and detailed structure of the silanol defect are not known. Here, the structure and formation mechanism of silanol dyads in the SSZ-55 zeolite have been investigated by both cluster and periodic density functional theory calculations. The calculated 1H NMR chemical shifts agree with the experimental values, showing that the silanol dyads are indeed commonly present at the vacancies and the vacancy density plays a role in the relaxation of the zeolite framework. The nature (size) of the silanol clusters influences their acidity.

Using a bimanual lever-driven wheelchair for arm movement practice early after stroke: A pilot, randomized, controlled, single-blind trial.

OBJECTIVE: Many patients with subacute stroke rely on the nonparetic arm and leg to propel manual wheelchairs. We designed a bimanual, lever-driven wheelchair (LARA) to promote overground mobility and hemiparetic arm exercise. This study measured the feasibility of using LARA to increase arm movement, achieve mobility, and improve arm motor recovery (clinicaltrials.gov/ct2/show/NCT02830893). DESIGN: Randomized, assessor-blind, controlled trial. SETTING: Two inpatient rehabilitation facilities. SUBJECTS: Nineteen patients with subacute stroke (1 week to 2 months post-stroke) received 30 minutes extra arm movement practice daily, while admitted to inpatient rehabilitation (n = 10) or before enrollment in outpatient therapy (n = 9). INTERVENTIONS: Patients were randomized to train with the LARA wheelchair (n = 11) or conventional exercises with a rehabilitation therapist (n = 8). MAIN MEASURES: Number of arm movements per training session; overground speed; Upper Extremity Fugl-Meyer score at three-month follow-up. RESULTS: Participants who trained with LARA completed 254 (median) arm movements with the paretic arm each session. For three participants, LARA enabled wheelchair mobility at practical indoor speeds (0.15-0.30 m/s). Fugl-Meyer score increased 19 ± 13 points for patients who trained with LARA compared to 14 ± 7 points with conventional exercises (P = 0.32). Secondary measures including shoulder pain and increased tone did not differ between groups. Mixed model analysis found significant interaction between LARA training and treatment duration (P = 0.037), informing power analysis for future investigation. CONCLUSIONS: Practising arm movement with a lever-driven wheelchair is a feasible method for increasing arm movement early after stroke. It enabled wheelchair mobility for a subset of patients and shows potential for improving arm motor recovery.

Chest tube output, duration, and length of stay are similar for pneumothorax and hemothorax seen only on computed tomography vs. chest radiograph.

PURPOSE: Whole-body computed tomography (CT) for blunt trauma patients is common. Chest CT (CCT) identifies "occult" pneumo- (PTX) and hemothorax (HTX) not seen on chest radiograph (CXR), one-third of whom get chest tubes, while CXR identifies "non-occult" PTX/HTX. To assess chest tube value for occult injury vs. expectant management, we compared output, duration, and length of stay (LOS) for chest tubes placed for occult vs. non-occult (CXR-visible) injury. METHODS: We compared chest tube output and duration, and patient length of stay for occult vs. non-occult PTX/HTX. This was a retrospective analysis of 5451 consecutive Level I blunt trauma patients, from 2010 to 2013. RESULTS: Of these blunt trauma patients, 402 patients (7.4%) had PTX, HTX or both, and both CXR and CCT. One third (n = 136, 33.8%) had chest tubes placed in 163 hemithoraces (27 bilateral). Non-occult chest tube output for all patients was 1558 ± 1919 cc (n = 54), similar to occult at 1123 ± 1076 cc (n = 109, p = 0.126). Outputs were similar for HTX-only patients, with non-occult (n = 34) at 1917 ± 2130 cc, vs. occult (n = 54) at 1449 ± 1131 cc (p = 0.24). Chest tube duration for all patients was 6.3 ± 4.9 days for non-occult vs. 5.0 ± 3.3 for occult (p = 0.096). LOS was similar between all occult injury patients (n = 46) and non-occult (n = 90, 17.0 ± 15.8 vs. 13.7 ± 11.9 days, p = 0.23). CONCLUSION: Mature clinical judgment may dictate which patients need chest tubes and explain the similarity between groups.

Development and Evaluation of MOVit: An Exercise-Enabling Interface for Driving a Powered Wheelchair.

Powered wheelchair users can experience negative health effects from reduced physical activity. If a user could exercise by driving the chair, it might improve fitness. This paper presents the development of MOVit, an exercise-enabling, wheelchair driving interface. The design goal of MOVit was that users cyclically move their arms to drive the chair, thereby providing a light level of exercise while driving. MOVit supports this arm movement with custom mobile arm supports that also serve as the sensors that provide controller inputs. Here, we first quantified how increasing the frequency and amplitude of arm movement increase oxygen consumption and heart rate. Then, we evaluated two novel control methods for driving by moving the arm supports. Participants without impairment ( N = 24 ) were randomized to one of the two methods, or conventional joystick control, and performed driving tests over two days on a simulator and test course. Our results indicate that driving speed and accuracy were significantly lowered with the exercise-enabling methods compared to joystick control (ANOVA, ), but the decreases were small (speed was ~0.1 m/s less and course tracking error ~1 cm greater). These results show, for the first time, the feasibility of exercising while driving a powered wheelchair.

Insight into on-wafer crystallization of pure-silica-zeolite films through nutrient replenishment.

Tetraethylorthosilicate (TEOS) is added to a pure-silica-zeolite MEL nanoparticle suspension and the mixture is subsequently used for preparing spin-on low-dielectric constant (low-k) films. The films are then characterized by ellipsometric porosimetry, transmission electron microscopy (TEM), and nanoindentation. Investigation into the film microstructure indicates that the addition of TEOS significantly increases the fraction of the crystalline domains, decreases the inter-crystal mesopore size, and narrows the pore size distribution. Films with 12% TEOS loading have a mean pore size distribution centered at 3.5 nm (diameter) with a full width at half maximum (fwhm) of 0.8 nm, while those with no TEOS have a distribution at 11.1 nm and fwhm of 7.9 nm. At 12% TEOS loading, the reduced modulus and hardness are 11.0 and 1.12 GPa, respectively; without TEOS, the values are 6.4 and 0.57 GPa.

Zeolite thin films: from computer chips to space stations.

Zeolites are a class of crystalline oxides that have uniform and molecular-sized pores (3-12 A in diameter). Although natural zeolites were first discovered in 1756, significant commercial development did not begin until the 1950s when synthetic zeolites with high purity and controlled chemical composition became available. Since then, major commercial applications of zeolites have been limited to catalysis, adsorption, and ion exchange, all using zeolites in powder form. Although researchers have widely investigated zeolite thin films within the last 15 years, most of these studies were motivated by the potential application of these materials as separation membranes and membrane reactors. In the last decade, we have recognized and demonstrated that zeolite thin films can have new, diverse, and economically significant applications that others had not previously considered. In this Account, we highlight our work on the development of zeolite thin films as low-dielectric constant (low-k) insulators for future generation computer chips, environmentally benign corrosion-resistant coatings for aerospace alloys, and hydrophilic and microbiocidal coatings for gravity-independent water separation in space stations. Although these three applications might not seem directly related, they all rely on the ability to fine-tune important macroscopic properties of zeolites by changing their ratio of silicon to aluminum. For example, pure-silica zeolites (PSZs, Si/Al = infinity) are hydrophobic, acid stable, and have no ion exchange capacity, while low-silica zeolites (LSZs, Si/Al < 2) are hydrophilic, acid soluble, and have a high ion exchange capacity. These new thin films also take advantage of some unique properties of zeolites that have not been exploited before, such as a higher elastic modulus, hardness, and heat conductivity than those of amorphous porous silicas, and microbiocidal capabilities derived from their ion exchange capacities. Finally, we briefly discuss our more recent work on polycrystalline zeolite thin films as promising biocompatible coatings and environmentally benign wear-resistant and antifouling coatings. When zeolites are incorporated into polymer thin films in the form of nanocrystals, we also show that the resultant composite membranes can significantly improve the performance of reverse osmosis membranes for sea water desalination and proton exchange membrane fuel cells. These diverse applications of zeolites have the potential to initiate new industries while revolutionizing existing ones with a potential economic impact that could extend into the hundreds of billions of dollars. We have licensed several of these inventions to companies with millions of dollars invested in their commercial development. We expect that other related technologies will be licensed in the near future.

On-wafer crystallization of ultralow-kappa pure silica zeolite films.

A higher goal: An on-wafer crystallization process to prepare pure silica zeolite (PSZ) MEL-type films that is superior to the previously used hydrothermal process is reported. These striation-free MEL-type films (right, see picture) outperform the traditional spin-on films (left) in terms of the kappa value, mechanical properties, surface roughness, mesopore size, and size distribution.

Hydrofluoric-acid-resistant and hydrophobic pure-silica-zeolite MEL low-dielectric-constant films.

A new technique for the silylation of pure-silica-zeolite MEL low-k films has been developed in which the spin-on films are calcined directly in trimethylchlorosilane or 1,1,1,3,3,3-hexamethyldisilazane (HMDS) in order to protect the films against corrosive wet etch chemicals and ambient moisture adsorption. In an alternative procedure, HMDS is also added to the zeolite suspension before film preparation. Fourier transform infrared spectroscopy, X-ray photoelectron spectroscopy, water-soak tests, and HF etch tests are performed to characterize the films. The dielectric constant is as low as 1.51, and the films resist HF attack up to 5.5 min. These properties are highly desirable by the semiconductor industry for next-generation microprocessors.

Pure-silica-zeolite MEL low-k films from nanoparticle suspensions.

Following our previous works on pure-silica-zeolite (PSZ) MFI, in this study we explore PSZ MEL as a new option for low-k dielectric films. Our motivation has been to increase the microporosity of the spin-on films by moving to structures with a framework density (FD) lower than MFI. Nanoparticle PSZ MEL suspensions were synthesized by a two-stage method that allowed the yield of nanocrystals to be significantly enhanced, while the zeolite nanocrystals remain small. For the first time zeolite nanocrystals of about 50 nm were synthesized with a yield as high as 57%. Nanoparticle suspensions with different particle sizes and crystallinities were spun on silicon wafers to prepare continuous thin films. An ultralow-k value as low as 1.5 was obtained with MEL nanoparticle suspension of high relative crystallinity. The surface roughness of the PSZ MEL film with high relative crystallinity is greatly improved (R(rms) approximately 5.6 nm) compared to MFI films with high relative crystallinity (R(rms) approximately 12 nm).