Search

Abstract While multiple factors impact disease, artificial intelligence (AI) studies in medicine often use small, non-diverse patient cohorts due to data sharing and privacy issues. Federated learning (FL) has emerged as a solution, enabling training across hospitals without direct data sharing. Here, we present FL-PedBrain, an FL platform for pediatric posterior fossa brain tumors, and evaluate its performance on a diverse, realistic, multi-center cohort. Pediatric brain tumors were targeted due to the scarcity of such datasets, even in tertiary care hospitals. Our platform orchestrates federated training for joint tumor classification and segmentation across 19 international sites. FL-PedBrain exhibits less than a 1.5% decrease in classification and a 3% reduction in segmentation performance compared to centralized data training. FL boosts segmentation performance by 20 to 30% on three external, out-of-network sites. Finally, we explore the sources of data heterogeneity and examine FL robustness in real-world scenarios with data imbalances.

Abstract LNA-containing oligonucleotides bind DNA more tightly than standard DNA, so they can interact with targeted sequences and affect multiple processes. When a desired DNA is present at low concentrations relative to nearly identical undesired DNAs, LNAs can block amplification of unwanted DNAs. Using a short rAAV and synthetic DNA sequence as a model, we studied the length, number, and positioning of LNA bases to improve blocker effectiveness. Oligonucleotides 18–24 bases long with LNAs at every other position were most effective. Highly degenerate targets were used to characterize the impact of mismatches on blocking. Mismatches at LNA ends had little impact on blocking activity. Single and double mismatches were tolerated with longer blockers, especially if the mismatches were near LNA ends. Shorter LNAs were more selective, with > 1 mismatch preventing effective blocking. Neither the strand to which a blocker bound nor the distance between the blocker and priming sites greatly impacted blocking efficiency. We used these findings to design blockers of wild-type DNA versus the single-base A1AT PiZ allele. Blockers are most specific when the mismatch is located away from the LNA 5′ end. Pairs of partially overlapping blockers on opposite strands with a centrally-located mismatch have maximal activity and specificity.

Abstract Background Osteoarthritis (OA) is a common and debilitating condition characterized by degeneration of hyaline cartilage. Currently, there is no treatment for OA that directly targets degradation of cartilage matrix. Capacitively coupled electric fields (CCEFs) represent a noninvasive and cost-effective treatment modality that can potentially restore articular cartilage homeostasis. Previous studies showed that stimulation of articular cartilage with CCEFs resulted in upregulation of anabolic factors and downregulation of catabolic factors. These studies didn’t explain the derivation of the CCEFs or verify their uniformity and field strength, so it’s possible that cartilage wasn’t exposed to uniform field strength. The present study aims to employ CCEFs with verified uniform field strength in two in-vitro models of OA to investigate its potential to preserve cartilage matrix and validate the results of the aforementioned studies. Methods Rabbit hyaline chondrocytes and full-thickness bovine articular cartilage explants were cultured in the absence or presence of CCEF and in the absence or presence of Interleukin1-B (IL-1B). Quantitative polymerase chain reaction (QPCR) was performed on chondrocytes to measure gene expression of ADAM-TS4, MMP3, MMP9, IL-6, TIMP1, and TIMP2. QPCR was performed on explants to measure gene expression of MMP3, Aggrecan, Collagen-2, and TIMP1. Aggrecan concentration in explants was measured with histology. Statistical analysis was performed using one-way analysis of variance and Tukey–Kramer multiple comparison test. Results The treatment of chondrocytes with IL-1B resulted in upregulated expression of ADAM-TS4, MMP3, MMP9, and IL-6, while simultaneous administration of IL-1B and CCEF led to a relative decrease in ADAM-TS4, MMP3, MMP9, and IL-6 expression and a relative increase in TIMP1 and TIMP2 expression. Application of IL-1B and CCEF to the explants resulted in decreased expression of MMP3 and increased expression of Aggrecan, Collagen-2, and TIMP1 when compared to application of IL-1B alone. Conclusion The data indicate that application of a CCEF with verified uniformity may result in upregulation of cartilage anabolic factors even in the presence of IL-1B while attenuating IL-1B induced upregulation of catabolic factors in both monolayer culture and whole tissue. These results demonstrate the potential of CCEFs to suppress the progression of OA and regenerate articular cartilage matrix.

Abstract Conventional neurostimulation systems for preclinical research can be bulky and invasive due to the need for batteries or wired interfaces. Emerging as a new neural interface technique, ultrasound-powered piezoelectric neural stimulators work by converting ultrasound energy to electrical charge for neural stimulation. In addition to the benefits of wireless powering and miniaturization leading to less traumatic surgery, piezoelectric neural stimulators can also exhibit prolonged operational lifetimes for a long-term stable neural interface, and show promise for clinical translation. As one of first steps to demonstrate the value of ultrasound-powered piezoelectric neural interface, Li et al. developed a piezoelectric stimulator to activate spinal cord neural circuits for locomotion restoration in a rat model with spinal cord injury (SCI) and compared its efficacy with conventional electrical stimulation (ES). From the point of view of materials science, neural engineering and microelectronics, we provide our commentary on the article, highlighting its importance and discussing the issues that remain to be addressed in future studies in the emerging field of ultrasound powered piezoelectric neurostimulation devices.

The search for extraterrestrial (alien) life is one of the greatest scientific quests, yet it raises fundamental questions about just what we should be looking for and how. We approach alien hunting from the perspective of an experimenter engaging in binary classification with some true- and confounding-positive probabilities (TPP and CPP). We derive the Bayes factor in such a framework between two competing hypotheses, which we use to classify experiments as either impotent, imperfect, or ideal. Similarly, the experimenter can be classified as dogmatic, biased, or agnostic. We show how the unbounded explanatory and evasion capability of aliens poses fundamental problems to experiments directly seeking aliens. Instead, we advocate framing the experiments as looking for that outside of known processes, which means the hypotheses we test do not directly concern aliens per se. To connect back to aliens requires a second level of model selection, for which we derive the final odds ratio in a Bayesian framework. This reveals that it is fundamentally impossible to ever establish alien life at some threshold odds ratio, ${{ \mathcal O }}_{\mathrm{crit}}$ , unless we deem the prior probability that some as-yet-undiscovered natural process could explain the event is less than ${(1+{{ \mathcal O }}_{\mathrm{crit}})}^{-1}$ . This elucidates how alien hunters need to carefully consider the challenging problem of how probable unknown unknowns are, such as new physics or chemistry, and how it is arguably most fruitful to focus on experiments for which our domain knowledge is thought to be asymptotically complete.

Agrowing avenue for determining the prevalence of life beyond Earth is to search for “technosignatures” from extraterrestrial intelligences/agents. Technosignatures require significant energy to be visible across interstellar space and thus intentional signals might be concentrated in frequency, in time, or in space, to be found in mutually obvious places. Therefore, it could be advantageous to search for technosignatures in parts of parameter space that are mutually derivable to an observer on Earth and a distant transmitter. In this work, we used the L -band (1.1–1.9 GHz) receiver on the Robert C. Byrd Green Bank Telescope to perform the first technosignature search presynchronized with exoplanet transits, covering 12 Kepler systems. We used the Breakthrough Listen turboSETI pipeline to flag narrowband hits (∼3 Hz) using a maximum drift rate of ±614.4 Hz s ^−1 and a signal-to-noise threshold of 5—the pipeline returned ∼3.4 × 10 ^5 apparently-localized features. Visual inspection by a team of citizen scientists ruled out 99.6% of them. Further analysis found two signals of interest that warrant follow up, but no technosignatures. If the signals of interest are not redetected in future work, it will imply that the 12 targets in the search are not producing transit-aligned signals from 1.1 to 1.9 GHz with transmitter powers >60 times that of the former Arecibo radar. This search debuts a range of innovative technosignature techniques: citizen science vetting of potential signals of interest, a sensitivity-aware search out to extremely high drift rates, a more flexible method of analyzing on-off cadences, and an extremely low signal-to-noise threshold.

Implanted vagus nerve stimulation (VNS) delivered concurrently with upper limb rehabilitation has been shown to improve arm function after stroke. Transcutaneous auricular VNS (taVNS) offers a non-invasive alternative to implanted VNS and may provide similar therapeutic benefit. There is much discussion about the optimal approach for combining VNS and physical therapy, as such we sought to determine whether taVNS administered during robotic training, specifically delivered during the premotor planning stage for arm extension movements, would confer additional motor improvement in patients with chronic stroke. Thirty-six patients with chronic, moderate-severe upper limb hemiparesis (>6 months; mean Upper Extremity Fugl-Meyer score = 25 ± 2, range 13–48), were randomized to receive 9 sessions (1 h in length, 3x/week for 3 weeks) of active (N = 18) or sham (N = 18) taVNS (500 ms bursts, frequency 30 Hz, pulse width 0.3 ms, max intensity 5 mA, ∼250 stimulated movements per session) delivered during robotic training. taVNS was triggered by the onset of a visual cue prior to center-out arm extension movements. Clinical assessments and surface electromyography (sEMG) measures of the biceps and triceps brachii were collected during separate test sessions. Significant motor improvements were measured for both the active and sham taVNS groups, and these improvements were robust at 3 month follow-up. Compared to the sham group, the active taVNS group showed a significant reduction in spasticity of the wrist and hand at discharge (Modified Tardieu Scale; taVNS = –8.94% vs. sham = + 2.97%, p < 0.05). The EMG results also demonstrated significantly increased variance for the bicep peak sEMG amplitude during extension for the active taVNS group compared to the sham group at discharge (active = 26.29% MVC ± 3.89, sham = 10.63% MVC ± 3.10, mean absolute change admission to discharge, p < 0.01), and at 3-month follow-up, the bicep peak sEMG amplitude was significantly reduced in the active taVNS group (P < 0.05). Thus, robot training improved the motor capacity of both groups, and taVNS, decreased spasticity. taVNS administered during premotor planning of movement may play a role in improving coordinated activation of the agonist-antagonist upper arm muscle groups by mitigating spasticity and increasing motor control following stroke.Clinical Trial Registration:www.ClinicalTrials.gov, identifier (NCT03592745).

Abstract Background Despite their growing popularity, the potential respiratory toxicity of electronic cigarettes (e-cigarettes) remains largely unknown. One potential aspect of e-cigarette toxicity is the effect of e-cigarette vapor on lung surfactant function. Lung surfactant is a mixture of lipids and proteins that lines the alveolar region. The surfactant layer reduces the surface tension of the alveolar fluid, thereby playing a crucial role in lung stability. Due to their small size, particulates in e-cigarette vapor can penetrate the deep lungs and come into contact with the lung surfactant. The current study sought to examine the potential adverse effects of e-cigarette vapor and conventional cigarette smoke on lung surfactant interfacial properties. Methods Infasurf®, a clinically used and commercially available calf lung surfactant extract, was used as lung surfactant model. Infasurf® films were spread on top of an aqueous subphase in a Langmuir trough with smoke particulates from conventional cigarettes or vapor from different flavors of e-cigarettes dispersed in the subphase. Surfactant interfacial properties were measured in real-time upon surface compression while surfactant lateral structure after exposure to smoke or vapor was examined using atomic force microscopy (AFM). Results E-cigarette vapor regardless of the dose and flavoring of the e-liquid did not affect surfactant interfacial properties. In contrast, smoke from conventional cigarettes had a drastic, dose-dependent effect on Infasurf® interfacial properties reducing the maximum surface pressure from 65.1 ± 0.2 mN/m to 46.1 ± 1.3 mN/m at the highest dose. Cigarette smoke and e-cigarette vapor both altered surfactant microstructure resulting in an increase in the area of lipid multilayers. Studies with individual smoke components revealed that tar was the smoke component most disruptive to surfactant function. Conclusions While both e-cigarette vapor and conventional cigarette smoke affect surfactant lateral structure, only cigarette smoke disrupts surfactant interfacial properties. The surfactant inhibitory compound in conventional cigarettes is tar, which is a product of burning and is thus absent in e-cigarette vapor.

Laboratory cyclic plate load tests are commonly used in the assessment of geosynthetic performance in pavement applications due to the repeatability of testing results and the smaller required testing areas than traditional Accelerated Pavement Testing facilities. While the objective of traditional plate load testing procedure is to closely replicate traffic conditions, the reality is that rolling wheel loads produce different stresses in pavement layers than traditional cyclic plate load tests. This two-fold study investigates the differences between the stress response of subgrade soil from a rolling wheel load (replicating rolling traffic conditions) and a unidirectional dynamic load (replicating traditional plate load test procedures) in order to obtain a more realistic stress response of pavement layers from rolling wheel traffic. Ultimately, results show that the testing specimens that experienced rolling wheel loading had an average of 17% higher pressure measurements in the top of the subgrade than vertically loaded (unidirectional dynamic load) specimens. The second segment of this study is used in conjunction with the first to analyze aggregate base material behavior when using a geosynthetic for reinforcement. The study aimed to determine the difference in the post-trafficked strength and stiffness of pavement foundation. A Dynamic Cone Penetrometer and Light Weight Deflectometer were utilized to determine material changes from this trafficking and revealed that all specimens that included a geosynthetic had a higher base stiffness and strength while the specimen with geotextile and geogrid in combination created the highest stiffness and strength after large-scale rolling wheel trafficking.

The gene encoding the Insulin-like Growth Factor 2 mRNA binding protein 2/IMP2 is amplified and overexpressed in many human cancers, accompanied by a poorer prognosis. Mice lacking IMP2 exhibit a longer lifespan and a reduced tumor burden at old age. Herein we show in a diverse array of human cancer cells that IMP2 overexpression stimulates and IMP2 elimination diminishes proliferation by 50–80%. In addition to its known ability to promote the abundance of Insulin-like Growth Factor 2/IGF2, we find that IMP2 strongly promotes IGF action, by binding and stabilizing the mRNA encoding the DNA binding protein HMGA1, a known oncogene. HMGA1 suppresses the abundance of IGF binding protein 2/IGFBP2 and Grb14, inhibitors of IGF action. IMP2 stabilization of HMGA1 mRNA plus IMP2 stimulated IGF2 production synergistically drive cancer cell proliferation and account for IMP2’s tumor promoting action. IMP2’s ability to promote proliferation and IGF action requires IMP2 phosphorylation by mTOR.

The American Academy of Pediatrics (AAP) has endorsed Critical Congenital Heart Disease (CCHD) screening using pulse oximetry nationwide, but, however, acknowledges that altitude may impact failure rates and alternative algorithms may be required at high altitudes. We therefore evaluated a modified screening protocol at an altitude of 6200 feet with the hypothesis that modifications could decrease failure rates. We evaluated 2001 well, newborn infants ≥35 weeks gestation using a modified protocol, which included a lower saturation cutoff for the first screen (85% instead of the AAP recommended 90%) and an oxygen hood intervention between the first two screens. Using our modified screening algorithm, we found a 0.3% failure rate, which was similar to the 0.2% sea-level rate and statistically different from the 1.1% rate identified in a recent study at similar altitude. Had the AAP protocol been used, the failure rate would have increased to 0.8%, which is similar to prior reports near this altitude. Echocardiograms were performed on failing newborns with no CCHD identified. A Birth Defects Registry Database review demonstrated one newborn with CCHD was missed after meeting AAP passing criteria. Overall, this study demonstrates that an alternative algorithm can be implemented at moderate altitude with decreased failure rate and comparable false negative rate.