Tyler O'Neal, Staff Editor PROCESSORS

Japanese researchers use Bayesian U-net architecture on CT images for personalized modeling of forces, stresses on muscles, bones

Personalized medicine has stirred the imagination of drugs and therapies that are individually tailored to patients. In the future, there will no longer be a need to worry about side effects, and patients will be screened to identify which treatment will be most effective for them, instead of for the average population. Deep learning is one tool that will be key to realizing personalized medicine. A new study by Japanese researchers describes a new deep learning tool that will advance personalized medicine for musculoskeletal diseases. The tool can segment individual muscles for a comprehensive model of the musculoskeletal system, which is expected to advance personalized biomechanics.

Accurate measurements of the musculoskeletal system can have a tremendous impact for the extremely ill, like those suffering from ALS or other severe forms of atrophy, to influence the design of rehabilitation devices, and for the extraordinarily gifted, like high-performance athletes who want to take their game to the next level. These measurements come from computed tomography or other imaging modality with which researchers build computer models to study the forces and stresses on muscles and bones. 

CAPTION Muscle segmentation from CT volume using Bayesian U-net. Skin mask is obtained using conventional U-net during the first stage, and then Bayesian U-net is utilized to assign a muscle label as well as uncertainty to each pixel. CREDIT Yoshinobu Sato
CAPTION Muscle segmentation from CT volume using Bayesian U-net. Skin mask is obtained using conventional U-net during the first stage, and then Bayesian U-net is utilized to assign a muscle label as well as uncertainty to each pixel. CREDIT Yoshinobu Sato

"Once we have the CT images, we need to segment the individual muscles for building our model," explains Professor Yoshinobu Sato of Nara Institute of Science and Technology (NAIST), Japan, who led the study. {module In-article}

"However, this segmentation was time-consuming and depended on expert-knowledge. The figure shows an overview of our system. We used deep learning to automate the segmentation of individual muscles to generate a musculoskeletal model that is personalized to the patient," he continues.

While it is normal to expect a physician to evaluate the images, this adds a level of subjectivity to the diagnosis. The system is especially beneficial for patients in remote areas with limited access to expert orthopedic surgeons, for decisions based on a more quantitative interpretation should improve outcomes.

The method depends on Bayesian U-net architecture.

"U-Net is a deep learning framework based on a fully convolutional neural network for the precise segmentation of images. Our colleague, Dr. Yuta Hiasa, extended U-net by combining Bayesian inference to formulate Bayesian U-net, in which uncertainty is associated with segmentation results," says Sato.

The challenge in segmenting individual muscles is the low contrast of the imaging at border regions of neighboring muscles. To test their system, the researchers examined 19 muscles in the thigh and hips. Bayesian U-Net had better segmentation accuracy than other methods including the hierarchical multi-atlas method, which is viewed as state-of-the-art, and did so while reducing the time to train and validate the system by a surgeon.

"Some pixels in the images had high uncertainty. It was these pixels that especially need confirmation by surgeons," notes Sato.

The researchers thus defined an uncertainty threshold to identify which pixels required human verification.

"Bayesian U-Net learned the musculoskeletal anatomy to create segmentations that would have been created by experts with high fidelity and our collaborator orthopedic surgeon, Prof. Nobuhiko Sugano of Osaka University Hospital, is quite satisfied with this achievement," says Sato.

HIV drug stops Zika infection, strategy could halt infections caused by related viruses

Tyler O'Neal, Staff Editor PROCESSORS

A research team at Lewis Katz School of Medicine at Temple University used computational biology

Like an adjustable wrench that becomes the "go-to" tool because it is effective and can be used for a variety of purposes, an existing drug that can be adapted to halt the replication of different viruses would greatly expedite the treatment of different infectious diseases. Such a strategy would prevent thousands of deaths each year from diseases like dengue and Ebola, but whether it can be done has been unclear. Now, in new work, researchers at the Lewis Katz School of Medicine at Temple University (LKSOM) show that repurposing an existing drug to treat viral diseases is, in fact, possible - potentially bypassing the decades needed to develop such a broad-spectrum drug from scratch. {module In-article}

In a new study published in the journal Molecular Therapy, the Temple researchers report that a drug used in the treatment of HIV also suppresses Zika virus infection. In cell and animal models, they show that the drug, called rilpivirine, stops the Zika virus by targeting enzymes that both HIV and Zika virus depends on for their replication. These enzymes occur in other viruses closely related to Zika, including the viruses that cause dengue, yellow fever, West Nile fever, and hepatitis C.

"HIV and Zika virus are distinct types of RNA viruses," explained Kamel Khalili, Ph.D., Laura H. Carnell Professor and Chair of the Department of Neuroscience, Director of the Center for Neurovirology, and Director of the Comprehensive NeuroAIDS Center at LKSOM. "By discovering that rilpivirine blocks Zika virus replication by binding to an RNA polymerase enzyme common to a family of RNA viruses, we've opened the way to potentially being able to treat multiple RNA virus infections using the same strategy."

Dr. Khalili, a senior investigator on the new study, attributed the breakthrough work to a productive collaboration with Temple University colleagues, including Dr. Michael L. Klein, FRS, Laura H. Carnell Professor of Science and Dean of the College of Science and Technology at Temple; and Ilker K. Sariyer, DVM, PhD, and Jennifer Gordon, PhD, Associate Professors of Neuroscience at Temple's Center for Neurovirology.

Historically rare and isolated to parts of Africa and Asia, the Zika virus is now present throughout the Americas and occurs in multiple other regions of the world. It has attracted increasing attention in recent years, owing to its damaging effects on the brain and nervous system. The virus is transmitted to humans by mosquitoes. Once in the body, it infects cells and replicates, typically taking up residence in cells in neural tissues. In severe cases, Zika virus infection can cause an autoimmune condition known as Guillain-Barré syndrome, which culminates in muscle paralysis. Infants born to mothers infected during pregnancy may experience delays in neurological development and may be affected by microcephaly (abnormal smallness of the head).

To replicate inside cells, the Zika virus requires an enzyme called non-structural protein 5 RNA-dependent RNA polymerase (NS5 RdRp). In the new study, Dr. Sariyer showed that rilpivirine suppresses the Zika virus infection in cells by blocking viral replication. Using structural biology and computational studies, Eleonora Gianti, Ph.D., a research assistant professor in Dr. Klein's laboratory, was able to show that rilpivirine prevents viral replication by binding specifically to the NS5 domain.

Dr. Gordon's team carried out experiments in mice, in which animals were infected with Zika virus through their footpads, similar to the way a person becomes infected through a mosquito bite. Mice that become infected with the Zika virus normally become very sick within about a week and eventually die. "We found, however, that when treated with rilpivirine, the animals survived," Dr. Gordon said. "Our conclusion is that rilpivirine disrupted the virus's usual course of infection."

Rilpivirine is one of several non-nucleoside reverse transcriptase inhibitor (NNRTI) drugs that have been developed for the treatment of HIV infection. Experiments in which the Temple researchers tested two other NNRTIs in Zika-infected cells revealed similar effects on viral replication, with the drugs specifically inhibiting NS5 activity.

"We now have a clear path forward," Dr. Khalili said. "We have a starting point from which we can find ways to make these drugs even more potent and more effective against flaviviruses." The researchers plan to soon step up their studies to develop ways to improve the effectiveness of NNRTIs in blocking infection with the Zika virus and other flaviviruses.

Epidemics involving flavivirus infections, particularly HIV, Zika, dengue, and hepatitis C, frequently overlap geographically and temporally. "The potential applications of this work are huge," Dr. Klein added.

UCSC astronomers perform supercomputer simulations to explain giant exoplanets with eccentric, close-in orbits

Tyler O'Neal, Staff Editor PROCESSORS

A giant-impacts phase in the evolution of planetary systems can explain the observations of close-in giant planets with eccentric orbits

As planetary systems evolve, gravitational interactions between planets can fling some of them into eccentric elliptical orbits around the host star, or even out of the system altogether. Smaller planets should be more susceptible to this gravitational scattering, yet many gas giant exoplanets have been observed with eccentric orbits very different from the roughly circular orbits of the planets in our own solar system.

Surprisingly, the planets with the highest masses tend to be those with the highest eccentricities, even though the inertia of a larger mass should make it harder to budge from its initial orbit. This counter-intuitive observation prompted astronomers at UC Santa Cruz to explore the evolution of planetary systems using supercomputer simulations. Their results, reported in a paper published in Astrophysical Journal Letters, suggest a crucial role for a giant-impacts phase in the evolution of high-mass planetary systems, leading to the collisional growth of multiple giant planets with close-in orbits. This artist's concept illustrates the collision of two rocky planets. A new study proposes a scenario in which collisions between gas giant planets can lead to mergers and the formation of high-mass gas giants with close-in orbits. (Image credit: NASA/JPL-Caltech){module In-article} 

"A giant planet is not as easily scattered into an eccentric orbit as a smaller planet, but if there are multiple giant planets close to the host star, their gravitational interactions are more likely scatter them into eccentric orbits," explained first author Renata Frelikh, a graduate student in astronomy and astrophysics at UC Santa Cruz.

Frelikh performed hundreds of simulations of planetary systems, starting each one with 10 planets in circular orbits and varying the initial total mass of the system and the masses of individual planets. As the systems evolved for 20 million simulated years, dynamical instabilities led to collisions and mergers to form larger planets as well as gravitational interactions that ejected some planets and scattered others into eccentric orbits.

Analyzing the results of these simulations collectively, the researchers found that the planetary systems with the most initial total mass-produced the biggest planets and the planets with the highest eccentricities.

"Our model naturally explains the counter-intuitive correlation of mass and eccentricity," Frelikh said.

Coauthor Ruth Murray-Clay, the Gunderson professor of theoretical astrophysics at UC Santa Cruz, said the only non-standard assumption in their model is that there can be several gas giant planets in the inner part of a planetary system. "If you make that assumption, all the other behavior follows," she said.

According to the classic model of planet formation, based on our own solar system, there is not enough material in the inner part of the protoplanetary disk around a star to make gas giant planets, so only small rocky planets form in the inner part of the system and giant planets from farther out. Yet astronomers have detected many gas giants orbiting close to their host stars. Because they are relatively easy to detect, these "hot Jupiters" accounted for the majority of early exoplanet discoveries, but they may be an uncommon outcome of planet formation.

"This may be an unusual process," Murray-Clay said. "We're suggesting that it is more likely to happen when the initial mass in the disk is high and those high-mass giant planets are produced during a phase of giant impacts."

This giant-impacts phase is analogous to the final stage in the assembly of our own solar system when the moon was formed in the aftermath of a collision between Earth and another planet. "Because of our solar system bias, we tend to think of impacts as happening to rocky planets and ejection as happening to giant planets, but there is a whole spectrum of possible outcomes in the evolution of planetary systems," Murray-Clay said.

According to Frelikh, collisional growth of high-mass giant planets should be most efficient in the inner regions, because encounters between planets in the outer parts of the system are more likely to lead to ejections than mergers. Mergers producing high-mass planets should peak at a distance from the host star of around 3 astronomical units (AU, the distance from Earth to the sun), she said.

"We predict that the highest-mass giant planets will be produced by mergers of smaller gas giants between 1 to 8 AU from their host stars," Frelikh said. "Exoplanet surveys have detected some extremely large exoplanets, approaching 20 times the mass of Jupiter. It may take a lot of collisions to produce those, so it's interesting that we see this giant-impacts phase in our simulations."