We have become accustomed to online algorithms trying to guess our preferences for everything from movies and music to news and shopping. This is based not only on what we have searched for, looked at, or listened to, but also on how these activities compare to others. Collaborative filtering, as the technique is called, uses hidden patterns in our behavior and the behavior of others to predict which things we may find interesting or appealing.

But what if the algorithms could use responses from our brain rather than just our behavior? It may sound a bit like science fiction, but a project combining computer science and cognitive neuroscience showed that brain-based collaborative filtering is indeed possible. By using an algorithm to match an individual’s pattern of brain responses with those of others, researchers from the University of Copenhagen and the University of Helsinki were able to predict a person’s attraction to a not-yet-seen face.

Previously the researchers had placed EEG electrodes onto the heads of study participants and showed them images of various faces, and thereby demonstrated that machine learning can use electrical activity from the brain to detect which faces the subjects found most attractive.

“Through comparing the brain activity of others, we’ve now also found it possible to predict faces each participant would find appealing prior to seeing them. In this way, we can make reliable recommendations for users — just as streaming services suggest new films or series based on the history of the users,” explains senior author Dr. Tuukka Ruotsalo of the University of Copenhagen’s Department of Computer Science.

Towards mindful computing and greater self-awareness

Industries and service providers are more and more often giving personalized recommendations and we are now starting to expect individually tailored content from them. Consequently, researchers and industries are interested in developing more accurate techniques of satisfying this demand. However, the current collaborative filtering techniques which are based on explicit behaviour in terms of ratings, click behaviour, content sharing etc. are not always reliable methods of revealing our real and underlying preferences.

“Due to social norms or other factors, users may not reveal their actual preferences through their behaviour online. Therefore, explicit behaviour may be biased. The brain signals we investigated were picked up very early after viewing, so they are more related to immediate impressions than carefully considered behaviour,” explains co-author Dr. Michiel Spapé.

“The electrical activity in our brains is an alternative and rather untapped source of information. In the longer term, the method can probably be used to provide much more nuanced information about people’s preferences than is possible today. This could be to decode the underlying reasons for a person’s liking of certain songs — which could be related to the emotions that they evoke,” explains Tuukka Ruotsalo.

But researchers don’t just see the new method as a useful way for advertisers and streaming services to sell products or retain users. As lead author Keith Davis points out:

“I consider our study as a step towards an era that some refer to as “mindful computing,” in which, by using a combination of computers and neuroscience techniques, users will be able to access unique information about themselves. Indeed, Brain-Computer Interfacing as it is known, could become a tool for understanding oneself better.”

Nevertheless, there is still a way to go before the technique can be applied beyond the laboratory. The researchers point out that brain-computer interface devices must become cheaper and easier to use before they find themselves in the hands or strapped to the heads of casual users. Their best guess is that this will take at least 10 years.

The researchers also underscore that the technology comes with a significant challenge for protecting brain-based data from misuse and that it is important for the research community to carefully consider data privacy, ownership and the ethical use of raw data collected by EEG.

ABOUT THE EXPERIMENT

In the experiment, participants were shown a large number of images of human faces and asked to look for those that they found attractive. While doing so, their brain signals were recorded. This data was used to train a machine learning model to distinguish between the brain activity when the participant saw a face that they found attractive versus when they saw a face that they did not find attractive.

With a different machine learning model, the brain-based data from a larger number of participants was used to calculate which new facial images each participant would find attractive. Thus, the prediction was based partly on individual participant’s own brain signals and partly on how other participants responded to the images.

Lead author of the study Dr Saidul Islam, from the University of Technology Sydney, said while previous research has revealed how virus aerosols travel through the upper airways including the nose, mouth and throat — this study was the first to examine how they flow through the lower lungs.

“Our lungs resemble tree branches that divide up to 23 times into smaller and smaller branches. Due to the complexity of this geometry it is difficult to develop a computer simulation, however we were able to model what happens in the first 17 generations, or branches, of the airways,” said Dr Islam.

“Depending on our breathing rate, between 32% and 35% of viral particles are deposited in these first 17 branches. This means around 65% of virus particles escape to the deepest regions of our lungs, which includes the alveoli or air sacs,” he said.

The alveolar system is critical to our ability to absorb oxygen, so significant amounts of virus in this region, along with inflammation caused by our body’s immune response, can cause severe damage, reducing the amount of oxygen in the blood and increasing the risk of death.

The study also revealed that more virus particles are deposited in the right lung, especially the right upper lobe and the right lower lobe, than in the left lung. This is due to the highly asymmetrical anatomical structure of the lungs and the way air flows through the different lobes.

The research is backed up by a recent study of chest CT scans of COVID-19 patients showing greater infection and disease in the regions predicted by the model.

The researchers modelled three different flow rates — 7.5, 15 and 30 litres per minute. The model showed greater virus deposition at lower flow rates.

As well as improving our understanding of coronavirus transmission, the findings have implications for the development of targeted drug delivery devices that can deliver medicine to the areas of the respiratory system most affected by the virus.

“Normally when we inhale drugs from a drug delivery device most of it is deposited in the upper airways, and only a minimum amount of drugs can reach the targeted position of the lower airways. However, with diseases like COVID-19 we need to target the areas most affected,” said Dr Islam.

“We are working to develop devices that can target specific regions, and we also hope to build age and patient specific whole lung models to increase understanding of how SARS CoV-2 aerosols affect individual patients,” said co-author and group leader of the UTS Computer Simulations and Modelling group, Dr Suvash Saha.

The World Health Organisation recently updated its advice about the importance of aerosol transmission, warning that because aerosols can remain suspended in the air, crowded indoor settings and areas with poor ventilation pose a significant risk for transmission of Covid-19.

“When we use an aerosol deodorant, the smallest particles of that liquid fall on us under extreme pressure in the form of gas. Similarly, when an infected person speaks, sings, sneezes or coughs, the virus is spread through the air and can infect those nearby,” said Dr Saha.

The study has further applications, with researchers using portable devices to examine air quality — including PM2.5 and PM10 concentration and gasses such as carbon dioxide, formaldehyde and sulphur dioxide — in spaces such as train carriages. The researchers can then use this data to model the impact on our lungs.

The study, SARS CoV-2 aerosol: How far it can travel to the lower airways, was recently published in the journal Physics of Fluids.

“It’s a huge problem,” says Swathi Kiran, director of BU’s Aphasia Research Lab, and College of Health & Rehabilitation Sciences: Sargent College associate dean for research and James and Cecilia Tse Ying Professor in Neurorehabilitation. “It’s something our lab is working to tackle at multiple levels.”

For the last decade, Kiran and her team have studied the brain to see how it changes as people’s language skills improve with speech therapy. More recently, they’ve developed new methods to predict a person’s ability to improve even before they start therapy. In a new paper published in Scientific Reports, Kiran and collaborators at BU and the University of Texas at Austin report they can predict language recovery in Hispanic patients who speak both English and Spanish fluently — a group of aphasia patients particularly at risk of long-term language loss — using sophisticated computer models of the brain. They say the breakthrough could be a game changer for the field of speech therapy and for stroke survivors impacted by aphasia.

“This [paper] uses computational modeling to predict rehabilitation outcomes in a population of neurological disorders that are really underserved,” Kiran says. In the US, Hispanic stroke survivors are nearly two times less likely to be insured than all other racial or ethnic groups, Kiran says, and therefore they experience greater difficulties in accessing language rehabilitation. On top of that, oftentimes speech therapy is only available in one language, even though patients may speak multiple languages at home, making it difficult for clinicians to prioritize which language a patient should receive therapy in.

“This work started with the question, ‘If someone had a stroke in this country and [the patient] speaks two languages, which language should they receive therapy in?'” says Kiran. “Are they more likely to improve if they receive therapy in English? Or in Spanish?”

This first-of-its-kind technology addresses that need by using sophisticated neural network models that simulate the brain of a bilingual person that is language impaired, and their brain’s response to therapy in English and Spanish. The model can then identify the optimal language to target during treatment, and predict the outcome after therapy to forecast how well a person will recover their language skills. They found that the models predicted treatment effects accurately in the treated language, meaning these computational tools could guide healthcare providers to prescribe the best possible rehabilitation plan.

“There is more recognition with the pandemic that people from different populations — whether [those be differences of] race, ethnicity, different disability, socioeconomic status — don’t receive the same level of [healthcare],” says Kiran. “The problem we’re trying to solve here is, for our patients, health disparities at their worst; they are from a population that, the data shows, does not have great access to care, and they have communication problems [due to aphasia].”

As part of this work, the team is examining how recovery in one language impacts recovery of the other — will learning the word “dog” in English lead to a patient recalling the word “perro,” the word for dog in Spanish?

“If you’re bilingual you may go back and forth between languages, and what we’re trying to do [in our lab] is use that as a therapy piece,” says Kiran.

Clinical trials using this technology are already underway, which will soon provide an even clearer picture of how the models can potentially be implemented in hospital and clinical settings.

“We are trying to develop effective therapy programs, but we also try to deal with the patient as a whole,” Kiran says. “This is why we care deeply about these health disparities and the patient’s overall well-being.”

Story Source:

Materials provided by Boston University. Original written by Jessica Colarossi. Note: Content may be edited for style and length.

The study, published in Science Advances, found that periodic changes in the intensity of monsoon rainfall over the past 900,000 years were associated with fluctuations in atmospheric carbon dioxide (CO₂), continental ice volume and moisture import from the southern hemisphere Indian Ocean. The findings bolster climate model predictions that rising CO₂ and higher global temperatures will lead to stronger monsoon seasons.

“We show that over the last 900,000 years, higher CO₂ levels along with associated changes in ice volume and moisture transport were associated with more intense monsoon rainfall,” said Steven Clemens, a professor of geological sciences (research) at Brown University and lead author of the study. “That tells us that CO₂ levels and associated warming were major players in monsoon intensity in the past, which supports what the models predict about future monsoons — that rainfall will intensify with rising CO₂ and warming global temperature.”

The South Asian monsoon is arguably the single most powerful expression of Earth’s hydroclimate, Clemens says, with some locations getting several meters of rain each summer. The rains are vital to the region’s agriculture and economy, but can also cause flooding and crop disruption in years when they’re particularly heavy. Because the monsoons play such a large role in the lives of nearly 1.4 billion people, understanding how climate change may affect them is critical.

For several years, Clemens has been working with an international team of researchers to better understand the major drivers of monsoon activity. In November 2014, the research team sailed aboard the research vessel JOIDES Resolution to the Bay of Bengal, off the coast of India, to recover sediment core samples from beneath the sea floor. Those core samples preserve a record of monsoon activity spanning millions of years.

The rainwater produced by the monsoons each summer eventually drains off the Indian subcontinent into the Bay of Bengal. The runoff creates a layer of dilute seawater in the bay that rides atop the denser, more saline water below. The surface water is a habitat for microorganisms called planktonic foraminifera, which use nutrients in the water to construct their shells, which are made of calcium carbonate (CaCO₃). When the creatures die, the shells sink to the bottom and become trapped in sediment. By taking core samples of sediment and analyzing the oxygen isotopes in those fossils, scientists can divine the salinity of the water in which the creatures lived. That salinity signal can be used as an indicator of changing rainfall amounts over time.

Other data from the samples complement the foraminifera data. River runoff into the bay brings sediment from the continent with it, providing another indicator of rain intensity. The carbon isotopic composition of plant matter washed into the ocean and buried in sediment offers yet another rainfall-related signal that reflects changes in vegetation type. The hydrogen isotope composition of waxes on plant leaves varies in different rainfall environments, and that signature can be reconstructed from sediment cores as well.

“The idea is that we can reconstruct rainfall over time using these proxies, and then look at other paleoclimate data to see what might be the important drivers of monsoon activity,” Clemens said. “That helps us to answer important questions about the factors driving the monsoons. Are they primarily driven by external factors like changes in Earth’s orbit, which alter the amount of solar radiation from the sun, or are factors internal to the climate system like CO₂, ice volume and moisture-transporting winds more important?”

The researchers found that periods of more intense monsoon winds and rainfall tended to follow peaks in atmospheric CO₂ and low points in global ice volume. Cyclical changes in Earth’s orbit that alter the amount of sunlight each hemisphere receives played a role in monsoon intensity as well, but on their own could not explain monsoon variability. Taken together, the findings suggest that monsoons are indeed sensitive to CO₂-related warming, which validates climate model predictions of strengthening monsoons in relation to higher CO₂.

“The models are telling us that in a warming world, there’s going to be more water vapor in the atmosphere,” Clemens said. “In general, regions that get a lot of rain now are going to get more rain in the future. In terms of the South Asians monsoons, that’s entirely consistent with what we see in this study.”

The research was supported by the U.S. National Science Foundation (OCE1634774), the Japanese Society for the Promotion of Science (JPMXS05R2900001 and 19H05595), the Japanese Agency for Marine-Earth Science and Technology, the United Kingdom Natural Environment Research Council (NERC; NE/L002493/1), the United States Geological Survey, and Technology and Research Initiative Fund (Arizona Board of Regents).