Archive for the ‘Machine Learning’ Category

Machine learning models for predicting early hemorrhage progression in traumatic brain injury | Scientific Reports – Nature.com

In emergency situations, the collection of precise clinical information from trauma patients can be challenging, with accurate data often being elusive. Accurately assessing the risk of progression in traumatic intracranial hemorrhage (ICH) is essential, particularly for patients who are relatively stable or exhibit minimal traumatic brain hemorrhage, compared to those immediately identified for emergency surgical intervention among the TBI cohort13,14. Additionally, accurately discerning details regarding the mechanism of head injury frequently proves difficult15.

In this study, the objective is to create a predictive model for the short-term prognosis of patients with traumatic brain injury. This model emphasizes the use of clear and readily accessible information from the emergency department setting. Specifically, it relies on data from initial head CT scans and findings from physical examinations, both of which are readily available and easily obtained in the emergency room.

Previous literature has explored the analysis of various traumatic ICH types16. While lvarez-Sabn et al. have reported on the phenomenon of delayed traumatic ICH17, studies that demonstrate a variance in the frequency of ICH progression according to the type of ICH are lacking. Additionally, systematic clinical analyses on the influence of each ICH type on patient prognosis remain unexplored. The ICH type characterized in this study as petechial hemorrhage has also been referred to as blossomed or exhibiting a salt and pepper appearance in prior research16,18. Pathologically, this phenotype signifies a severe manifestation of traumatic subarachnoid hemorrhage that extends into the brain parenchyma, arising from progressive microvascular rupture and consequent bleeding. We hypothesized that PH type would have the worst prognosis due to these pathological differences, and this was confirmed by the XGboost model's feature importance analysis.

The clinical significance of counter coup head injury, characterized by brain injury occurring on the side opposite to the point of impact, has been suggested as a potential indicator of the severity of head trauma19. This perspective is based on the understanding that counter coup injuries are frequently associated with a higher risk of complications, including brain swelling and bleeding, compared to injuries that occur solely at the site of impact, known as coup injuries9.

In this study, we observed that the incidence of counter coup ICH was 17.9% in patients with occipital fractures, a rate higher than in patients with skull fractures at other locations (3.7% in frontal fractures, 7.2% in temporal fractures, and 3.7% in parietal fractures). This led to a notably increased frequency of ICH in the frontal lobe among patients whose initial impact was on the occipital skull. This observed trend may be linked to brain contusions that occur on the irregular surfaces of the anterior cranial fossa of the skull and structures like the anterior clinoid process. This could account for the prevalent association of counter coup ICH in the frontal lobe with TBIs involving occipital skull impacts9.

In our study, we successfully developed an algorithm capable of predicting an individual's prognosis using CT findings and clinical information. By integrating both clinical and radiological factors, such as counter coup injury and the specific type of ICH, we achieved high accuracy in predicting ICH progression among patients with mild to moderate traumatic brain injury (TBI).

The proposed XGBoost model demonstrated an average accuracy of 91% in predicting ICH progression, surpassing the logistic regression model, which achieved an AUC of 0.82. This enhanced performance emphasizes the efficacy of the XGBoost model in predicting ICH progression, highlighting the benefits of applying advanced machine learning techniques over traditional statistical methods for clinical predictions. Furthermore, our analysis validated the significant utility of SHAP values derived from the XGBoost model in assessing individual ICH progression risks. The incorporation of SHAP values enhances the visualization of individual risk factors, offering clinicians a crucial tool for interpreting the effects of various predictors on ICH progression at a personalized level. This capability facilitates more precise and tailored clinical decision-making.

To the best of our knowledge, this study represents the first attempt to develop a machine-learning model specifically for predicting ICH progression using image data from CT scans. We anticipate that our findings will contribute to the early identification of patients at risk for ICH progression, thereby informing treatment decisions and monitoring strategies. This approach has the potential to mitigate the risk of complications and enhance overall outcomes in patients with traumatic brain injury (TBI).

The current study is subject to several limitations. Firstly, due to the limited number of patients in each age group, we were unable to analyze the risk of ICH progression across different age demographics. Secondly, we did not account for the potential impact of variables such as current medication use and underlying health conditions on ICH progression in TBI patients. Due to the challenges in obtaining a complete medical history from patients presenting to the emergency room with traumatic brain injury, our study focused primarily on factors that can be quickly and readily obtained in the ER, particularly radiological factors, to investigate their association with ICH progression. Although we investigated the history of antiplatelet and anticoagulation medication use, only a small proportion of patients (27 out of 650, or 4.2%) were confirmed to have used these medications. This limited number of patients was insufficient to establish a statistical correlation with ICH progression. This likely reflects the unreliability of initial medical history investigations and suggests that patients who were on antiplatelet or anticoagulation therapy might have presented with more severe ICH, thus potentially excluding them from this study due to their immediate need for surgical intervention.

Thirdly, our machine learning model was developed using data from a single institution, highlighting the need for future studies to perform general validation of the models with external datasets.

In forthcoming research, we aim to enhance the accuracy of our algorithm in predicting the progression of TBI. To improve the predictability of our current machine learning algorithm, it will be crucial to gather more comprehensive individual information from patient medical records. Furthermore, future research should investigate the factors influencing the necessity of surgery among patients exhibiting ICH progression, particularly focusing on changes in the Glasgow Coma Scale (GCS) following follow-up and the subsequent need for surgical intervention. Such analysis is anticipated to hold substantial clinical significance.

Follow this link:
Machine learning models for predicting early hemorrhage progression in traumatic brain injury | Scientific Reports - Nature.com

Scientists leverage machine learning to decode gene regulation in the developing human brain – EurekAlert

image:

The study is part of the PsychENCODE Consortium, which brings together multidisciplinary teams to generate large-scale gene expression and regulatory data from human brains across several major psychiatric disorders and stages of brain development. (From left: first authors Sean Whalen and Chengyu Deng, and senior authors Katie Pollard and Nadav Ahituv.)

Credit: Gladstone Institutes / Michael Short

SAN FRANCISCOMay 24, 2024In a scientific feat that broadens our knowledge of genetic changes that shape brain development or lead to psychiatric disorders, a team of researchers combined high-throughput experiments and machine learning to analyze more than 100,000 sequences in human brain cellsand identify over 150 variants that likely cause disease.

The study, from scientists at Gladstone Institutes and University of California, San Francisco (UCSF), establishes a comprehensive catalog of genetic sequences involved in brain development and opens the door to new diagnostics or treatments for neurological conditions such as schizophrenia and autism spectrum disorder. Findings appear in the journal Science.

We collected a massive amount of data from sequences in noncoding regions of DNA that were already suspected to play a big role in brain development or disease, says Senior Investigator Katie Pollard, PhD, who also serves as director of the Gladstone Institute for Data Science and Biotechnology. We were able to functionally test more than 100,000 of them to find out whether they affect gene activity, and then pinpoint sequence changes that could alter their activity in disease.

Pollard co-led the sweeping study with Nadav Ahituv, PhD, professor in the Department of Bioengineering and Therapeutic Sciences at UCSF and director of the UCSF Institute for Human Genetics. Much of the experimental work on brain tissue was led by Tomasz Nowakowski, PhD, associate professor of neurological surgery in the UCSF Department of Medicine.

In all, the team found 164 variants associated with psychiatric disorders and 46,802 sequences with enhancer activity in developing neurons, meaning they control the function of a given gene.

These enhancers could be leveraged to treat psychiatric diseases in which one copy of a gene is not fully functional, Ahituv says: Hundreds of diseases result from one gene not working properly, and it may be possible to take advantage of these enhancers to make them do more.

Organoids and Machine Learning Take the Spotlight

Beyond identifying enhancers and disease-linked sequences, the study holds significance in two other key areas.

First, the scientists repeated parts of their experiment using a brain organoid developed from human stem cells and found that the organoid was an effective stand-in for the real thing. Notably, most of the genetic variants detected in the human brain tissue replicated in the cerebral organoid.

Our organoid compared very well against the human brain, Ahituv says. As we expand our work to test more sequences for other neurodevelopmental diseases, we now know that the organoid is a good model for understanding gene regulatory activity.

Second, by feeding massive amounts of DNA sequence data and gene regulatory activity to a machine learning model, the team was able to train the computer to successfully predict the activity of a given sequence. This type of program can enable in-silico experiments that allow researchers to predict the outcomes of experiments before doing them in the lab. This strategy enables scientists to make discoveries faster and using fewer resources, especially when large quantities of biological data are involved.

Sean Whalen, PhD, a senior research scientist in the Pollard Lab at Gladstone and a co-first author of the study, says the team tested the machine learning model using sequences held out from model training to see if it could predict the results already gathered on gene expression activity.

The model had never seen this data before and was able to make predictions with great accuracy, showing it had learned the general principles for how genes are impacted by noncoding regions of DNA in developing brain cells, Whalen says. You can imagine how this could open up a lot of new possibilities in research, even predicting how combinations of variants might function together.

A New Chapter for Brain Discoveries

The study was completed as part of the PsychENCODE Consortium, which brings together multidisciplinary teams to generate large-scale gene expression and regulatory data from human brains across several major psychiatric disorders and stages of brain development.

Through the consortiums publication of multiple studies, it seeks to shed light on poorly understood psychiatric conditions, from autism to bipolar disorder, and ultimately jumpstart new treatment approaches.

Our study contributes to this growing body of knowledge, showing the utility of using human cells, organoids, functional screening methods, and deep learning to investigate regulatory elements and variants involved in human brain development, says Chengyu Deng, PhD, a postdoctoral researcher at UCSF and a co-first author of the study.

About the Study

The study, Massively Parallel Characterization of Regulatory Elements in the Developing Human Cortex, appears in the May 24, 2024 issue of Science. Authors include: Chengyu Deng, Sean Whalen, Marilyn Steyert, Ryan Ziffra, Pawel Przytycki, Fumitaka Inoue, Daniela Pereira, Davide Capauto, Scott Norton, Flora Vaccarino, PsychENCODE Consortium, Alex Pollen, Tomasz Nowakowski, Nadav Ahituv, and Katherine Pollard.

The work was funded in part by the National Institute of Mental Health, the New York Stem Cell Foundation, the National Human Genome Research Institute, and Coordination for the Improvement of Higher Education Personnel. The data generated was part of thePsychENCODE Consortium.

About Gladstone Institutes

Gladstone Institutesis an independent, nonprofit life science research organization that uses visionary science and technology to overcome disease. Established in 1979, it is located in the epicenter of biomedical and technological innovation, in the Mission Bay neighborhood of San Francisco. Gladstone has created a research model that disrupts how science is done, funds big ideas, and attracts the brightest minds.

Massively parallel characterization of regulatory elements in the developing human cortex

24-May-2024

Read the original:
Scientists leverage machine learning to decode gene regulation in the developing human brain - EurekAlert

AI Startup Says California AI Bill Will Hamper Innovation – BroadbandBreakfast.com

AI

The bill increases regulatory requirements for machine learning systems in California.

May 24, 2024 In a Tuesday press release, Haltia AI, an artificial intelligence startup based in Dubai, warned leaders in machine learning that Californias new AI bill will cripple innovation with overly burdensome regulations.

Haltia said that the bill throws a wrench into the growth of AI startups with its unrealistic requirements and stifling compliance costs.

The legislation, titled the Understanding the Safe and Secure Innovation for Frontier Artificial Intelligence Act, was introduced in February and passed the California State Senate on Tuesday. The act mandates that developers of AI tools comply with various safety requirements and report any safety concerns.

AI systems are defined by the act as machine-based systems that can make predictions, recommendations, decisions, and formulate options. Safety tests include ensuring that an AI model does not have the capability to enable harms, such as creation of chemical and biological weapons or cyberattacks on critical infrastructure. Third party testers will be required to determine the safety of these systems.

Haltia said that on the surface, the act aims for responsible AI development. However, its implementation creates a labyrinth of red tape that disproportionately impacts startups. Because the bill requires ongoing annual reviews, Haltia argues that it adds significant technical and financial burdens.

Arto Bendiken, co-founder and CTO at Haltia, said that the act is a prime example of how well-intentioned regulations can morph into a bureaucratic nightmare. He added that the financial penalties for non-compliance only exacerbate the issue, potentially deterring groundbreaking ideas before they even take flight.

Haltia called for other AI startups to follow its lead and move operations to the United Arab Emirates where its thriving ecosystem, coupled with its commitment to the future of AI, makes it the ideal launchpad for the next generation of groundbreaking AI technologies in the Silicon Valley of the East.

In 2023, California Governor Gavin Newson signed an executive order that announced new directives aimed at understanding the risks of machine learning technologies in order to ensure equitable outcomes when used and to prepare the states workforce for its use.

Read this article:
AI Startup Says California AI Bill Will Hamper Innovation - BroadbandBreakfast.com

Airbnb using machine learning technology to prevent parties – KYW

PHILADELPHIA (KYW Newsradio) With the help of machine learning technology, Airbnb says it will be cracking down on parties this summer.

Its really important that those spaces are respected and treated with care, and that, you know, people are not showing up and taking advantage of that, said Airbnbs Global Director of Corporate and Policy Communications Christopher Nulty.

The best part about staying in an Airbnb is often that you're staying in a neighborhood, and the only way to continue staying in a neighborhood is to be a good neighbor.

Nulty says the company will be using the technology to prevent any disruptive parties, paying close attention to bookings on Memorial Day, Fourth of July and Labor Day. It looks at how long guests are staying, past rental ratings, distance from home, and the number of guests.

So far, it has resulted in a 50% reduction in unauthorized parties. In 2023, more than 67,000 people across the U.S., including 950 in Philadelphia, were deterred from booking entire home listings over those weekends.

Those who are flagged, but arent actually planning on throwing a party, can call Airbnbs customer service line.

Read the original:
Airbnb using machine learning technology to prevent parties - KYW

Bringing generative artificial intelligence to space – SpaceNews

TAMPA, Fla. Amazon Web Services is busy positioning its cloud infrastructure business to capitalize on the promise of generative artificial intelligence for transforming space and other industries.

More than 60% of the companys space and aerospace customers are already using some form of AI in their businesses, according to AWS director of aerospace and satellite Clint Crosier, up from single digits around three years ago.

Crosier predicts similar growth over the next few years in space for generative AI, which uses deep-learning models to answer questions or create content based on patterns detected in massive datasets, marking a major step up from traditional machine-learning algorithms.

Mathematical advances, an explosion in the amount of available data and cheaper and more efficient chips for processing it are a perfect storm for the rise of generative AI, he told SpaceNews in an interview, helping drive greater adoption of cloud-based applications.

In the last year, AWS has fundamentally reorganized itself internally so that we could put the right teams [and] organizational structure in place so that we can really double down on generative AI, he said.

He said AWS has created a generative AI for space cell of a handful of people to engage with cloud customers to help develop next-generation capabilities.

These efforts include a generative AI laboratory for customers to experiment with new ways of using these emerging capabilities.

Crosier sees three main areas for using generative AI in space: geospatial analytics, spacecraft design and constellation management.

Earth observation satellite operators such as BlackSky and Capella Space already use these tools to help manage search queries and gain more insights into their geospatial data.

Its early days in the manufacturing sector, but Crosier said engineers are experimenting with how a generative AI model fed with design parameters could produce new concepts by drawing from potentially overlooked data, such as from the automotive industry.

Whether youre designing a satellite, rocket or spacecraft, youre letting the generative AI go out and do that exploratory work around the globe with decades of data, he said, and then it will come back and bring you novel design concepts that nobody has envisioned before for your team to use as a baseline to start refining.

He said generative AI also has the potential to help operators manage increasingly crowded orbits by helping to simulate testing scenarios.

If I have a constellation of 600 satellites, I want to model how that constellation will behave under various design parameters, he said.

Well, I can get a model of two concepts, which leaves me woefully inadequate but it costs time and money to model them, or I can model an infinite number. Gen AI will tell me what are the top 25 cases I should model for my modeling simulation capability that will give me the best design optimization, and so were seeing it used that way.

AWS efforts to accelerate the adoption of these emerging computing capabilities also include scholarships and a commitment announced in November to provide free AI training for two million people worldwide before the end of 2025.

Continue reading here:
Bringing generative artificial intelligence to space - SpaceNews