Deep learning which is also termed as hierarchical learning or deep structured learning uses a layered algorithmic architecture to analyze data. Its peculiarity helps organizations and researchers achieve the unachievable in a most innovative manner. Recently, researchers at the University of Eastern Finland developed a new deep learning model that can identify sleep stages as accurately as an experienced physician. This discovery paves the way for better diagnosis and treatments of sleep disorders, including obstructive sleep apnea.

To note, obstructive sleep apnea (OSA) is a nocturnal breathing disorder that causes a major burden on public health care systems and national economies. As noted by a report, it is estimated that up to one billion people worldwide suffer from obstructive sleep apnea, and the number is expected to grow due to population aging and increased prevalence of obesity. When untreated, OSA increases the risk of cardiovascular diseases and diabetes, among other severe health consequences.

Therefore, it wise to have a system that can identify sleep stages for the diagnostics of sleep disorders, including obstructive sleep apnea. Traditionally, sleep is manually classified into five stages, which are wake, rapid eye movement (REM) sleep and three stages of non-REM sleep. However, manual scoring of sleep stages is time-consuming, subjective and costly as well.

Hence to win against such challenges, researchers used polysomnographic recording data from healthy individuals and individuals with suspected OSA to develop an accurate deep learning model for automatic classification of sleep stages. In addition, they wanted to find out how the severity of OSA affects classification accuracy.

In healthy individuals, the model was able to identify sleep stages with an 83.7 percent accuracy when using a single frontal electroencephalography channel (EEG), and with an 83.9 percent accuracy when supplemented with electrooculogram (EOG). In patients with suspected OSA, the model achieved accuracies of 82.9 percent (single EEG channel) and 83.8 percent (EEG and EOG channels). The single-channel accuracies ranged from 84.5 percent for individuals without OSA to 76.5 percent for severe OSA patients. The accuracies achieved by the model are equivalent to the correspondence between experienced physicians performing manual sleep scoring. However, the model has the benefit of being systematic and always following the same protocol, and conducting the scoring in a matter of seconds.

According to the researchers, deep learning enables automatic sleep staging for suspected OSA patients with high accuracy.

The Sleep Technology and Analytics Group, STAG, at the University of Eastern Finland solves sleep diagnostics challenges by using a variety of different approaches. The methods developed by the group are based on wearable, non-intrusive sensors, better diagnostic parameters and modern computational solutions that are based on artificial intelligence. The new methods developed by the group are hoped to improve OSA severity assessment, promote individualized treatment planning and more reliable prediction of OSA-related daytime symptoms and comorbidities.

Future Applications

As noted by the American Academy of Sleep Medicine, the researchers hope the deep learning model can be used to improve the consistency of sleep staging across providers and systems while also completing the scoring in mere seconds. They also noted the potential cost savings by measuring sleep with fewer channels. Ultimately, the researchers think their methods could improve sleep apnea severity assessment, promote individualized treatment planning, and more reliably predict sleep apnea-related daytime symptoms and comorbidities. The research originated at the Sleep Technology and Analytics Group at the University of Eastern Finland, which was created to examine challenges in sleep diagnostics.

The post RESEARCHERS DEVELOPED DEEP LEARNING FOR AUTOMATIC CLASSIFICATION OF SLEEP STAGES appeared first on Artificial Intelligence.

Earlier this year, Google took the wraps off of AutoML Natural Language, an extension of its Cloud AutoML machine learning platform to the natural language processing domain. After a months-long beta, AutoML today launched in general availability for customers globally, with support for tasks like classification, sentiment analysis, and entity extraction, as well as a range of file formats, including native and scanned PDFs.

By way of refresher, AutoML Natural Language taps machine learning to reveal the structure and meaning of text from emails, chat logs, social media posts, and more. It can extract information about people, places, and events both from uploaded and pasted text or Google Cloud Storage documents, and it allows users to train their own custom AI models to classify, detect, and analyze things like sentiment, entities, content, and syntax. It furthermore offers custom entity extraction, which enables the identification of domain-specific entities within documents that don’t appear in standard language models.

AutoML Natural Language has over 5,000 classification labels and allows training on up to 1 million documents up to 10MB in size, which Google says makes it an excellent fit for “complex” use cases like comprehending legal files or document segmentation for organizations with large content taxonomies. It has been improved in the months since its reveal, specifically in the areas of text and document entity extraction — Google says that AutoML Natural Language now considers additional context (such as the spatial structure and layout information of a document) for model training and prediction to improve the recognition of text in invoices, receipts, resumes, and contracts.

Additionally, Google says that AutoML Natural Language is now FedRAMP-authorized at the Moderate level, meaning it has been vetted according to U.S. government specifications for data where the impact of loss is limited or serious. It says that this — along with newly introduced functionality that lets customers create a data set, train a model, and make predictions while keeping the data and related machine learning processing within a single server region — makes it easier for federal agencies to take advantage.

Already, Hearst is using AutoML Natural Language to help organize content across its domestic and international magazines, and Japanese publisher Nikkei Group is leveraging AutoML Translate to publish articles in different languages. Chicory, a third early adopter, tapped it to develop custom digital shopping and marketing solutions for grocery retailers like Kroger, Amazon, and Instacart.

The ultimate goal is to provide organizations, researchers, and businesses who require custom machine learning models a simple, no-frills way to train them, explained product manager for natural language Lewis Liu in a blog post. “Natural language processing is a valuable tool used to reveal the structure and meaning of text,” he said. “We’re continuously improving the quality of our models in partnership with Google AI research through better fine-tuning techniques, and larger model search spaces. We’re also introducing more advanced features to help AutoML Natural Language understand documents better.”

Notably, the launch of AutoML follows on the heels of AWS Textract, Amazon’s machine learning service for text and data extraction, which debuted in May. Microsoft offers a comparable service in Azure Text Analytics.

The post Google launches AutoML Natural Language with improved text classification and model training appeared first on Artificial Intelligence.