Connect with us

Healthcare

AI technology from IBM detects breast cancer risk before it happens

Avatar

Published

on

Share this:

IBM has taken a step forward in disease prevention by designing an artificial intelligence technology that can predict the risk of breast cancer developing up to one year before the first signs of cancer appear.

With this new step in medical diagnosis, IBM has developed an artificial intelligence model which is capable of predicting malignant breast cancer within a year with an 87 percent accuracy rate (when the output from the machine is compared with expert radiologists.) In addition, the technology could correctly predict 77 percent of non-cancerous cases.

The prediction methods uses both mammogram images and medical records in order to make the assessment, based on a data review of the medical evidence. This is the first application of AI to draw upon both images and data to make a prediction in relation to breast cancer.

At the heart of the deep neural network technology is an algorithm, which was trained by IBM technologists along with medical professionals from Israel’s largest healthcare organizations. The training of the artificial intelligence took place using anonymized mammography images which were linked to biomarkers (like patient reproductive history) together with clinical data. The training data-set consisted of 52,936 images from 13,234 women who underwent at least one mammogram between 2013 and 2017.

The aim of the medtech is not to replace the physician, but to act as a ‘second pair of eyes’, providing a backup in the event that something has been missed through conventional patient assessment. This could prove especially useful in areas with staff shortages where a second medical professional is not available to provide a second assessment.

An assessment of the technology has been published in the journal Radiology. The research paper is titled “Predicting Breast Cancer by Applying Deep Learning to Linked Health Records and Mammograms.”

In related news, IBM is applying artificial intelligence to catch Type 1 diabetes much earlier. IBM’s other health technology project could help identify patients at risk and help chart a course for tracking the condition. The predictive tool is a joint project between IBM and JDRF (formerly known as the Juvenile Diabetes Research Foundation).

Share this:

Healthcare

IoT + Data = Healthcare Intelligence

In the equation IoT + X = Intelligence, what role can patient and asset data play as the X factor?

Avatar

Published

on

Share this:

The essential premise of the Internet of Things — that a device can pick up vital information and relay that for processing through a Wi-Fi connection — is systematically shaving off inefficiencies in healthcare. This is especially welcome news in the United States where healthcare spending rose to a crushing $3.6 trillion in 2018, which amounts to nearly 18% of the overall economy. 

IoT in healthcare roughly falls into two buckets: Data from patients and data from institutional assets. Here’s where judicious implementation of an IoT strategy helps:

Effective patient monitoring

Consumers are already familiar with the Apple Watch or FitBit to monitor physical activity, heart rate, sleep patterns and more. Physicians can rely on condition-specific monitors to deliver more relevant data. Monitoring of glucose levels from diabetics through an IoT-enabled wearable can help track insulin needs and gain a better handle on preventing complications. Such monitoring also empowers patients — they can read levels through a related mobile app — and gives them greater ability to participate in their healthcare strategy.

Decreases post-op costs

IoT-enabled wearables/sensors can monitor patient health after they have returned home from major surgeries and automatically alert the hospital if certain vitals look worrisome. Wearable and implantable stickers monitor heart rate. Smart bandages can keep an eye on wounds and watch for infection. Such remote monitoring of fairly routine vitals eliminates the need for the patient to be tethered to the hospital for extended periods after surgery. Remote IoT-enabled monitoring also enables tele-health where physicians can remotely work with patients who report problems. Pre-screening like this has the potential to decrease the need for readmission.

Tracking medical assets

IoT-enabled sensors on medical devices — and even staff — can help track assets more efficiently. Staff can bring X-ray machines and traveling IV units into service as needed, instead of wasting time tracking them down. Such data also helps hospitals forecast device utilization so they can better plan for need. Hospitals can also restrict access to specific drugs by allowing remote IoT-based monitoring of these medicines. Room sensors can read ID badges and only allow approved personnel into sensitive areas.

Predictive maintenance

IoT can spit out data not just about patient health — but also that of machines. If a refrigerator holding critical medicines is about to break down, a sensor connected to the unit can alert maintenance who can proactively attend to the machine before it goes out of order. Healthcare organizations must keep equipment running smoothly and IoT enables them to do so.

Reduce ER wait times

By IoT-enabled tracking of assets such as hospital beds and aligning them with patient needs healthcare organizations can dramatically decrease wait times in the emergency room. Mount Sinai Medical Center in New York City has used this strategy to cut wait times for half of its 59,000 annual ER patients.

While IoT is a powerful tool in healthcare, it must also play by the same rules that govern other devices and systems. IoT-enabled healthcare devices go through elaborate certification processes and conform to country-specific patient privacy laws. Since IoT in healthcare will often involve sharing and relaying of sensitive patient health information (PHI), organizations need to encrypt data and remove all identifiers before they can work with them at scale.

Skyrocketing healthcare costs demand efficiencies at scale. IoT and patient or asset data can deliver such prescriptive price reductions while still maintaining high care standards.

Share this:
Continue Reading

Healthcare

North America’s first digital hospital launches second generation Command Centre

Avatar

Published

on

From left to right: Jhanvi Solanki, Dr. Susan Tory and Jane Casey in Humber River Hospital’s Command Centre. - Photo by Humber River Hospital
Share this:

Do the words ‘command centre’ make you think of huge rooms with NASA scientists, expertly making sure a Mars rover lands safely on the Red Planet?

What if a command centre could revolutionize the patient experience in one of the busiest hospitals in North America, bringing a new standard of patient-centered, quality healthcare?

Combining Artificial Intelligence, Machine Learning, and professional expertise, Humber River Hospital in Toronto has launched the world’s first clinical analytic applications, in partnership with GE Healthcare Partners (GEHC). 

Displayed on large-screen monitors at HRH’s 4,500 square-foot Command Centre, these four new applications or analytic ’tiles’ use “standardized early warning systems, predictive analytics, real-time information from multiple digital systems,” alongside the professional expertise of experienced nurses.

Canadian Patient Safety Institute and Canadian Institute for Health Information data shows that 1 in 18 hospital stays in Canada involved at least one harmful event. This addition to the Command Centre means quicker alerts for clinical staff, and better protection for patients with conditions that make them vulnerable to risks of adverse events, or adverse outcomes.

The Humber River Hospital is the Greater Toronto Area’s (GTA) largest acute care centre, serving a catchment area of more than 850,000 in the city’s northwest. Opening in 2015, it was also North America’s first fully digital hospital. 

Just two years later, HRH opened the first generation of its Command Centre, a data-driven ‘mission control’ offering real-time insight on patient flow, via advanced algorithms and predictive analytics. As a result, the hospital has “unlocked” the equivalent of 35 additional beds — and the ability to treat thousands of additional patients.

Get an inside look at HRH’s Generation 2 Command Centre:

Share this:
Continue Reading

Healthcare

Canadian startup Deep Genomics uses AI to speed up drug discovery

Avatar

Published

on

genetics
Share this:

One of the biggest challenges pharmaceutical companies face is with the time taken to discover new drugs, develop them and get them to market. This lengthy process is punctuated with false starts. Startup Deep Genomics uses AI to accelerate the process.

Canadian startup Deep Genomics has been using artificial intelligence as a mechanism to speed up the drug discovery process, combining digital simulation technology with biological science and automation. The company has built a platform which uses machine learning to delve into the molecular basis of genetic diseases. The platform can analyze potential candidate drugs and identify those which appear most promising for further development by scientists.

The drug development process is dependent upon many factors, such as those relating to combining molecules (noting the interactions between hundreds of biological entities) and with the assessment of biomedical data. The data review required at these stages is highly complex. For these reasons, many researchers are seeking algorithms to help to extract data for analysis.

According to MaRS, Deep Genomics is addressing the time consuming element involved in the initial stages of drug discovery. The artificial intelligence system that the company has designed is able to process 69 billion molecules, comparing each one against around one million cellular processes. This type of analysis would have taken a conventional computer (or a team of humans) many years to run the necessary computations.

Within a few months, Deep Genomics AI has narrowed down the billions of combination to shortlist of 1,000 potential drugs. This process is not only faster, it narrows down the number of experiments that would need to be run, saving on laboratory tests and ensuring that only those drugs with a high chance of success are progressed to the clinical trial stage.

This type of system goes some way to addressing the lengthy typical time to market, which stands at around 14 years for a candidate drug; as well as reducing the costs for drug development, which run into the billions of dollars per drug.

Share this:
Continue Reading

Featured