Archive for the ‘Artificial Intelligence’ Category

Artificial Intelligence Greatly Speeds Radiation Therapy Treatment Planning – Imaging Technology News

The traditional treatment planning process takes days to create an optimized radiation therapy delivery plan, but new artificial intelligence (AI) technologies are helping speed this process. In some cases autonomous AI-generated plans can be generated in a couple minutes.

"AI can help treatment planners and dosimetrists by saving a lot of time doing simpler, repetitive tasks that you do again and again, now the AI can do that for you and you can spend your time onthose more challenging, difficult cases and you can do a better job there," explained Steve Jiang, Ph.D., director of the medical artificial intelligence and automation lab and vice-chair of the Department of Radiation Oncology, University of Texas Southwestern.

Varian and RaySearch have both developed machine-learning technologies to automate treatment plans.

"The fully automated system takes in the patient imaging and the target defined by the physician, and out on the other end comes a fully deliverable therapy plan," said Kevin Moore, Ph.D., DABR, deputy director of medical physics and associate professor, University of California San Diego, who is using the Varian AI TSP software.

He said UCSD began using the software in tandem with traditional treatment planning to ensure the plans were as good as human created plans. After a human-made plan was completed, he said they ran the AI and it took 5-20 minutes to complete a plans start to finish depending on the complexity. "The comparisons were very good," Moore said. The site is now running the AI treatment plans first and and the human planner looks at it to see if it can be further optimized.

The system has helped with speed and efficacy at UCSD Moore said, and the site has now treated well over 1,000 patients with its AI-assisted planning.

Machine learning was incorporated into the RaySearch 8B TPS in 2018 and began it began to beused clinicallyin 2019. The system is trained to take the treatment planning computed tomography (CT) scan and automatically segment the anatomy and auto contour to help speed the planning process.

Princess Margaret Cancer Center, part of the University Health Network in Toronto, Canada, was an early adopter of AI-based treatment planning in 2019. The center conducted a study where it used the RaySearch machine learning treatment plan system to automatically generate plans along side traditional human made plans. The radiation oncologist then compared the two plans to decide if they are acceptable and which is their favorite plan.

"The automated treatment planning system works by training the algorithm with curated sets of similar treatment plans and it is able to detect the patients who are most similar to a novel patient and create a new treatment plan for that patient with no user interaction, beyond pressing the play button," explained Leigh Conroy, Ph.D., physics resident, at Princess Margaret Cancer Center, who has been working on the AI implementation.

She said the machine-generated plan can be modified and optimized by the human planners before being sent to the therapy system. The center began using the system clinically last summer.

Some plans are easier to create than others, so Conroy said the AI system might be used to help free up treatment planners to work on more complex cases, such as head and neck.

TheRayStation 8B TPS software can be trained to automate treatment planning as well as organ segmentation, either using the clinics own data or by pre-trained models provided by RaySearch.

RaySearch is developing several other machine learning applications, including target volume estimation and large-scale data extraction and analysis.

Varian received FDA clearance for its Ethos AI-driven radiation therapy system in February 2020. It is an adaptive intelligence solution that uses AI in the treatment system to take the onboard cone beam CT imaging and compare it to the treatment plan and deliver an entire adaptive treatment plan in a typical 15-minute treatment time slot, from patient setup through treatment delivery.

Tumors change position and size during a course of radiotherapy treatments.Changes in anatomical positioning that move tumors outside TPS margins also occur due to weight loss, position of the uterus, how full the bladder is, and intestinal gas, explained David Sjostrom, Ph.D., deputy chief physicist, Herlev Hospital, Department of Oncology, Division of Radiotherapy, Herlev, Denmark. Hehadthe first clinical experience treating patients with the Varian Ethos system in September 2019.

"Normally it would take days to modify a treatment plan, and you don't do that online with the patient on the table. What we have done up until today was treating within the margins of the plan, or maybe had a study where we had different selections of plans, but it is still not the optimal way of doing it," Sjostrom said.

"So you would have a plan for a small bladder, a medium sized bladder and a large bladder, but what if it is anything in between? With the AI drive workflow, we now get results where we don't need to edit anything and that is the beauty," he explained.

Sjostrom said the system takes a couple minutes to create the plan and then it can be compared to the original plan and you can tell the system which one you want to use, or you can modify one of the plans. However, Sjostrom said so far they have picked the AI generated plan without modification. He said they have averaged a 40 percent reduction in the target margins using the AI-based system.

In early 2018, Mirada Medical received U.S. Food and Drug Administration (FDA) clearance for the worlds first AI-powered auto-contouring solution. The DLCExpert software automatically contours of computed tomography (CT) scans based on next-generation deep learning contouring (DLC) and is now in use in multiple academic medical centers worldwide.

The DLCExpert deep learning software automatically identifies organs, segments and auto-contours them as the first step in creating radiation oncology treatment plans without any human intervention. The files created by the software are vendor neutral and can be imported into any vendors treatment planning system.

Siris Medical also gained FDA clearance in 2018 for its AI-driven, real-time editing of plan contours in its PlanMD decision support software. As the user draws the contours of the target treatment area, the software lets the user see the result of their efforts in real time, without re-optimizing or replanning, which can lead to significant time savings. A newer version of the software was released in January 2019.

The PlanMD software is a complement to Siris other AI product, QuickMatch, which automatically pulls up prior cases from its archive that are most similar to the current one. Once the QuickMatch algorithm has been properly trained, it can reduce the time required for treatment planning by up to 70 percent, according to the vendor.

As radiation oncology moves towarduse of magnetic resonance imaging (MRI) radiotherapy treatment systems because of the better soft tissue delineation and ability to provide real-time imaging during treatment, AI will play a role to help eliminate the need for CT scans. CT is the basis for all current treatment planning because the radiation absorption of various tissues can be calculated from the Hounsfield units of the image grayscales that makeup the image. So all plans, even MR-guided Linacs, require CT imaging for treatment planning.

However, AI can create MRI-derived CT-like image reconstructions for treatment planning and eliminate the need for CT exams, Jiang explained. He said this can help reduce costs and save hospital resources. Eliminating CT scans also can help speed the planning process and reduce patient wait times for treatment, he said.

VIDEO: Artificial Intelligence Driven Adaptive Radiotherapy System Begins Treating Patients Interview with David Sjostrom, Ph.D.

VIDEO: Artificial Intelligence Automatic Contouring and Segmentation For Radiotherapy

New Treatment Planning System Technologies

VIDEO: Real-world Implementation of Deep Learning for Treatment Planning Interview with Kevin Moore, Ph.D.

VIDEO: Varian Showcases Latest Developments at ASTRO 2019

Top Technology Trends at ASTRO 2018

Varian Receives FDA 510(k) Clearance for Ethos Therapy

VIDEO: The Impact of Artificial Intelligence on Radiation Therapy Interview with Steve Jiang, Ph.D.

Visit link:
Artificial Intelligence Greatly Speeds Radiation Therapy Treatment Planning - Imaging Technology News

Is it right to use artificial intelligence in aviation security? – Airport Technology

]]> Greater Toronto Airports Authority recently announced it would be testing HEXWAVE, an AI-enhanced weapon detection technology. Credit: Florian Weihmann (Pexels).

Humans, who are limited by slow biological evolution, couldnt compete and would be superseded. On the face of it, a chilling statement made by the late Stephen Hawking. The world-renowned theoretical physicist was speaking of his fears of an unleashed artificial intelligence (AI). The development of full artificial intelligence could spell the end of the human race, he said, It would take off on its own, and re-design itself at an ever-increasing rate.

The role AI will play in tomorrows world has long been debated, with supporters and sceptics often happy to promote the merits of their argument, whilst almost all the time questioning that of the other side. Some people call this artificial intelligence, but the reality is this technology will enhance us. So instead of artificial intelligence, I think well augment our intelligence, counters IBM CEO Ginni Rometty.

For now, what is almost irrefutable is the impact AI has on our everyday life. In many cases, however, we often dont know its there, working in the background to make daily routines that little bit easier, and even safer. They are the hopes of the Greater Toronto Airports Authority (GTAA) which recently announced it would be testing HEXWAVE, an AI-enhanced weapon detection technology.

Pearson International Airport will be one of the test sites for Liberty Defenses system, which uses 3D radar imaging to detect and identify weapons. The intention, in aviation at least, is to use the system at a facilitys perimeter with the hope of identifying threats before they reach the terminal.

Keen to stress the purpose is not to replace security measures, the companys CEO Bill Riker said it would further enhance already employed security systems. As well as its trial at the airport, HEXWAVE will be tested at other venues, such as sports stadiums, shopping centres, education facilities and government sites across North America and Europe.

It is a product that detects metal and non-metal objects, and alerts responders through the use of AI technology, says GTAA director of corporate safety and security Dwayne MacIntosh. He said the Authority is looking at testing it at multiple locations to determine how we can best operationalise such a product in the future during the week-long project, although no date has yet been set.

The AI technology looks for prohibited items and alerts airport responders, who then work to mitigate the incident.

HEXWAVEs 3D imaging capabilities mean it can identify even the smallest of weapons including concealed handguns and knives as well as the likes of bombs and suicide vests with what the company says is a low-energy radar. This scans the outline of the body to reveal any abnormal contours, as well as detecting prohibited items in baggage. While the use of technology like this in this realm isnt new, it is the speed with which HEXWAVE carries out the operation that stands out. Rather than a subject needing to stand still, it can carry out its role on moving targets.

The AI component means the system itself assesses and determines whether an individual poses a threat. The AI technology looks for prohibited items and alerts airport responders, who then work to mitigate the incident, says MacIntosh. It can be used both inside and out and is scalable to any setting, helping manage throughput screening in real-time.

He is a firm believer in the advantages AI can offer sites such as airports. He says it facilitates the tracking and identification of threats, integrated into wider airport systems in such a way that it makes response seamless. Using AI to speed the process of identifying, tracking and dispatching resources allows for faster responses and better threat mitigation, thereby protecting passengers and the airport community, he continues.

However, critics may question whether this might infringe on individuals rights and even privacy. Its a concern MacIntosh wants to dispel, saying the system will simply not collect anyones data. Privacy is as important to us as safety, and we will always take measures to preserve both for passengers and the airport community, he adds.

But are systems of this kind necessary? Since the terror attacks on New York and Washington in September 2001, airport security has changed dramatically. In the last few years, AI technologies have been rolled out, with demand continuing to grow as the sector looks to strengthen itself against evolving threats. They have become increasingly crucial in that battle to stay ahead; indeed in the security sphere, they are becoming integral to the work of personnel.

Although there are new issues such as drones, many of the same concerns that we face today existed a decade ago.

Security scanners boasting AI technologies are fast becoming commonplace. Body scanning similar to HEXWAVE is already being used by some airports; facial recognition and other biotechnologies are now complimenting these systems and assisting the increased physical security presence since 9/11. Although there are new issues such as drones, many of the same concerns that we face today existed a decade ago, says MacIntosh. The difference is that the tools we use to mitigate those issues have continued to advance, giving us greater ability to protect passengers and the airport community.

See the article here:
Is it right to use artificial intelligence in aviation security? - Airport Technology

Artificial intelligence myths: Reality check – Livemint

Very few subjects in science and technology have caused much excitement right now as artificial intelligence as some of the worlds brightest minds have said that its potential to revolutionise all aspects of our lives.

AI makes it practical for machines to understand from experience, act human-like jobs, and adapt to the latest inputs. The concept works by amalgamating enormous data with quick, smart algorithms, and iterative processing, enabling the software to decipher by analysing patterns in the data in an automatic way.

There is science and well thought algorithm behind all the artificial solutions, where you need to set up proper expectations and clarification to avoid any rumours and myths around the outputs.

While the notion of AI is turning into a massive component of business and consumer transformations, its execution is generally stagnated because of some misconceptions associated with it.

Myth 1: AI will deliver magical resultsimmediately

The path to AI success is hard and takes time, and not just because of the technology. You also need a strategic framework and an iterative approach to avoid delivering a random set of disconnected AI solutions. The temptation is to go for moonshots to deliver the magic, but such projects often fail to live up to expectations because you dont have the basics homework done.

AI is not a magic, it requires rigour, logical thinking and long term strategy with a patience to do multiple iteration to get to the result.

Myth 2: AI Will Replace Human Jobs

Most of the times, management look at AI solutions to replace human and reduce the operational cost, creating a sense of fear among the employees.

So, if you think that AI solutions might strip human from their jobs, then you are undeniably wrong.

Reality is, AI and human need each other. AI is at its most valuable when it augments peoples capabilities. It can remove the duplicate work, freeing people up for more strategic activities. That has the added benefit of making people more motivated, productive, and loyal. Enterprise AI also relies on people to feed it the right data and work with it the right way. Often, AI doesnt provide conclusive answers to issues, but rather highly informed recommendations that an actual human can weigh to make the final decision.

Myth 3: AI Implementation Needs Huge Investment

Artificial developments resolutions appear to be tremendously scientific and complicated. This inclination recommends that just a modern tech organisation, including Google, Amazon, or Apple, with an extended team of experts and billion-dollar budgets can pay for implementing AI. In reality, there are a lot of smart tools existing for an enormous variety of organisation, which can be utilised to implement AI in their business procedures.

Myth 4: AI Algorithms are Competent to Process Any Data

Most of you must believe that ML algorithms are one of the most crucial elements in the entire system. An algorithm might appear to be robust and linked with the human brain, which can make intellect of any untidy data.

It is not possible, for algorithms, to make decisions without human intervention as they dont have magic power. It requires a specific piece of data to get impeccable results.

Myth 5: AI will Conquer Humanity

Machines are powerless to imagine similar to people and will barely be taught to do so. In fact, computers are going to have an optimistic impact on the world by supporting people in a lot of fields, building innovative business models, communities, and skills. Its certainly true that the advent of AI and automation has the potential to seriously disrupt labour and in many situations it is already doing just that. However, seeing this as a straightforward transfer of labour from humans to machines is a vast over-simplification. In fact, a lot of AI focus has been on reducing the drudgery" of day-to-day aspects of the work. AI gives an opportunity to upgrade your skills and move up in your career ladder at the same time.

About the Author: A technology and product leader, Rahul Kumar is Group Chief Product Officer with HT Media Group. An alumni of BIT Mesra, who later on honed his technology management skills from IIT Delhi, has been leveraging AI, ML and IOT to solve business and consumer problems across technology led startups and conglomerate.

See the article here:
Artificial intelligence myths: Reality check - Livemint

Battery Researchers Look to Artificial Intelligence to Slash Recharging Times – Greentech Media News

The battery sector is turning to artificial intelligence for clues on how to improve recharging rates without increasing the degradation of lithium-ion batteries.

Last month, a team from Stanford University, the Massachusetts Institute of Technology and the Toyota Research Institute published findings from battery testing aimed at cutting electric-vehicle charging times down to 10 minutes. The research, published in Nature, revealed how artificial intelligence could speed up the testing process required for novel charging techniques.

The researchers wrote a program that predicted how batteries would respond to different charging approaches and was able to cut the testing process from almost two years to 16 days, Stanford reported. The technique was used to evaluate 224 possible high-cycle-life charging processes in just over two weeks, the researchers said.

The research effort has been in progress for at least three years. In 2017, the Toyota Research Institute committed $35 million to artificial intelligence battery research, initially focusing on new materials.

Last year, the research partners claimed artificial intelligence could help predict the useful life of lithium-ion batteries to within 9 percent of the actual life cycle of the products.

The standard way to test new battery designs is to charge and discharge the cells until they die,co-lead author Peter Attia, now of Tesla but then a Stanford doctoral candidate in materials science and engineering, said in a press release at the time.

Since batteries have a long lifetime, this process can take many months and even years. Its an expensive bottleneck in battery research.

Independentof these efforts, a Canadian firm called GBatteries is using artificial intelligence in a bid to cut lithium-ion battery charging times down to five minutes. The company has succeeded in recharging an electric scooter battery in less than 10 minutes.

The main challenge with extremely fast charging is that it heats up and degrades the battery, GBatteries co-founder and Chief Commercial Officer Tim Sherstyuk told GTM.

The rates that can be achieved with todays fast-charging technology, which are slow by gas-station filling standards, are already problematic for batteries, he said.

Most fast-charging initiatives focus on novel chemistries that wont degrade easily, Sherstyuk said. GBatteries, meanwhile, uses artificial intelligence to monitor the state of the battery as it is charging.

Once the impedance of the battery reaches a critical level, the GBatteries algorithm pauses charging long enough to avoid irreversible damage. This allows charging to proceed in a series of high-intensity pulses at a rate much faster than is possible with traditional methods.

The GBatteries technology works for small batteries and has been demonstrated on power tools, cutting charging times from between 30 to 60 minutes down to 11. But scaling it up to cope with an electric vehicle battery pack is going to take a while, said Sherstyuk.

Even if artificial intelligence can help crack the means to charge electric vehicles as quickly as you now fill your tank with gas, it will take a while for the auto industry to incorporate the technology into the mainstream. The time horizon is years, not months.

Nevertheless, there is plenty of industry interest in tackling the problem.

Charging time is usually the fourth concern that people raise when considering to go electric or not, after upfront cost, range of the vehicle and where [to] charge, said Aaron Fishbone, director of communications at GreenWay, which operates a fast-charging network across Eastern Europe.

So, while not a top-tier issue, its still one raised by many people.

GBatteries pulse charging will require a lot more testing before it might be considered appropriate for the 50+ kilowattpower ratings required for electric vehicles, Fishbone said. In the meantime, high-power recharging is already reducing the time it takes to charge a battery.

Although there are not yet many cars that can take them, a 150-kilowatt charger can add 100 kilometers (62 miles) of range to an electric vehicle within a little over seven minutes, Fishbone said.

Nonetheless, anything which can speed up charging time without degrading battery life is a welcome development and can lead to other innovations which push the whole industry."

View original post here:
Battery Researchers Look to Artificial Intelligence to Slash Recharging Times - Greentech Media News

Artificial intelligence taking lessons on how to second-guess us – Times of India

'; var randomNumber = Math.random(); var isIndia = (window.geoinfo && window.geoinfo.CountryCode === 'IN') && (window.location.href.indexOf('outsideindia') === -1 ); console.log(isIndia && randomNumber Currently, AI may do a plausible job at detecting the intent of another person. It may even have a list of predefined, possible responses that a human will respond within a given situation, they said.

However, when an AI system or machine only has a few clues or partial observations to go on, its responses can sometimes be a little, noted the researchers.

What were doing in these early phases is to help machines learn to act like humans based on our daily interactions and the actions that are influenced by our own judgment and expectations so that they can be better placed to predict our intentions, said Lina Yao, a lecturer at UNSW. This may even lead to new actions and decisions of our own, so that we establish a cooperative relationship, Yao said.

The researchers want to see awareness of less obvious examples of human behaviour integrated into AI systems to improve intent prediction. However, doing so is a tall order, as humans themselves are not infallible when trying to predict the intention of another person, the researchers said.

Sometimes people may take some actions that deviate from their own regular habits, which may have been triggered by the external environment or the influence of another persons actions, she said.

Yao and her team are developing a prototype humanmachine interface system designed to capture the intent behind human movement.

Excerpt from:
Artificial intelligence taking lessons on how to second-guess us - Times of India