Live adaptation of organ models in the OR

The non-deformed liver model (red) adapts to the deformed surface profile (blue). Credit: Graphics: Dr. Stefanie Speidel, KIT, in Medical Physics, 41

During minimally invasive operations, a surgeon has to trust the information displayed on the screen: A virtual 3D model of the respective organ shows where a tumor is located and where sensitive vessels can be found. Soft tissue, such as the tissue of the liver, however, deforms during breathing or when the scalpel is applied. Endoscopic cameras record in real time how the surface deforms, but do not show the deformation of deeper structures such as tumors. Young scientists of the Karlsruhe Institute of Technology (KIT) have now developed a real-time capable computation method to adapt the virtual organ to the deformed surface profile.

The principle appears to be simple: Based on computer tomography image data, the scientists construct a virtual 3D model of the respective organ, including the tumor, prior to operation. During the operation, cameras scan the surface of the organ and generate a stiff profile mask. To this virtual mold, the 3D model then is to fit snuggly, like jelly to a given form. The Young Investigator Group of Dr. Stefanie Speidel analyzed this geometrical problem of shape adaptation from the physical perspective. "We model the surface profile as electrically negative and the volume model of the organ as electrically positive charged," Speidel explains. "Now, both attract each other and the elastic volume model slides into the immovable profile mask." The adapted 3D model then reveals to the surgeon how the tumor has moved with the deformation of the organ.

Simulations and experiments using a close-to-reality phantom liver have demonstrated that the electrostatic-elastic method even works when only parts of the deformed surface profile are available. This is the usual situation at the hospital. The human liver is surrounded by other organs and, hence, only partly visible by endoscopic cameras. "Only those structures that are clearly identified as parts of the liver by our system are assigned an electric charge," says Dr. Stefan Suwelack who, as part of Speidel's group, wrote his Ph.D. thesis on this subject. Problems only arise, if far less than half of the deformed surface is visible. To stabilize computation in such cases, the KIT researchers can use clear reference points, such as crossing vessels. Their method, however, in contrary to others does not rely on such references from the outset.

In addition, the model of the KIT researchers is more precise than conventional methods, because it also considers biomechanical factors of the liver, such as the elasticity of the tissue. So for instance, the phantom liver used by the scientists consists of two different silicones: A harder material for the capsule, i.e. the outer shell of the liver, and a softer material for the inner liver tissue.

As a result of their physical approach, the young scientists also succeeded in accelerating the computation process. As shape adaptation was described by electrostatic and elastic energies, they found a single mathematical formula. Using this formula, even conventional computers equipped with a single processing unit only work so quickly that the method is competitive. Contrary to conventional computation methods, however, the new method is also suited for parallel computers. Using such a computer, the Young Investigator Group now plans to model organ deformations stably in real time.

Source: Karlsruhe Institute of Technology

Video game teaches kids how to code

A screen shot for the video game. Credit: Image courtesy of University of California - San Diego

Computer scientists at the University of California, San Diego have successfully funded on Kickstarter a new and improved version of CodeSpells, a first-person player game they developed that teaches players how to code.

The game's previous iteration, developed by UC San Diego computer science Ph.D. students Sarah Esper and Stephen Foster, has been in use in dozens of schools throughout the world for more than a year. The researchers have been using the game as a platform to learn about the best ways to teach children how to code. They have presented their findings at a wide range of academic conferences, including the upcoming Koli Calling International Conference on Computing Education Research Nov. 20 to 23 in Koli, Finland.

In this latest paper, "CodeSpells: Bridging Educational Language Features with Industry-Standard Languages," the researchers demonstrate that after playing CodeSpells for either four hours over four weeks or 10 hours over seven days, children ages 8 to 12 were able to write code by hand in Java.

"It is the goal of CodeSpells to provide a rich experience of computer science education to students who may not have access to an educator," Esper said.

Researchers now want to make the game more attractive and more fun to play. But they need funds to improve the game's graphics and coding interface. Enter Kickstarter, where the project has already met and exceeded its $50,000 fundraising goal.

"We want the game to be educational, but our biggest goal is to make sure it's fun," Foster said.

He and Esper have co-founded ThoughtSTEM, along with UC San Diego biochemistry Ph.D. 

student Lindsey Handley, to teach children ages 8 to 18 how to code, via onsite classes and video games, including CodeSpells and Minecraft.

In its previous iteration, CodeSpells sent players on quests, which helped them master spells, written in Java. This new version is more open-ended much like Minecraft -- a so-called sandbox game. The players are wizards that can modify the world around them at will. They can build mountains and valleys, levitate objects and start fires. They do so by using Blocky, a visual programming language created by Google, or Javascript.

The hope is that players will come up with their own quests. Researchers also hope that as players tinker with the game, they'll come up with their own exciting spells and share those. The goal is to create a vibrant online community, much like the one that has developed around Minecraft.

The game will feature several modes out of the box, but players will be able to create their own modes too. They'll have the tools to create everything from modes to survive in the wilderness to modes to balance an eco-system. They can even create multi-player magic-based sports to play with their friends.

The game will feature four elements: earth, fire, water and air, which the players can manipulate via spells. So far, computer scientists have completed an early version of gameplay for earth magic. The Kickstarter will fund the development of magic for fire, air and water, with an alpha version to be released on Christmas Day 2014, a beta version in June 2015, and the final copy of the game's creative mode to be released September 2015.

If the Kickstarter exceeds its $50,000 goal, the game's multiplayer functions will be enhanced. In addition, the game will add a fifth element, life, which would give wizards control over animals and plants within the game. That feature would be released in summer 2016. By early 2017, players would be able to create their own species and custom characters within the game.

Rewards for the Kickstarter range from a digital copy of CodeSpells for a $10 donation to access to district-wide licenses for CodeSpells in alpha and beta versions and computational thinking courses to a teacher at each school within a school district for a $5,000 donation.

Source: University of California - San Diego

Playing action video games can boost learning, study finds

A new study shows for the first time that playing action video games improves not just the skills taught in the game, but learning capabilities more generally.

A new study shows for the first time that playing action video games improves not just the skills taught in the game, but learning capabilities more generally.

"Prior research by our group and others has shown that action gamers excel at many tasks. In this new study, we show they excel because they are better learners," explained Daphne Bavelier, a research professor in brain and cognitive sciences at the University of Rochester. "And they become better learners," she said, "by playing the fast-paced action games."

According to Bavelier, who also holds a joint appointment at the University of Geneva, our brains keep predicting what will come next -- whether when listening to a conversation, driving, or even preforming surgery. "In order to sharpen its prediction skills, our brains constantly build models, or 'templates,' of the world," she explained. "The better the template, the better the performance. And now we know playing action video game actually fosters better templates."

Action Players vs. Non-Action Players

In the current study, published in the Proceedings of the National Academy of Sciences, Bavelier and her team first used a pattern discrimination task to compare action video game players' visual performance with that of individuals who do not play action video games.
The action-gamers outperformed the non-action gamers. The key to the action-gamers success, the researchers found, was that their brains used a better template for the task at 
hand.

Video Training

Then, the team conducted another experiment to determine if habitual players of fast-paced, action-rich video games may be endowed with better templates independently of their game play, or if the action game play lead them to have better templates.

Individuals with little video game experience were recruited, and as part of the experiment, they were asked to play video games for 50 hours over the course of nine weeks. One group played action video games, e.g., Call of Duty. The second group played 50 hours of non-action video games, such as The Sims.

The trainees were tested on a pattern discrimination task before and after the video game "training." The test showed that the action video games players improved their templates, compared to the control group who played the non-action video games. The authors then turned to neural modeling to investigate how action video games may foster better templates.

Measuring Learning

When the researchers gave action gamers a perceptual learning task, the team found that the action video game players were able to build and fine tune templates quicker than non-action game control participants. And they did so on the fly as they engaged in the task.
Being a better learner means developing the right templates faster and thus better performance. And playing action video games, the research team found boosts that process.
"When they began the perceptual learning task, action video gamers were indistinguishable from non-action gamers; they didn't come to the task with a better template," said Bavelier. "Instead, they developed better templates for the task, much, much faster showing an accelerated learning curve."

The researchers also found that the action gamers' improved performance is a lasting effect. When tested several months to a year later, the action-trained participants still outperformed the other participants, suggesting that they retained their ability to build better templates.

Bavelier's team is currently investigating which characteristics in action video games are key to boost players' learning. "Games other than action video games may be able to have the same effect," she said. "They may need to be fast paced, and require the player to divide his or her attention, and make predictions at different time scales."

Vikranth R. Bejjanki of the University of Rochester and Princeton University, and Ruyuan Zhang of the University of Rochester are co-lead authors of the study. In addition to Bavelier and the lead authors, researchers from the University of Geneva, University of Wisconsin-Madison, and Ohio State University also contributed to the study.

The Office of Naval Research, the Swiss National Foundation, The Human Frontier Science Program, and the National Eye Institute supported the research.

Source: University of Rochester