Radiologists get some help from AI

Can AI beat radiologists at detecting brain he hemorrhages?

A study at the University of California at Berkeley and at San Francisco demonstrated that an algorithm developed by a team of researchers can beat two out of four radiologists at the detection of tiny haemorrhages. This is important because it could be of timely help to those who have strokes, aneurysms or injury.

The algorithm combined with a CT Scan can carry out its detection in a second, locate the region in which they occur and even classify the problem it detects, colour coding it for quick identification. Radiologists look at thousands of images each day often looking for tiny barely visible signs. The idea behind developing the algorithm is to assist busy radiologists be more efficient. 

Protests in VR

Protesters in Hong Kong have hit upon a novel way to increase awareness for their cause, according to a Reuters report. Students have developed a virtual reality game to enable others to experience the clashes with government forces without actually being at the scene. The game is described as short but detailed.

The clash takes place in first person perspective at a shabby high rise block and in it, players have to avoid tear gas, run from riot police, and make sure they don’t get caught near fires. The students hope the VR experience will make the events more real and encourage non-participants to ‘fight for freedom’.

The nature of VR is that it allows for a 360-degree view so that when a user turns around, the scene will pan out just as it would in real life. The game is not yet available to the public but it has been submitted for distribution. 

What three words?

If you’ve ever used a Maps app to get somewhere and not landed up exactly at the expected destination, you’ve experienced the frustration that an app called  what3words is trying to address.

Instead of using the usual coordinates to get to a place which may not even have an address, a startup divides the whole world into three metre squares. You use the what3words app to generate a three word combination for your destination and that’s what is used to get exactly to the place. The words could be anything at all: beer-lamp-banana, wallet-chair-tree etc.

The start-up claims this combo can get emergency help to those lost in forests where there’s of course no address. This is a precise method, but not one without its critics. Though it’s available in some different languages, it can be an interesting problem if a user tries to translate it directly into another. The word combinations can also sometimes be inappropriate such as one that contained weep for the Auschwitz concentration camp. 

Maintaining robotic balance

Four-legged robots — and there are se frightening looking ones not unlike dogs, thanks to Boston Dynamics and other companies — can balance better than the two-legged ones that want to look like human beings.

Yet, it’s important that these humanoids too learn the art of balance so they can be put to work in critical situations such as rescue work where anything like rubble or uneven steps could cause both human and humanoid to take a tumble. Engineers at MIT and the University of Illinois at Urbana-Champaign have come up with a way to help upright robots move through tricky places.

The MIT team uses the obvious to teach their robot to understand balance — a human being. As the robot is controlled remotely, a system of physical feedback on a vest worn by the controller takes place as the robot moves. If it stumbles, the controlling person feels a corresponding pull sensation via the vest. As the controller adjusts position to correct balance, that feedback goes to the robot to do likewise to re-balance and machine-learn the art of balancing when walking, jumping and so on.

 

Complied by Mala Bhargava

comment COMMENT NOW