This week we look at a deep learning application that can detect pancreatic cancer early. We investigate a new record for carbon based energy storage systems and we discover a way to make carbon fibre from coal waste. Finally we examine a new way to predict a centuries old problem, Rogue Waves.
Pancreatic Cancer Detection
Pancreatic ductal adenocarcinoma (PDAC) is the most deadly solid malignancy. There are no early symptoms therefor it is typically detected late at an inoperable stage. Screening of asymptomatic individuals remains unfeasible. One alternative would be to use a non contrast computed tomography (CT) scan however early identification of PDAC has long been considered impossible. Roughly 466,000 people die from Pancreatic Cancer worldwide each year.
A diverse team from China, the US and Europe have developed a deep leaning application called PANDA that uses the data from a CT scan to identify Pancreatic Cancer. The model cascades in three network stages.
The first stage involves pancreas localization using an nU-Net model (a deep learning method for biomedical image segmentation) . The second stage carries out lesion detection using convolutional neural networks together with a classification head. This allows the identification of subtle texture changes of lesions in the non-contrast CT scan. A specificity of 99% is used for lesion detection to reduce false positives. The third stage involves the differential diagnosis of pancreatic lesions to help identify texture, position and pancreas shape for a more accurate classification of the lesion.
The system has been tested in real world scenarios on a population of 20,530 patients. PANDA may serve as a new tool for large scale pancreatic cancer screening.
Carbon based Energy Storage Record
A team at the Department of Energy in Oak Ridge in Tennessee has designed a record setting carbonaceous super capacitor material that stores four times the energy than the best available commercial material.
Batteries are the most common form of energy storage. They convert chemical energy to electrical energy when required. Capacitors store energy as an electric field, somewhat like static electricity. Capacitors can not (yet) store as much energy as batteries in a given volume however they can recharge repeatedly and do not lose the ability to hold a charge. Super capacitors hold more charge than a simple capacitor and charge and discharge more quickly.
The team used machine learning to develop an optimal outcome for a capacitor. The model predicted the highest capacitance for a carbon electrode would be 570 farads per gram if the carbon were co-doped with oxygen and nitrogen. They then designed an extremely porous doped carbon that would provide huge surface areas for interfacial electrochemical reactions. They then synthesized an oxygen rich carbon framework for storing and transporting charge.
The synthesized material actually had a capacitance of 611 farads per gram. The surface area of the material was more than 4,000 square meters per gram. The material looks like a golf ball with deep dimples when viewed under a microscope.
The new material could improve regenerative brakes, power electronics and auxiliary power supplies. The team now has even more data from which to build the next machine learning model. They hope to push the boundaries of super capacitors even further.
Carbon Fibre from Coal Waste
A team at the University of Kentucky have developed a way to turn coal waste into high value carbon products. Prior to burning coal for power generation, it is washed at preparation plants to remove any non combustible materials. This slurry is called underflow. It is usually stored in impoundments for disposal.
The slurry is comprised of fine coal, small rock and clay particles suspended in water. It is estimated that approximately 58 gigaton of waste slurry has been deposited in impoundments around the world.
The initial work is to separate the coal waste from the other particles. Once separated the waste coal is subjected to direct coal liquefaction to produce a coal extract liquid. That is then filtered. The filtered coal extract was then vacuum distilled forming an isotropic pitch bottom product. This was then recovered and melt spun to produce green fibers. These fibers showed mechanical properties in line with other general purpose carbon fibers.
The technology could be used for low cost carbon fibre and remediation of waste coal impoundments. Carbon fibre is critical to transportation, national security and renewable energy.
Predicting Rogue Waves
Sailors have known about rogue waves since the 17th century. Stories about monster waves have been part of the lore of sailors for centuries. It wasn’t when a 26 meter rogue wave hit a Norwegian Oil platform in 1995 that digital instruments were there to capture and measure the wave. Not only did it prove the phenomenon of rogue waves it also provided scientific evidence.
Since 1995 these extreme waves have been subject to extensive study. A team at the University of Copenhagen have used an AI to discover a mathematical model to determine how and when rogue waves can occur. Using a combined 700 years worth of wave data from more than a billion waves the researchers can now predict the likelihood of being struck at sea by a monster wave at any given time. Data was collected from 158 different locations around the world, 24 hours a day.
These waves have been found to be caused by a combination of factors that can be combined into a single risk estimate. The team found that rogue waves can occur all the time. Their data set registered 100,000 waves that met the definition of a rogue wave, equivalent to one a day somewhere on the planet.
Until recently it was thought that rogue waves were caused by waves combining and uniting their energy. However the team found that the most dominant factor is linear superposition. This is when two wave systems cross each other and reinforce one another for a brief period of time.
There are 50,000 cargo ships on the oceans at any point in time. Shipping companies can now use this algorithm to plan routes in advance and reduce the chance of running into rogue waves. The algorithm and data are all now publicly available.
Paying it Forward
If you have a start-up or know of a start-up that has a product ready for market please let me know. I would be happy to have a look and feature the startup in this newsletter. Also if any startups need introductions please get in touch and I will help where I can.
If you have any questions or comments please comment below.
I would also appreciate it if you could forward this newsletter to anyone that you think might be interested.
Till next week.
Perhaps they can predict when a 26 m wave is arriving for the surfers...