Tyler O'Neal, Staff Editor APPLICATIONS

Dr. Conant shows how AI improves efficiency, accuracy of digital breast tomosynthesis

At A Glance

  • Researchers trained an AI system on large digital breast tomosynthesis (DBT) data sets to identify suspicious findings.
  • The system was tested by having 24 radiologists, including 13 breast subspecialists, each read 260 DBT examinations with and without AI assistance.
  • Cancer detection sensitivity and specificity increased, and reading times decreased.

Artificial intelligence (AI) helps improve the efficiency and accuracy of an advanced imaging technology used to screen for breast cancer, according to a new study published in the journal Radiology: Artificial Intelligence.

Digital breast tomosynthesis (DBT) is an advanced method for cancer detection in which an X-ray arm sweeps over the breast, taking multiple images in a matter of seconds.

Research has shown that DBT improves cancer detection and reduces false-positive recalls compared to screening with digital mammography (DM) alone. However, the DBT exam can take almost twice as long to interpret as DM due to the time it takes for the radiologist to scroll through all the images. This increased time is likely to be more consequential as DBT increasingly becomes the standard-of-care for mammographic imaging.

Emily F. Conant, M.D{module In-article}

For the study, researchers developed a deep learning system, a type of AI that can mine vast amounts of data to find subtle patterns beyond human recognition. They trained the AI system on large DBT data sets to identify suspicious findings in the DBT images.

After developing and training the system, the researchers tested its performance by having 24 radiologists, including 13 breast subspecialists, each read 260 DBT examinations with and without AI assistance. The examinations included 65 cancer cases.

Use of AI was associated with improved accuracy and shorter reading times. Sensitivity increased from 77 percent without AI to 85 percent with it. Specificity increased from 62.7 percent without AI to 69.6 percent with it. The recall rate for non-cancers, or the rate at which women were called back for follow-up examinations based on benign findings, decreased from 38 percent without AI to just 30.9 percent with it. On average, reading time decreased from just over 64 seconds without AI to only 30.4 seconds with it.

“Overall, readers were able to increase their sensitivity by 8 percent, lower their recall rate by 7 percent and cut their reading time in half when using AI concurrently while reading DBT cases compared to reading without using AI,” said study lead author Emily F. Conant, M.D., professor and chief of breast imaging from the Department of Radiology at the Perelman School of Medicine at the University of Pennsylvania in Philadelphia.

Also showing improvement was the area under the receiver operating characteristic curve (AUC), a graphing variable that combines sensitivity and specificity into a single measure for a better representation of overall radiologist performance. Radiologist performance, measured by mean AUC, increased from 0.795 without AI to 0.852 with AI.

“We know that DBT imaging increases cancer detection and lowers recall rate when added to 2-D mammography and even further improvement in these key metrics is clinically very important,” Dr. Conant said. “And, since adding DBT to the 2-D mammogram approximately doubles radiologist reading time, the concurrent use of AI with DBT increases cancer detection and may bring reading times back to about the time it takes to read DM-alone exams.”

The researchers expect the deep learning approach to improve as it is exposed to larger and larger data sets, making its potential impact on patient care even more significant.

“The results of this study suggest that both improved efficiency and accuracy could be achieved in clinical practice using an effective AI system,” Dr. Conant said.

Kovács delivers an alternative method for exoplanet stability analysis

Tyler O'Neal, Staff Editor APPLICATIONS

Starting with observational data and scalar time series, this method uses complex network topology to deduce the underlying dynamics of systems in a fast, efficient way without the need for n-body simulations

Exoplanets revolving around distant stars are coming quickly into focus with advanced technology like the Kepler space telescope. Gaining a full understanding of those systems is difficult because the initial positions and velocities of the exoplanets are unknown. Determining whether the system dynamics are quasi-periodic or chaotic is cumbersome, expensive and computationally demanding.

In this week's Chaos, from AIP Publishing, Tamás Kovács delivers an alternative method for stability analysis of exoplanetary bodies using only the observed time-series data to deduce dynamical measurements and quantify the unpredictability of exoplanet systems.

"If we don't know the governing equations of the motion of a system, and we only have the time series -- what we measure with the telescope -- then we want to transform that time series into a complex network. In this case, it is called a recurrence network," Kovács said. "This network holds all of the dynamical features of the underlying system we want to analyze." Gaining a full understanding of systems of exoplanets and distant stars is difficult, because the initial positions and velocities of the exoplanets are unknown. Determining whether the system dynamics are quasi-periodic or chaotic is cumbersome, expensive and computationally demanding. In this week's Chaos, Tamás Kovács delivers an alternative method for stability analysis of exoplanetary bodies using only the observed time series data to deduce dynamical measurements and quantify the unpredictability of exoplanet systems. This image shows a stability map of Saturn obtained by chaos indicator MEGNO (a) and recurrence network measures average path length (b) and transitivity (c). The latter two panels are based on transit timing variation of Jupiter and the radial velocity of the sun, respectively.{module In-article}

The paper draws on the work of physicist Floris Takens, who proposed in 1981 that the dynamics of a system could be reconstructed using a series of observations about the state of the system. With Takens' embedding theorem as a starting point, Kovács uses time-delay embedding to reconstruct a high-dimensional trajectory and then identify recurrence points, where bodies in the phase space are close to each other.

"Those special points will be the vertices and the edges of the complex network," Kovács said. "Once you have the network, you can reprogram this network to be able to apply measures like transitivity, average path length or others unique to that network."

Kovács tests the reliability of the method using a known system as a model, the three-body system of Saturn, Jupiter and the sun, and then applies it to the Kepler 36b and 36c system. His Kepler system results agree with what is known.

"Earlier studies pointed out that Kepler 36b and 36c is a very special system because, from the direct simulation and the numerical integrations, we see the system is at the edge of the chaos," Kovács said. "Sometimes, it shows regular dynamics, and at other times, it seems to be chaotic."

The author plans to next apply his methods to systems with more than three bodies, testing its scalability and exploring its ability to handle long time series and sharper datasets.

Japanese scientists start travelling towards a quantum internet at light speed

Tyler O'Neal, Staff Editor APPLICATIONS

Specialists successfully transferred and verified the angular momentum basis of quantum information from laser light to an electron trapped on a quantum dot. This work is a major step towards realizing hacker-proof interconnected quantum supercomputers

A research team led by Osaka University demonstrated how information encoded in the circular polarization of a laser beam can be translated into the spin state of an electron in a quantum dot, each being a quantum bit and a quantum computer candidate. The achievement represents a major step towards a "quantum internet," in which future computers can rapidly and securely send and receive quantum information.

Quantum supercomputers have the potential to vastly outperform current systems because they work in a fundamentally different way. Instead of processing discrete ones and zeros, quantum information, whether stored in electron spins or transmitted by laser photons, can be in a superposition of multiple states simultaneously. Moreover, the states of two or more objects can become entangled, so that the status of one cannot be completely described without this other. Handling entangled states allow quantum supercomputers to evaluate many possibilities simultaneously, as well as transmit information from place to place immune from eavesdropping. This is a schematic image of the spin detection of a circularly polarized photon exciting an electron spin. The yellow nano-fabricated metal electrodes form the pockets required to trap the electrons, move them, and sense them.{module In-article}

However, these entangled states can be very fragile, lasting only microseconds before losing coherence. To realize the goal of a quantum internet, over which coherent light signals can relay quantum information, these signals must be able to interact with electron spins inside distant computers.

Researchers led by Osaka University used laser light to send quantum information to a quantum dot by altering the spin state of a single electron trapped there. While electrons don't spin in the usual sense, they do have angular momentum, which can be flipped when absorbing circularly polarized laser light.

"Importantly, this action allowed us to read the state of the electron after applying the laser light to confirm that it was in the correct spin state," says first author Takafumi Fujita. "Our readout method used the Pauli exclusion principle, which prohibits two electrons from occupying the exact same state. On the tiny quantum dot, there is only enough space for the electron to pass the so-called Pauli spin blockade if it has the correct spin."

Quantum information transfer has already been used for cryptographic purposes. "The transfer of superposition states or entangled states allows for completely secure quantum key distribution," senior author Akira Oiwa says. "This is because any attempt to intercept the signal automatically destroys the superposition, making it impossible to listen in without being detected."

The rapid optical manipulation of individual spins is a promising method for producing a quantum nano-scale general supercomputing platform. An exciting possibility is that future supercomputers may be able to leverage this method for many other applications, including optimization and chemical simulations.