Computers Math

Significant advances in artificial intelligence (AI) over the past decade have relied upon extensive training of algorithms using massive, open-source databases. But when such datasets are used “off label” and applied in unintended ways, the results are subject to machine learning bias that compromises the integrity of the AI algorithm, according to a new study by researchers at the University of California, Berkeley, and the University of Texas at Austin.
The findings, published this week in the Proceedings of the National Academy of Sciences, highlight the problems that arise when data published for one task are used to train algorithms for a different one.
The researchers noticed this issue when they failed to replicate the promising results of a medical imaging study. “After several months of work, we realized that the image data used in the paper had been preprocessed,” said study principal investigator Michael Lustig, UC Berkeley professor of electrical engineering and computer sciences. “We wanted to raise awareness of the problem so researchers can be more careful and publish results that are more realistic.”
The proliferation of free online databases over the years has helped support the development of AI algorithms in medical imaging. For magnetic resonance imaging (MRI), in particular, improvements in algorithms can translate into faster scanning. Obtaining an MR image involves first acquiring raw measurements that code a representation of the image. Image reconstruction algorithms then decode the measurements to produce the images that clinicians use for diagnostics.
Some datasets, such as the well-known ImageNet, include millions of images. Datasets that include medical images can be used to train AI algorithms used to decode the measurements obtained in a scan. Study lead author Efrat Shimron, a postdoctoral researcher in Lustig’s lab, said new and inexperienced AI researchers may be unaware that the files in these medical databases are often preprocessed, not raw.
As many digital photographers know, raw image files contain more data than their compressed counterparts, so training AI algorithms on databases of raw MRI measurements is important. But such databases are scarce, so software developers sometimes download databases with processed MR images, synthesize seemingly raw measurements from them, and then use those to develop their image reconstruction algorithms. More

One of the ultimate goals of medical science is to develop personalized disease diagnostics and therapeutics. With a patient’s genetic information, doctors could tailor treatments to individuals, leading to safer and more effective care.
Recent work from a team of Northwestern Engineering researchers has moved the field closer to realizing this future.
Led by Professor Horacio Espinosa, the research team developed a new version of its Nanofountain Probe Electroporation (NFP-E), a tool used to deliver molecules into single-cells using electricity. The enhanced method leverages artificial intelligence (AI) to execute cell engineering tasks such as cell nuclei localization and probe detection. Other processes such as probe motion, probe-to-cell contact detection, and electroporation-mediated delivery of foreign cargo into single cells are also automated, minimizing user intervention.
“NFP-E can handle small starting samples without any significant cell loss in the entire protocol,” said Espinosa, James N. and Nancy J. Farley Professor in Manufacturing and Entrepreneurship at the McCormick School of Engineering and the study’s corresponding author. “This is an advantage over other cell engineering methods such as bulk electroporation, which require millions of cells and lead to significant cell losses. The automated NFP-E, combined with its ability to selectively target and manipulate single cells in micro-arrays, can be useful in fundamental research, such as deciphering intracellular dynamics and cell-to-cell communication studies as well as biological applications such as cell line generation.”
Espinosa and graduate students Prithvijit Mukherjee, Cesar A. Patino, and Nibir Pathak reported their work in the paper “Deep Learning Assisted Automated Single Cell Electroporation Platform for Effective Genetic Manipulation of Hard-to-Transfect Cells” published March 21 in Small.
“Genetic manipulation of human induced pluripotent stem cells (hiPSCs) by introducing exogenous cargo has a wide range of applications in disease diagnostics, therapeutic discovery, and regenerative medicine,” said Mukherjee, a PhD student in the Espinosa group who is joining the microfluidics group at Illumina. More

In a world first, a team co-led by a physicist at City University of Hong Kong (CityU) has discovered that excitons — excited electrons bound to empty electron “holes” — can exist stably and travel rapidly through metal. Because excitons can be generated by energy from light and have no electrical charge, this discovery makes them potential candidates as a higher-speed alternative to free electrons as a carrier of digital information.
Excitons form when certain materials absorb energy from light to excite electrons, the negatively charged particles in atoms. The electrons are boosted to a higher energy level to leave positively charged spaces or “holes” in their original position. Owing to electrostatic attraction, a hole and an excited electron can pair up without recombining, forming an exciton that behaves like an uncharged particle.
“When an exciton’s electron recombines with a hole, energy is emitted as light, which could be harnessed for data transfer in the optoelectronics industry,” says team co-leader Dr Ma Junzhang, Assistant Professor in the CityU Department of Physics. “Excitons would be better data carriers than free electrons, whose negative charge slows them down, but excitons are very unstable, especially in metals. In fact, before our study, stable and mobile excitons were thought to be impossible in metals.”
The researchers succeeded in generating and detecting excitons in metal because of a combination of optimal test conditions and unique characteristics of their chosen material, tantalum triselenide, TaSe3. The research was headed by CityU and the Paul Scherrer Institute (PSI) in Switzerland, and the results were published in Nature Materials in an article titled “Multiple mobile excitons manifested as sidebands in quasi-one-dimensional metallic TaSe3.” The joint corresponding authors of the paper were Dr Ma Junzhang, and Professor Shi Ming and Dr Markus Müller from PSI. Collaborators included researchers from Rutgers University, Princeton University, Stanford University, and other institutions.
Importance of excitons as robust information carriers
The exciton is expected to play an important role in the future of information transmission thanks to both its charge neutrality and ability to move through a solid. Unlike negatively charged free electrons, excitons are unhindered by external electric fields, magnetic fields, and defects in the surrounding material. More

A research team reported the direct measurement of dielectric tensors of anisotropic structures including the spatial variations of principal refractive indices and directors. The group also demonstrated quantitative tomographic measurements of various nematic liquid-crystal structures and their fast 3D nonequilibrium dynamics using a 3D label-free tomographic method. The method was described in Nature Materials.
Light-matter interactions are described by the dielectric tensor. Despite their importance in basic science and applications, it has not been possible to measure 3D dielectric tensors directly. The main challenge was due to the vectorial nature of light scattering from a 3D anisotropic structure. Previous approaches only addressed 3D anisotropic information indirectly and were limited to two-dimensional, qualitative, strict sample conditions or assumptions.
The research team developed a method enabling the tomographic reconstruction of 3D dielectric tensors without any preparation or assumptions. A sample is illuminated with a laser beam with various angles and circularly polarization states. Then, the light fields scattered from a sample are holographically measured and converted into vectorial diffraction components. Finally, by inversely solving a vectorial wave equation, the 3D dielectric tensor is reconstructed.
Professor YongKeun Park said, “There were a greater number of unknowns in direct measuring than with the conventional approach. We applied our approach to measure additional holographic images by slightly tilting the incident angle.”
He said that the slightly tilted illumination provides an additional orthogonal polarization, which makes the underdetermined problem become the determined problem. “Although scattered fields are dependent on the illumination angle, the Fourier differentiation theorem enables the extraction of the same dielectric tensor for the slightly tilted illumination,” Professor Park added.
His team’s method was validated by reconstructing well-known liquid crystal (LC) structures, including the twisted nematic, hybrid aligned nematic, radial, and bipolar configurations. Furthermore, the research team demonstrated the experimental measurements of the non-equilibrium dynamics of annihilating, nucleating, and merging LC droplets, and the LC polymer network with repeating 3D topological defects.
“This is the first experimental measurement of non-equilibrium dynamics and 3D topological defects in LC structures in a label-free manner. Our method enables the exploration of inaccessible nematic structures and interactions in non-equilibrium dynamics,” first author Dr. Seungwoo Shin explained.
Story Source:
Materials provided by The Korea Advanced Institute of Science and Technology (KAIST). Note: Content may be edited for style and length. More

Researchers at Linköping University have, by way of a number of theoretical calculations, shown that magnesium diboride becomes superconductive at a higher temperature when it is stretched. The discovery is a big step toward finding superconductive materials that are useful in real-world situations.
“Magnesiumdiboride or MgB2 is an interesting material. It’s a hard material that is used for instance in aircraft production and normally it becomes superconductive at a relatively high temperature, 39 K, or -234 C°,” says Erik Johansson, who recently completed his doctorate at the Division of Theoretical Physics.
Erik Johansson is also principal author of an article published in the Journal of Applied Physics that have attracted broad attention. The results have been identified by the editor as particularly important for the future.
“Magnesium boride has an uncomplicated structure which means that the calculations on the supercomputers here at the National Supercomputer Centre in Linköping can focus on complex phenomena like superconductivity,” he says.
Access to renewable energy is fundamental for a sustainable world, but even renewable energy disappears in the form of losses during transmission in the electrical networks. These losses are due to the fact that even materials that are good conductors have a certain resistance, resulting in losses in the form of heat. For this reason, scientists worldwide are trying to find materials that are superconductive, that is, that conduct electricity with no losses at all. Such materials exist, but superconductivity mostly arises very close to absolute 0, i.e. 0 K or -273,15 °C. Many years of research have resulted in complicated new materials with a maximum critical temperature of maybe 200 K, that is, -73 °C. At temperatures under the critical temperature, the materials become superconductive. Research has also shown that superconductivity can be achieved in certain metallic materials at extremely high pressure.
If the scientists are successful in increasing the critical temperature, there will be greater opportunities to use the phenomenon of superconductivity in practical applications.
“The main goal is to find a material that is superconductive at normal pressure and room temperature. The beauty of our study is that we present a smart way of increasing the critical temperature without having to use massively high pressure, and without using complicated structures or sensitive materials. Magnesium diboride behaves in the opposite way to many other materials, where high pressure increases the ability to superconduct. Instead, here we can stretch the material by a few per cent and get a huge increase in the critical temperature,” says Erik Johansson.
In the nanoscale, the atoms vibrate even in really hard and solid materials. In the scientists’ calculations of magnesium diboride, it emerges that when the material is stretched, the atoms are pulled away from each other and the frequency of the vibrations changes. This means that in this material, the critical temperature increases — in one case from 39 K to 77 K. If magnesium diboride is instead subjected to high pressure, its superconductivity decreases.
The discovery of this phenomenon paves the way for calculations and tests of other similar materials or material combinations that can increase the critical temperature further.
“One possibility could be to mix magnesium diboride with another metal diboride, creating a nanolabyrinth of stretched MgB2 with a high superconductive temperature,” says Björn Alling, docent and senior lecturer at the Division of Theoretical Physics and director of the National Supercomputer Centre at Linköping University.
The research has been funded by the Knut and Alice Wallenberg Foundation, the Swedish Research Council and the Swedish Foundation for Strategic Research, among others. It has been conducted with support from the government’s strategic venture, Advanced Functional Materials, AFM, at Linköping University.
Story Source:
Materials provided by Linköping University. Original written by Monica Westman Svenselius. Note: Content may be edited for style and length. More

Computers, smartphones, GPS: quantum physics has enabled many technological advances. It is now opening up new fields of research in cryptography (the art of coding messages) with the aim of developing ultra-secure telecommunications networks. There is one obstacle, however: after a few hundred kilometers within an optical fiber, the photons that carry the qubits or ‘quantum bits’ (the information) disappear. They therefore need ‘repeaters’, a kind of ‘relay’, which are partly based on a quantum memory. By managing to store a qubit in a crystal (a “memory”) for 20 milliseconds, a team from the University of Geneva (UNIGE) has set a world record and taken a major step towards the development of long-distance quantum telecommunications networks. This research can be found in the journal npj Quantum Information.
Developed during the 20th century, quantum physics has enabled scientists to describe the behavior of atoms and particles as well as certain properties of electromagnetic radiation. By breaking with classical physics, these theories generated a real revolution and introduced notions without equivalent in the macroscopic world such as superposition, which describes the possibility for a particle to be in several places at once, or entanglement, which describes the ability of two particles to affect each other instantaneously even at a distance (‘spooky action at a distance’).
Quantum theories are now at the heart of much research in cryptography, a discipline that brings together techniques for encoding a message. Quantum theories make it possible to guarantee perfect authenticity and confidentiality for information (a qubit) when it is transmitted between two interlocutors by a particle of light (a photon) within an optical fiber. The phenomenon of superposition let the sender know immediately whether the photon conveying the message has been intercepted.
Memorizing the signal
However, there is a major obstacle to the development of long-distance quantum telecommunication systems: beyond a few hundred kilometers, the photons are lost and the signal disappears. Since the signal cannot be copied or amplified — it would lose the quantum state that guarantees its confidentiality — the challenge is to find a way of repeating it without altering it by creating ‘repeaters’ based, in particular, on a quantum memory.
In 2015, the team led by Mikael Afzelius, a senior lecturer in the Department of Applied Physics at the Faculty of Science of the University of Geneva (UNIGE), succeeded in storing a qubit carried by a photon for 0.5 milliseconds in a crystal (a ‘memory’). This process allowed the photon to transfer its quantum state to the atoms of the crystal before disappearing. However, the phenomenon did not last long enough to allow the construction of a larger network of memories, a prerequisite for the development of long-distance quantum telecommunications.
Storage record
Today, within the framework of the European Quantum Flagship program, Mikael Afzelius’ team has managed to increase this duration significantly by storing a qubit for 20 milliseconds. “This is a world record for a quantum memory based on a solid-state system, in this case a crystal. We have even managed to reach the 100 millisecond mark with a small loss of fidelity,” enthuses the researcher. As in their previous work, the UNIGE scientists used crystals doped with certain metals called ‘rare earths’ (europium in this case), capable of absorbing light and then re-emitting it. These crystals were kept at -273,15°C (absolute zero), because beyond 10°C above this temperature, the thermal agitation of the crystal destroys the entanglement of the atoms.
“We applied a small magnetic field of one thousandth of a Tesla to the crystal and used dynamic decoupling methods, which consist in sending intense radio frequencies to the crystal. The effect of these techniques is to decouple the rare-earth ions from perturbations of the environment and increase the storage performance we have known until now by almost a factor of 40,” explains Antonio Ortu, a post-doctoral fellow in the Department of Applied Physics at UNIGE. The results of this research constitute a major advance for the development of long-distance quantum telecommunications networks. They also bring the storage of a quantum state carried by a photon to a time scale that can be estimated by humans.
An efficient system in ten years
However, there are still several challenges to be met. “The challenge now is to extend the storage time further. In theory, it would be enough to increase the duration of exposure of the crystal to radio frequencies, but for the time being, technical obstacles to their implementation over a longer period of time prevent us from going beyond 100 milliseconds. However, it is certain that these technical difficulties can be resolved,” says Mikael Afzelius.
The scientists will also have to find ways of designing memories capable of storing more than a single photon at a time, and thus of having ‘entangled’ photons which will guarantee confidentiality. “The aim is to develop a system that performs well on all these points and that can be marketed within ten years,” concludes the researcher.
Story Source:
Materials provided by Université de Genève. Note: Content may be edited for style and length. More

An interdisciplinary team of researchers from the University of Missouri, Children’s Mercy Kansas City and Texas Children’s Hospital has used a new data-driven approach to learn more about persons with Type 1 diabetes, who account for about 5-10% of all diabetes diagnoses. The team gathered its information through health informatics and applied artificial intelligence (AI) to better understand the disease.
In the study, the team analyzed publicly available, real-world data from about 16,000 participants enrolled in the T1D Exchange Clinic Registry.By applying a contrast pattern mining algorithm developed at the MU College of Engineering, the team was able to identify major differences in health outcomes among people living with Type 1 diabetes who do or do not have an immediate family history of the disease.
Chi-Ren Shyu, the director of the MU Institute for Data Science and Informatics (MUIDSI), led the AI approach used in the study, and said the technique is exploratory in nature.
“Here we let the computer do the work of connecting millions of dots in the data to identify only major contrasting patterns between individuals with and without a family history of Type 1 diabetes, and to do the statistical testing to make sure we are confident in our results,” said Shyu, the Paul K. and Dianne Shumaker Professor in the MU College of Engineering.
Erin Tallon, a graduate student in the MUIDSI and the lead author on the study, said the team’s analysis resulted in some unfamiliar findings.
“For instance, we found individuals in the registry who had an immediate family member with Type 1 diabetes were more frequently diagnosed with hypertension, as well as diabetes-related nerve disease, eye disease and kidney disease,” Tallon said. “We also found a more frequent co-occurrence of these conditions in individuals who had an immediate family history of Type 1 diabetes. Additionally, individuals who had an immediate family history of Type 1 diabetes also more frequently had certain demographic characteristics.”
Tallon’s passion for this project began with a personal connection, and quickly grew as a result of her experience working as a nurse in an intensive critical care unit (ICU). She would often see patients with Type 1 diabetes who were also dealing with other co-existing conditions such as kidney disease and high blood pressure. Knowing that a person’s Type 1 diabetes diagnosis often occurs only when the disease is already very advanced, she wanted to find better ways for prevention and diagnosis, starting with finding a way to analyze the large amounts of publicly available data already collected about the disease. More

A new computational approach developed by researchers at The University of Texas MD Anderson Cancer Center successfully combines data from parallel gene-expression profiling methods to create spatial maps of a given tissue at single-cell resolution. The resulting maps can provide unique biological insights into the cancer microenvironment and many other tissue types.
The study was published today in Nature Biotechnology and will be presented at the upcoming American Association for Cancer Research (AACR) Annual Meeting 2022 (Abstract 2129).
The tool, called CellTrek, uses data from single-cell RNA sequencing (scRNA-seq) together with that of spatial transcriptomics (ST) assays — which measure spatial gene expression in many small groups of cells — to accurately pinpoint the location of individual cell types within a tissue. The researchers presented findings from analysis of kidney and brain tissues as well as samples of ductal carcincoma in situ (DCIS) breast cancer.
“Single-cell RNA sequencing provides tremendous information about the cells within a tissue, but, ultimately, you want to know where these cells are distributed, particularly in tumor samples,” said senior author Nicholas Navin, Ph.D., professor of Genetics and Bioinformatics & Computational Biology. “This tool allows us to answer that question with an unbiased approach that improves upon currently available spatial mapping techniques.”
Single-cell RNA sequencing is an established method to analyze the gene expression of many individual cells from a sample, but it cannot provide information on the location of cells within a tissue. On the other hand, ST assays can measure spatial gene expression by analyzing many small groups of cells across a tissue but are not capable of providing single-cell resolution.
Current computational approaches, known as deconvolution techniques, can identify different cell types present from ST data, but they are not capable of providing detailed information at the single-cell level, Navin explained.
Therefore, co-first authors Runmin Wei, Ph.D., and Siyuan He of the Navin Laboratory led the efforts to develop CellTrek as a tool to combine the unique advantages of scRNA-seq and ST assays and create accurate spatial maps of tissue samples.
Using publicly available scRNA-seq and ST data from brain and kidney tissues, the researchers demonstrated that CellTrek achieved the most accurate and detailed spatial resolution of the methods evaluated. The CellTrek approach also was able to distinguish subtle gene expression differences within the same cell type to gain information on their heterogeneity within a sample.
The researchers also collaborated with Savitri Krishnamurthy, M.D., professor of Pathology, to apply CellTrek to study DCIS breast cancer tissues. In an analysis of 6,800 single cells and 1,500 ST regions from a single DCIS sample, the team learned that different subgroups of tumor cells were evolving in unique patterns within specific regions of the tumor. Analysis of a second DCIS sample demonstrated the ability of CellTrek to reconstruct the spatial tumor-immune microenvironment within a tumor tissue.
“While this approach is not restricted to analyzing tumor tissues, there are obvious applications for better understanding cancer,” Navin said. “Pathology really drives cancer diagnoses and, with this tool, we’re able to map molecular data on top of pathological data to allow even deeper classifications of tumors and to better guide treatment approaches.”
This research was supported by the National Institutes of Health/National Cancer Institute (RO1CA240526, RO1CA236864, CA016672), the Cancer Prevention and Research Institute of Texas (CPRIT) (RP180684), the Chan Zuckerberg Initiative SEED Network Grant, and the PRECISION Cancer Grand Challenges Grant. Navin is supported by the American Association for the Advancement of Science (AAAS) Martin and Rose Wachtel Cancer Research Award, the Damon Runyon-Rachleff Innovation Award, the Andrew Sabin Family Fellowship, and the Jack and Beverly Randall Prize for Excellence in Cancer Research. Wei is supported by a Damon Runyon Quantitative Biology Fellowship Award.
Collaborating MD Anderson authors include Shanshan Bai, Emi Sei, Ph.D., and Min Hu, all of Genetics; and Ken Chen, Ph.D., of Bioinformatics. Additional authors include Alastair Thompson, M.D., of Baylor College of Medicine, Houston. The authors have no conflicts of interest. More