The use of AI-based processing of bone-marrow smears is being explored to assist in the diagnosis of leukemia. This involves the extraction of single-cell images from high-resolution data through unsupervised learning methods. The cell images are then analyzed using neural networks to identify any visual anomalies that may have a genetic origin. The neural network highlights important areas for decision-making using "explainable AI" strategies, which are indicated by colored highlights. The credit for this image goes to AG Risse.
The use of AI-based processing of bone-marrow smears is being explored to assist in the diagnosis of leukemia. This involves the extraction of single-cell images from high-resolution data through unsupervised learning methods. The cell images are then analyzed using neural networks to identify any visual anomalies that may have a genetic origin. The neural network highlights important areas for decision-making using "explainable AI" strategies, which are indicated by colored highlights. The credit for this image goes to AG Risse.
Tyler O'Neal, Staff Editor ACADEMIA

Harnessing the power of artificial intelligence in leukemia diagnosis

Groundbreaking study presents a new way to predict genetic aberrations, bringing hope to leukemia patients

In a remarkable breakthrough, researchers from the University of Münster and the University Hospital Münster in Germany have unveiled an innovative method that utilizes artificial intelligence (AI) to predict important genetic features of acute myeloid leukemia (AML). This cutting-edge approach holds immense potential for revolutionizing the field of diagnostics and guiding targeted treatments at an early stage, offering renewed hope to those battling this highly aggressive form of leukemia.

To effectively treat AML patients, timely access to genetic information is crucial. However, conventional genetic testing can be both expensive and time-consuming, resulting in a significant gap between diagnosis and the availability of critical genetic data. This gap has led a team of IT specialists and physicians to develop a transformative solution that could bridge this divide.

Published in the esteemed journal "Blood Advances," their study showcases how AI, combined with high-resolution microscopic images of bone marrow smears, can accurately predict various genetic aberrations. The team extracted genetic information from massive datasets, comprising multi-gigabyte scans from over 400 AML patients. These high-resolution images, with pixel sizes of 270,000 by 135,000 on average, provided the necessary foundation for the deep learning algorithms.

Under the leadership of Prof. Benjamin Risse, the team developed a groundbreaking deep-learning method that enables the automatic recognition of genetic features and intricate patterns present in cytological images. The algorithm effectively categorizes different cell types and identifies genetic aberrations that may elude human observers. The AI system's exceptional sensitivity allows it to detect faint patterns and hidden textures that would typically go unnoticed by the human eye.

What sets this method apart is its end-to-end AI pipeline, unifying unsupervised, self-supervised, and supervised learning processes. The integration of these approaches significantly reduces manual preliminary work, providing real-time monitoring of results. When faced with problematic cell images, the AI system works in synergy with human experts to ensure accuracy, enhancing the overall effectiveness of the diagnostic process.

While this groundbreaking method cannot replace genetic analyses, it serves as a vital tool in the early stages of diagnosis for leukemia patients. For individuals with aggressive forms of the disease, who cannot afford delays associated with complete genetic analyses, this approach offers crucial insights into the underlying genetic abnormalities.

The implications of this study reach beyond leukemia treatment. The researchers envision a future where digital methods and artificial intelligence play an increasingly pivotal role in analyzing large medical datasets. The ability to make personalized treatment recommendations for patients with malignant diseases could be revolutionized, laying the groundwork for similar breakthroughs in the diagnosis of other bone marrow disorders.

The research project received funding from the European Union and the German Research Foundation, highlighting the collaborative efforts and dedication of the scientific community in combating complex medical challenges.

It is inspiring to witness how AI and cutting-edge technology are transforming the landscape of medical diagnostics. With every breakthrough, we inch closer to personalized and targeted treatment strategies that offer hope, strength, and ultimately, a brighter future for patients facing daunting health battles.

Figure 1 displays the observed gravitational anomaly from 2,463 pure wide binaries that are free of hidden additional companions. The left panel shows the anomaly that was derived from the algorithm calculating kinematic acceleration, while the right panel shows the anomaly directly from the observed sky-projected relative velocities between the two stars with respect to the sky-projected separations.
Figure 1 displays the observed gravitational anomaly from 2,463 pure wide binaries that are free of hidden additional companions. The left panel shows the anomaly that was derived from the algorithm calculating kinematic acceleration, while the right panel shows the anomaly directly from the observed sky-projected relative velocities between the two stars with respect to the sky-projected separations.

South Korean study finds modified gravity in low acceleration situations with wide binary stars

Tyler O'Neal, Staff Editor ACADEMIA

A recent study conducted by Kyu-Hyun Chae, a professor of physics and astronomy at Sejong University in Seoul, South Korea, has reinforced the evidence for modified gravity and its idiosyncratic breakdown at low acceleration. The study analyzed gravitational anomalies that were reported back in 2023 on widely separated or long-period binary stars, known as wide binaries. The data for the study was gathered by the European Space Agency's Gaia space telescope.

The findings of Chae's study hold great promise for theoretical physicists and mathematicians, as the breakdown of standard gravity at low acceleration has significant implications for astrophysics and cosmology. The team's analysis revealed that orbital motions in binaries experience larger accelerations than what is predicted by Newtonian laws in gravitational systems weaker than about 1 nanometer per second squared. The acceleration boost factor increases to about 1.4 at accelerations lower than about 0.1 nanometers per second squared.

The study also found that the observed gravitational anomaly is remarkably consistent with the Modified Newtonian Dynamics (MOND) gravity phenomenology. However, the underlying theoretical possibilities that encompass the MOND gravity phenomenology are open, which may be welcome news to theoretical physicists and mathematicians.

In previous studies, researchers have suggested that the gravitational acceleration of observed objects can only be explained by introducing dark matter. However, a new study has found that the required dark matter density exceeds what has been observed in galactic dynamics and cosmological observations. As a result, this provides further evidence for modified gravity.

The research focused on a sample of "pure" wide binaries, which were selected by removing all systems that could potentially harbor unobserved additional stars. The team selected up to 2,463 pure binaries, which was less than 10% of the sample used in the earlier study, to avoid potential errors in supercomputing hidden additional gravity effects.

The study's two different algorithms to test gravity produced consistent results that agree with previously reported gravitational anomalies. The observed acceleration or relative velocity between two stars naturally satisfies the Newton-Einstein standard gravity at small separation or high acceleration. However, the acceleration or relative velocity deviates from the Newtonian prediction at a separation of about 2,000 AU and an acceleration of about 1 nanometer per second squared. There is then a consistent boost of about 40 to 50% in acceleration or 20% boost in relative velocity at separation greater than about 5,000 AU or acceleration lower than about 0.1 nanometer per second squared up to the probed limit of about 20,000 AU or 0.01 nanometer per second squared.

The research concludes that at least three independent quantitative analyses by two independent groups have revealed essentially the same gravitational anomaly. Therefore, the gravitational anomaly is real, and a new scientific paradigm shift is on its way.

In conclusion, the study provides us with direct evidence that modified gravity at low acceleration is a reality. These findings have implications for theoretical physicists, mathematicians, and everyone involved in the study of our universe.

Mika Gustafsson, professor. Credit goes to Thor Balkhed
Mika Gustafsson, professor. Credit goes to Thor Balkhed

Severe MS predicted using machine learning: A breakthrough in personalized treatment

Tyler O'Neal, Staff Editor ACADEMIA

In a groundbreaking study, researchers from Linköping University, the Karolinska Institute, and the University of Skövde in Sweden have made significant progress in predicting the long-term disability outcomes in patients with multiple sclerosis (MS) using machine learning. By analyzing a combination of just 11 proteins, the team has developed a tool that can tailor treatments based on the expected severity of the disease for individual patients.

Multiple sclerosis, a chronic autoimmune disease, affects millions of people worldwide. The immune system of MS patients attacks the body's own nerves, leading to damage in the brain and spinal cord. The primary target of this attack is myelin, a fatty compound that surrounds and insulates nerve axons. When the myelin is damaged, the transmission of electrical signals becomes less efficient, resulting in various neurological symptoms.

One of the major challenges in treating MS is the considerable variation in disease progression from person to person. Early detection of those who are likely to experience a more severe disease course is crucial for providing timely and effective treatments. To address this challenge, the research team sought to identify early markers that could predict disease severity using cutting-edge machine learning techniques.

The study involved analyzing nearly 1,500 proteins in samples from 92 patients suspected or recently diagnosed with MS. By combining this data with information from their medical records and advanced machine learning algorithms, the researchers successfully identified a panel of 11 proteins that accurately predicted disease progression. This streamlined approach not only enhances convenience but also reduces the cost of analysis, making it more accessible for further research and potential clinical applications.

Dr. Mika Gustafsson, the lead researcher and professor of bioinformatics at the Department of Physics, Chemistry, and Biology at Linköping University, believes that their work brings us one step closer to a tool that can guide clinicians in selecting more effective treatments for patients in the early stages of the disease. However, he also highlights the need to strike a balance, as some patients may not require aggressive treatment and could be spared the potential side effects and costs.

The research team also discovered a specific protein called neurofilament light chain (NfL), which has proven to be a reliable biomarker for short-term disease activity. The presence of this protein indicates nerve damage and correlates with the disease's level of activity. This finding not only confirms earlier research but also provides valuable insight into monitoring disease progression and response to treatment.

An essential strength of this study lies in the extensive validation conducted. The combination of proteins identified in the patient group at Linköping University Hospital was confirmed in a separate group of MS patients at the Karolinska University Hospital in Stockholm. This cross-validation enhances the reliability of the findings and underscores their significance.

The implications of this research are immense, offering better insights into individualized treatment plans and improving the quality of life for MS patients. By utilizing machine learning and state-of-the-art protein analysis technologies, physicians can now make more informed decisions regarding the most suitable treatment strategies. Tremendous progress has been made toward early intervention and personalized care for those living with MS.

This study was funded by various organizations, including the Swedish Foundation for Strategic Research, the Swedish Brain Foundation, the Knut and Alice Wallenberg Foundation, and the Swedish Research Council.

As this groundbreaking research continues to evolve, scientists and medical professionals are hopeful that it will pave the way for a future where early detection and personalized treatment will significantly improve the lives of individuals battling multiple sclerosis.