Integrating medical imaging and cancer biology with deep neural networks

Neural network framework may increase radiologist's confidence in assessing the type of lung cancer on CT scans, informing individualized treatment planning

2021-05-10

(Press-News.org) Despite our remarkable advances in medicine and healthcare, the cure to cancer continues to elude us. On the bright side, we have made considerable progress in detecting several cancers in earlier stages, allowing doctors to provide treatments that increase long-term survival. The credit for this is due to "integrated diagnosis," an approach to patient care that combines molecular information and medical imaging data to diagnose the cancer type and, eventually, predict treatment outcomes.

There are, however, several intricacies involved. The correlation of molecular patterns, such as gene expression and mutation, with image features (e.g., how a tumor appears in a CT scan), is commonly referred to as "radiogenomics." This field is limited by its frequent use of high-dimensional data, wherein the number of features exceeds that of observations. Radiogenomics is also plagued by several simplifying model assumptions and a lack of validation datasets. While machine learning techniques such as deep neural networks can alleviate this situation by providing accurate predictions of image features from gene expression patterns, there arises a new problem: we do not know what the model has learned.

"The ability to interrogate the model is critical to understanding and validating the learned radiogenomic associations," explains William Hsu, associate professor of radiological sciences at the University of California, Los Angeles, and director of the Integrated Diagnostics Shared Resource. Hsu's lab works on problems related to data integration, machine learning, and imaging informatics. In an earlier study, Hsu and his colleagues used a method of interpreting a neural network called "gene masking" to interrogate trained neural networks to understand learned associations between genes and imaging phenotypes. They demonstrated that the radiogenomic associations discovered by their model were consistent with prior knowledge. However, they only used a single dataset for brain tumor in their previous study, which means the generalizability of their approach remained to be determined.

Against this backdrop, Hsu and his colleagues, Nova Smedley, former graduate student and lead author, and Denise Aberle, a thoracic radiologist, have carried out a study investigating whether deep neural networks can represent associations between gene expression, histology (microscopic features of biological tissues), and CT-derived image features. They found that the network could not only reproduce previously reported associations but also identify new ones. END

[Attachments] See images for this press release:

ELSE PRESS RELEASES FROM THIS DATE:

Timing is everything in new implant tech

2021-05-10

HOUSTON - (May 10, 2021) - Implants that require a steady source of power but don't need wires are an idea whose time has come. Now, for therapies that require multiple, coordinated stimulation implants, their timing has come as well. Rice University engineers who developed implants for electrical stimulation in patients with spinal cord injuries have advanced their technique to power and program multisite biostimulators from a single transmitter. A peer-reviewed paper about the advance by electrical and computer engineer Kaiyuan Yang and his colleagues at Rice's Brown School of Engineering won the best paper award at the IEEE's Custom Integrated Circuits Conference, held virtually in the last ...

Informed tourists make whale watching wafer for whales

2021-05-10

According to the International Whaling Commission, whale-watching tourism generates more than $2.5 billion a year. After the COVID-19 pandemic, this relatively safe outdoor activity is expected to rebound. Two new studies funded by a collaborative initiative between the Smithsonian Tropical Research Institute (STRI) in Panama and Arizona State University (ASU) show how science can contribute to whale watching practices that ensure the conservation and safety of whales and dolphins. "The Smithsonian's role is to provide scientific advice to policy makers as they pioneer management strategies to promote whale conservation," said STRI marine biologist, Hector Guzmán, whose previous work led the International Maritime Organization to establish shipping corridors ...

For twins, gesture and speech go hand-in-hand in language development

2021-05-10

ATLANTA--Gestures--such as pointing or waving--go hand in hand with a child's first words, and twins lag behind single children in producing and using those gestures, two studies from Georgia State University psychology researchers show. Twins produce fewer gestures and gesture to fewer objects than other children, said principal researcher Seyda Ozcaliskan, an associate professor in the Department of Psychology. Language use also lags for twins, and language--but not gesture--is also affected by sex, with girls performing better than boys, Ozcaliskan said. "The implications are fascinating," said Ozcaliskan. ...

UM scientist joins team partnering with UN's initiative to map ungulate migrations

2021-05-10

MISSOULA - University of Montana Professor Mark Hebblewhite has joined an international team of 92 scientists and conservationists to create the first-ever global atlas of ungulate (hoofed mammal) migrations. Working in partnership with the Convention on the Conservation of Migratory Species of Wild Animals, a U.N. treaty, the Global Initiative on Ungulate Migration (GIUM) launches May 7 with the publication of a commentary in Science titled " END ...

Brain regions involved in vision also encode how to hold tools

2021-05-10

Visual brain areas involved in processing hands also encode information about the correct way to hold tools, according to new research published in JNeurosci. Each part of the brain's visual system activates in response to a certain type of item -- whether it's faces, tools, objects, or hands. Scientists assumed the brain segregates visual information in this way to optimize motor actions with tools. Yet most studies investigating the brain mechanisms for tool grasping used images of tools or hands, and no actual hand movements were performed. Knights et al. used fMRI to measure the brain activity of participants as they grasped 3D-printed kitchen tools (spoon, knife, and pizza cutter) and similar-sized non-tools. ...

UEA team reads minds to understand human tool use

2021-05-10

Researchers at the University of East Anglia have made an astonishing discovery about how our brains control our hands. They used MRI data to study which parts of the brain are used when we handle tools, such as a knives. They read out the signal from certain brain regions and tried to distinguish when participants handled tools appropriately for use. Humans have used tools for millions of years, but this research is the first to show that actions such grasping a knife by its handle for cutting are represented by brain areas that also represent images of human hands, our primary 'tool' for interacting with the world. The research ...

'Unmaking' a move: Correcting motion blur in single-photon images

2021-05-10

Imaging technology has come a long way since the beginning of photography in the mid-19th century. Now, many state-of-the-art cameras for demanding applications rely on mechanisms that are considerably different from those in consumer-oriented devices. One of these cameras employs what is known as "single-photon imaging," which can produce vastly superior results in dark conditions and fast dynamic scenes. But how does single-photon imaging differ from conventional imaging? When taking a picture with a regular CMOS camera, like the ones on smartphones, ...

Monash study may help boost peptide design

2021-05-10

Peptides " short strings of amino acids" play a vital role in health and industry with a huge range of medical uses including in antibiotics, anti-inflammatory and anti-cancer drugs. They are also used in the cosmetics industry and for enhancing athletic performance. Altering the structure of natural peptides to produce improved compounds is therefore of great interest to scientists and industry. But how the machineries that produce these peptides work still isn't clearly understood. Associate Professor Max Cryle from Monash University's Biomedicine Discovery Institute (BDI) has revealed a key aspect of peptide machinery in a paper published in Nature Communications today that provides a key to the "Holy Grail" of re-engineering peptides.. The findings will advance his lab's work into ...

Should we panic over declining sperm counts? Harvard researchers say not so fast

2021-05-10

A new study from the Harvard GenderSci Lab in the journal Human Fertility, "The Future of Sperm: A Biovariability Framework for Understanding Global Sperm Count Trends" questions the panic over apparent trends of declining human sperm count. Recent studies have claimed that sperm counts among men globally, and especially from "Western" countries, are in decline, leading to apocalyptic claims about the possible extinction of the human species. But the Harvard paper, by Marion Boulicault, Sarah S. Richardson, and colleagues, reanalyzes claims of precipitous human sperm declines, re-evaluating evidence presented in the widely-cited 2017 meta-analysis by Hagai Levine, Shanna ...

As global climate shifts, forests' futures may be caught in the wind

2021-05-10

Berkeley -- Forests' ability to survive and adapt to the disruptions wrought by climate change may depend, in part, on the eddies and swirls of global wind currents, suggests a new study by researchers at the University of California, Berkeley. Unlike animals, the trees that make up our planet's forests can't uproot and find new terrain if conditions get tough. Instead, many trees produce seeds and pollen that are designed to be carried away by the wind, an adaptation that helps them colonize new territories and maximize how far they can spread their genes. The ...