Detecting Brain Imaging Anomalies Using Generative AI

Post by Amanda Engstrom

The takeaway

Generative Artificial Intelligence (AI) has become a useful tool for synthesizing large brain imaging datasets and detecting pathological anomalies, but not without error. The introduction of metrics that focus on evaluating normative representation of healthy brain tissue can increase anomaly detection and diagnosis.

What's the science?

The advancement of medical imaging technologies has increased doctors’ ability to diagnose a variety of diseases. Still, it has created the challenge of how to integrate and analyze large volumes of complex imaging data. To capture the complexity and rarity of human pathologies, generative AI has been harnessed for the automated detection of pathological anomalies. Normative representation learning in the brain aims to understand the typical anatomy of the brain using large human datasets. This week in Nature Communications, Bercea and colleagues test three novel metrics that evaluate normative representation in generative AI models, focusing on understanding typical anatomical distributions in healthy individuals, and tested them against various brain pathologies.

How did they do it?

The authors propose three metrics that evaluate the quality of the pseudo-healthy restorations by generative AI models. These metrics are: 

1) Restoration Quality Index (RQI) which evaluates the perceived quality of the synthesized images, 

2) Anomaly to Healthy Index (AHI), measuring the closeness of the distribution of restored pathological images to a healthy reference set

3) Healthy Conservation and Anomaly Correction Index (CACI) measures how the model can both maintain the integrity of healthy regions as well as correct anomalies in pathological areas.

The authors used these metrics to evaluate current generative AI frameworks to assess the ability of each model to learn and apply normative representations to their images. Models were trained on over 500 healthy scans, and evaluated using two datasets that encompassed a wide spectrum of brain pathologies. After evaluating the performance of each model based on normative learning metrics, the authors then determined the relationship between this ranking and anomaly detection metrics.

Finally, the authors performed a clinical evaluation of their metrics with 16 radiologists. Experts were given 180 randomized images (30 pathology-free originals and 30 generated from 5 different AI models) and asked to rate each image for ‘Realness’, ‘Image Quality’, and ‘Health Status’. These evaluations helped evaluate the effectiveness of both the new metrics as well as help measure the clinical relevance of the learned representations.

What did they find?

After applying their three normative learning metrics, the authors found that each individual metric offers a unique perspective on the performance of the AI models. Methods that simply replicate input images, like autoencoders, have a high RQI but score poorly on AHI and CACI. However, models that remove anomalies such as variational autoencoders or latent transfer models, have improved CACI, but poor RQI because the output images are typically blurry. The AHI metric was the most challenging for all models. Guided restoration techniques using intelligent masking tend to achieve the highest overall scores. When all three metrics (RQI, AHI, and CACI) were collectively optimized, those AI models demonstrated enhanced predictive anomaly detection power highlighting the importance of balancing all three metrics rather than relying on one individually.

When clinically validating AI-generated images with radiologists, there was no significant difference between AI-generated and real images. Even real non-pathological images showed variability in scoring, particularly in health score. Real images scored only marginally higher in ‘Realness’ scores. Models such as AutoDDPM and the RA method, both scored within the top 5 for normative learning and, received scores similar to the real images in ‘Realness’ and ‘Health’ respectively. The comprehensive clinical validation concluded that the proposed RQI, and to a lesser extent the AHI (CACI could not be evaluated in this study design), correlated well with clinical assessments.

What's the impact?

This study found that generative AI models that score highly in normative learning metrics can more aptly detect diverse brain pathologies and are more proficient at anomaly detection. These metrics provide a framework for evaluating AI models with greater clinical relevance. Advanced AI medical imaging is an advantageous diagnostic tool to assist clinicians in increasing workflow efficiency, diagnostic accuracy and ultimately improving patient care. 

Access the original scientific publication here.

Microplastics Identified in Human Brain Tissue

Post by Meagan Marks

The takeaway

Microplastics and nanoplastics were found in postmortem human brain tissue at concentrations significantly higher than other organ systems analyzed, calling for further research into how these particles accumulate and affect neurological and psychiatric health.

What's the science?

Over the past half-century, the prevalence of microplastics and nanoplastics (MNPs) in our environment has risen exponentially, leading to widespread pollution and potentially harmful effects on our health. MNPs are produced when plastic products—such as clothing, food packaging, and automobile parts—break down into tiny, non-biodegradable polymers, entering ecosystems, food, water, and eventually, our bodies. 

Preclinical studies have found MNPs in the organs of animals, and have linked their presence to inflammation, toxicity, and disease. However, the implications for human health remain unclear, especially regarding the average levels of MNPs within the human body, and how they distribute across organ systems. This week in Nature Medicine, Campen and colleagues analyze postmortem human brain, liver, and kidney tissue to assess the relative concentration of MNPs in the brain and how they compare to other organ systems within the body. 

How did they do it?

To measure the concentration of MNPs in human brain, liver, and kidney tissue, the authors used pyrolysis gas chromatography-mass spectrometry, a precise technique for detecting and identifying micro and nanoparticles. They first isolated the MNPs from each sample by chemically digesting the tissue, leaving behind non-biodegradable products. The remaining solids were then compacted into a pellet and subjected to several analytical steps to determine both the quantity and identity of the MNPs present. This approach allowed the authors to compare the concentrations of various plastic types within each sample and across different subjects and organ systems. Notably, the authors included samples collected in both 2016 and 2024 to identify trends in MNP levels over the past eight years. Additionally, brain samples from patients with dementia were analyzed to assess MNP levels in neurologically diseased brains. It’s also important to note that all brain samples were taken from the frontal cortex, so more work is needed to explore MNP distributions across other brain regions. 

What did they find?

After comparing samples from the liver, kidney, and brain, the authors found that the brain contained 7-30 times as many MNPs (with a median of 3,345 micrograms per gram of sample) as compared to the liver and kidney (with median values of 433 and 404 micrograms per gram of sample, respectively). The majority of the MNPs appeared as plastic shards or flakes and were made of a specific plastic called polyethylene, a common polymer found in food packaging, bottles, and automobile parts. The brain had the highest concentration of this plastic, which made up about 75% of all MNPs. 

Interestingly, liver and brain samples from 2024 showed significantly higher concentrations of MNPs compared to the 2016 samples, with certain types of plastic—including polyethylene, polypropylene, polyvinyl chloride, and styrene butadiene rubber—specifically increasing. The total mass of these concentrations within samples had increased by 50% over the past 8 years, suggesting that environmental MNPs may be growing and leading to higher uptake by our bodies. 

Additionally, dementia patients had significantly higher MNPs than other brain samples, with a median value of over 26,000 micrograms per gram of sample. This is likely due to a more permeable blood-brain barrier and impaired clearance mechanisms, which are hallmarks of the disease. 

What's the impact?

This study found that the brain contains significantly higher levels of MNPs compared to the liver and kidney. While this discovery is important, there is still much work to be done, particularly in understanding how MNPs are taken up and dispersed throughout the brain, how they concentrate in different regions, and how they are cleared from the system. Given the rising levels of MNPs in our environment, it is crucial to investigate their potential role in neurological and psychiatric health.

Access the original scientific publication here 

How the Brain Regulates the Body’s Immune Response

Post by Meredith McCarty

The takeaway

The brain can modulate immune responses to prepare the immune system for potential threats or even induce placebo-like effects. This research reveals the functional circuitry underlying the coordination of behavioral and immunological responses in the mouse brain. 

What's the science?

The immune system is responsible for taking protective action to protect the body against perceived external threats. The insular cortex (IC) of the brain is essential for integrating sensory information with bodily states. Prior work has implicated interactions between the insular cortex and the immune system, but the underlying functional dynamics remain unknown. 

This week in Nature Neuroscience, Kayyal and colleagues use behavioral and experimental tools in mouse models to explore the role of IC in regulating immune responses.

How did they do it?

To study the interactions between the body’s immune response and sensory information, the authors utilize a conditioned immune response (CIR) task design. In this design, experimental mice are exposed to a conditioned stimulus (a specific scent) that is paired with an unconditioned stimulus that activates or inhibits the immune system. They learn the association between the scent and the unconditioned stimulus, and the researchers can study what is going on in the brain when this CIR is learned and tested. 

To study brain activity, the researchers utilize retrograde labeling techniques to identify neurons within IC that project from anterior to posterior regions (aIC-to-pIC), or posterior to anterior regions (pIC-to-aIC). They tagged these populations to quantify the degree of connectivity, and changes in connectivity across various task conditions. 

To study changes in the synaptic properties of the neurons, they conducted electrophysiological recordings from tagged neurons to quantify changes in excitability and inhibitory currents. 

To test whether aIC-to-pIC neurons are necessary and sufficient for aspects of the measured CIR, they injected modified receptors into projecting neurons to allow them to experimentally activate or inhibit specific populations of neurons and measure changes in immune and behavioral responses.    

What did they find?

When studying the immune response and behavior after CIR, they found the experimental mice had a strong aversion to the conditioned stimulus and an elevated immune response. This suggests that CIR primes the immune system for potential infection after exposure to harmful bacteria. 

When comparing the neurons that project from aIC-to-pIC and vice versa, they found an increased percentage activation of aIC-to-pIC neurons following CIR, suggesting that these insular projections play a critical role in retrieving memory associated with the conditioned scent stimulus. When quantifying the excitatory to inhibitory ratio of projection neurons, they found reduced excitability and an increased number of active aIC-to-pIC projecting neurons, highlighting a potential mechanism by which information is flexibly retrieved during the retrieval of CIR-related memories.  

Through selectively activating and inhibiting the aIC-to-pIC and pIC-to-aIC projections, they found that inhibiting the aIC-to-pIC pathway reduced learned aversive behavior. Their findings suggest that aIC regions encode taste and its conditioned response (immune system threat) and that the aIC-to-pIC pathway is necessary for the successful retrieval of the CIR. Their results highlight the importance of the aIC-to-pIC projecting neurons in modulating the body’s immune response following exposure to a conditioned aversive sensory stimulus. 

What's the impact?

This study is the first to suggest a novel role for specific regions of the insular cortex in retrieving immune-related information and flexibly tuning behavioral responses. It was previously unknown how the functional connectivity of the insular cortex relates to the body’s immune response and how the brain and body interact to promote immune function.