Where, why, and how is bias learned in medical image analysis models? A study of bias encoding within convolutional networks using synthetic data.

Background: Understanding the mechanisms of algorithmic bias is highly challenging due to the complexity and uncertainty of how various unknown sources of bias impact deep learning models trained with medical images. This study aims to bridge this knowledge gap by studying where, why, and how biases from medical images are encoded in these models.

Methods: We systematically studied layer-wise bias encoding in a convolutional neural network for disease classification using synthetic brain magnetic resonance imaging data with known disease and bias effects.

HarmonyTM: multi-center data harmonization applied to distributed learning for Parkinson's disease classification.

Purpose: Distributed learning is widely used to comply with data-sharing regulations and access diverse datasets for training machine learning (ML) models. The traveling model (TM) is a distributed learning approach that sequentially trains with data from one center at a time, which is especially advantageous when dealing with limited local datasets. However, a critical concern emerges when centers utilize different scanners for data acquisition, which could potentially lead models to exploit these differences as shortcuts.

Foundation model-driven distributed learning for enhanced retinal age prediction.

Objectives: The retinal age gap (RAG) is emerging as a potential biomarker for various diseases of the human body, yet its utility depends on machine learning models capable of accurately predicting biological retinal age from fundus images. However, training generalizable models is hindered by potential shortages of diverse training data. To overcome these obstacles, this work develops a novel and computationally efficient distributed learning framework for retinal age prediction.

Towards objective and systematic evaluation of bias in artificial intelligence for medical imaging.

Objective: Artificial intelligence (AI) models trained using medical images for clinical tasks often exhibit bias in the form of subgroup performance disparities. However, since not all sources of bias in real-world medical imaging data are easily identifiable, it is challenging to comprehensively assess their impacts. In this article, we introduce an analysis framework for systematically and objectively investigating the impact of biases in medical images on AI models.

Factors associated with older adults' cognitive decline 6 months after gamma-variant SARS-CoV-2 infection.

Background: Cognitive deficits are commonly reported after COVID-19 recovery, but little is known in the older population. This study aims to investigate possible cognitive damage in older adults 6 months after contracting COVID-19, as well as individual risk factors.

Methods: This cross-sectional study involved 70 participants aged 60-78 with COVID-19 6 months prior and 153 healthy controls.

A multi-center distributed learning approach for Parkinson's disease classification using the traveling model paradigm.

Distributed learning is a promising alternative to central learning for machine learning (ML) model training, overcoming data-sharing problems in healthcare. Previous studies exploring federated learning (FL) or the traveling model (TM) setup for medical image-based disease classification often relied on large databases with a limited number of centers or simulated artificial centers, raising doubts about real-world applicability. This study develops and evaluates a convolution neural network (CNN) for Parkinson's disease classification using data acquired by 83 diverse real centers around the world, mostly contributing small training samples.

The context of COVID-19 affected the long-term sleep quality of older adults more than SARS-CoV-2 infection.

Introduction: Sleep problems are one of the most persistent symptoms of post-COVID syndrome in adults. However, most recent research on sleep quality has relied on the impact of the pandemic, with scarcely any data for older adults on the long-term consequences of COVID infection. This study aims to understand whether older individuals present persistently impaired sleep quality after COVID-19 infection and possible moderators for this outcome.

What you have, not who you know: food-enhanced social capital and changes in social behavioural relationships in a non-human primate.

Social network position in non-human primates has far-reaching fitness consequences. Critically, social networks are both heterogeneous and dynamic, meaning an individual's current network position is likely to change due to both intrinsic and extrinsic factors. However, our understanding of the drivers of changes in social network position is largely confined to opportunistic studies.

Identifying Biases in a Multicenter MRI Database for Parkinson's Disease Classification: Is the Disease Classifier a Secret Site Classifier?

Sharing multicenter imaging datasets can be advantageous to increase data diversity and size but may lead to spurious correlations between site-related biological and non-biological image features and target labels, which machine learning (ML) models may exploit as shortcuts. To date, studies analyzing how and if deep learning models may use such effects as a shortcut are scarce. Thus, the aim of this work was to investigate if site-related effects are encoded in the feature space of an established deep learning model designed for Parkinson's disease (PD) classification based on T1-weighted MRI datasets.

Simulation-based training in Leprosy: development and validation of a scenario for community health workers.

Objectives: To build and validate a clinical simulation scenario designed to instruct community health workers (CHWs) in active leprosy case detection.

Methods: Methodological study involving the development of a simulated clinical scenario and content validation by experts. The Content Validity Index (CVI) was used to determine the level of agreement among the judging commitee, and a descriptive analysis of their recommendations was performed.

Image-encoded biological and non-biological variables may be used as shortcuts in deep learning models trained on multisite neuroimaging data.

Objective: This work investigates if deep learning (DL) models can classify originating site locations directly from magnetic resonance imaging (MRI) scans with and without correction for intensity differences.

Material And Methods: A large database of 1880 T1-weighted MRI scans collected across 41 sites originally for Parkinson's disease (PD) classification was used to classify sites in this study. Forty-six percent of the datasets are from PD patients, while 54% are from healthy participants.

Explainable classification of Parkinson's disease using deep learning trained on a large multi-center database of T1-weighted MRI datasets.

Introduction: Parkinson's disease (PD) is a severe neurodegenerative disease that affects millions of people. Early diagnosis is important to facilitate prompt interventions to slow down disease progression. However, accurate PD diagnosis can be challenging, especially in the early disease stages.

The influence of six polymorphisms of uncoupling protein 3 (UCP3) gene and childhood obesity: a case-control study.

Background: Obesity is defined as a multifactorial disease, marked by excessive accumulation of body fat, responsible for compromising the individual's health over the years. The energy balance is essential for the proper functioning of the body, as the individual needs to earn and spend energy in a compensatory way. Mitochondrial Uncoupling Proteins (UCP) help in energy expenditure through heat release and genetic polymorphisms could be responsible for reducing energy consumption to release heat and consequently generate an excessive accumulation of fat in the body.

An analysis of the effects of limited training data in distributed learning scenarios for brain age prediction.

Objective: Distributed learning avoids problems associated with central data collection by training models locally at each site. This can be achieved by federated learning (FL) aggregating multiple models that were trained in parallel or training a single model visiting sites sequentially, the traveling model (TM). While both approaches have been applied to medical imaging tasks, their performance in limited local data scenarios remains unknown.

Multi-scale phase separation by explosive percolation with single-chromatin loop resolution.

The 2 m-long human DNA is tightly intertwined into the cell nucleus of the size of 10 μm. The DNA packing is explained by folding of chromatin fiber. This folding leads to the formation of such hierarchical structures as: chromosomal territories, compartments; densely-packed genomic regions known as Topologically Associating Domains (TADs), or Chromatin Contact Domains (CCDs), and loops.

Sandpile cascades on oscillator networks: The BTW model meets Kuramoto.

Cascading failures abound in complex systems and the Bak-Tang-Weisenfeld (BTW) sandpile model provides a theoretical underpinning for their analysis. Yet, it does not account for the possibility of nodes having oscillatory dynamics, such as in power grids and brain networks. Here, we consider a network of Kuramoto oscillators upon which the BTW model is unfolding, enabling us to study how the feedback between the oscillatory and cascading dynamics can lead to new emergent behaviors.

[Individual and contextual factors associated with smoking among young Brazilian adults].

Young people receive special attention regarding smoking as it is a period of life in which the use of this and other substances generally starts and is consolidated. There are no studies on risk factors associated with young adults with a representative sample from Brazil that take into consideration individual and contextual aspects. The objective was to identify factors associated with smoking among young Brazilian adults aged 18 to 24 years, considering the combined influence of individual and contextual factors assessed through the Municipal Human Development Index (MHDI).

Effectiveness of Pedagogical Tools for Teaching Medical Gross Anatomy During the COVID-19 Pandemic.

Unlabelled: Medical institutions have been forced to modify gross anatomy pedagogy to comply with the health restrictions imposed by the novel coronavirus (COVID-19). Boston University School of Medicine (BUSM) is one such institution that temporarily restructured its course. We replaced cadaveric dissection activities with prosections and placed a greater emphasis on a flipped classroom model.

Unusual bilateral variation of the flexor digitorum accessorius longus muscle and its relation on tarsal tunnel syndrome.

Purpose: During standard anatomical dissection for a medical anatomy course, we encountered an unusual bilateral variant of a unipennate flexor digitorum accessorius longus (FDAL) muscle, a supernumery muscle of the deep posterior leg and medial ankle.

Methods: We documented the muscles course and measured the diameter and length of the FDAL muscle belly, as well as the full length of its tendinous attachments.

Results: On both right and left legs, the FDAL originated from the proximal posterior fibula and distal one-third of the flexor hallucis longus muscle.

Mutualistic networks emerging from adaptive niche-based interactions.

Mutualistic networks are vital ecological and social systems shaped by adaptation and evolution. They involve bipartite cooperation via the exchange of goods or services between actors of different types. Empirical observations of mutualistic networks across genres and geographic conditions reveal correlated nested and modular patterns.

Homophily Based on Few Attributes Can Impede Structural Balance.

Homophily between agents and structural balance in connected triads of agents are complementary mechanisms thought to shape social groups leading to, for instance, consensus or polarization. To capture both processes in a unified manner, we propose a model of pair and triadic interactions. We consider N fully connected agents, where each agent has G underlying attributes, and the similarity between agents in attribute space (i.

Prevention of infections associated with peripheral catheters: construction and validation of clinical scenario.

Objectives: to describe the process of construction and content validation of a clinical setting for teaching and learning the recommended practices for preventing bloodstream infection, associated with peripheral venous catheters, for nursing professionals.

Methods: methodological study of the construction of the scenario based on the National League Nursing Jeffries Simulation Theory, International Nursing Association for Clinical Simulation and Learning, and the Brazilian Health Regulatory Agency. Twelve experts performed content validation.