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. Therefore, all layers of the PD classifier were frozen, except for the last layer of the network, which was replaced by a linear layer that was exclusively re-trained to predict three potential bias types (biological sex, scanner type, and originating site). Our findings based on a large database consisting of 1880 MRI scans collected across 41 centers show that the feature space of the established PD model (74% accuracy) can be used to classify sex (75% accuracy), scanner type (79% accuracy), and site location (71% accuracy) with high accuracies despite this information never being explicitly provided to the PD model during original training. Overall, the results of this study suggest that trained image-based classifiers may use unwanted shortcuts that are not meaningful for the actual clinical task at hand. This finding may explain why many image-based deep learning models do not perform well when applied to data from centers not contributing to the training set.

Image-Based Tools and Analysis for Human RVOT/RV Structures.

The right-ventricular (RV) outflow tract (RVOT) and the transition to the RV free wall are recognized sources of arrhythmia in human hearts. However, we do not fully understand myocardial tissue structures in this region. Human heart tissue was processed for optical clarity, labelled with wheat-germ agglutin (WGA) and anti-Cx43, and imaged on a custom-built line scanning confocal microscope. The 3D images were analyzed for myocyte gross structures and cell morphology. There were regions of high organization as well as rapid changes to more heterogeneous regions. Preliminary cell segmentations were used to estimate cell morphology. Observed RVOT/RV structure is consistent with known arrhythmic substrates.Clinical Relevance- New views of human tissue structure enable clearer clinical understanding of arrhythmogenic activation pathways and targets for invasive treatment such as RF ablation.

The causal link between cardiometabolic risk factors and gray matter atrophy: An exploratory study.

Although gray matter atrophy is commonly observed with aging, it is highly variable, even among healthy people of the same age. This raises the question of what other factors may contribute to gray matter atrophy. Previous studies have reported that risk factors for cardiometabolic diseases are associated with accelerated brain aging. However, these studies were primarily based on standard correlation analyses, which do not unveil a causal relationship. While randomized controlled trials are typically required to investigate true causality, in this work, we investigated an alternative method by exploring data-driven causal discovery and inference techniques on observational data. Accordingly, this feasibility study used clinical and quantified gray matter volume data from 22,793 subjects from the UK biobank cohort without any known neurological disease. Our method identified that age, sex, body mass index (BMI), body fat percentage (BFP), and smoking exhibit a causal relationship with gray matter volume. Interventions on the causal network revealed that higher BMI and BFP values significantly increased the chance of gray matter atrophy in males, whereas this was not the case in females.

Extended-volume image-derived models of coronary microcirculation.

OBJECTIVE: Recent advances in tissue clearing and high-throughput imaging have enabled the acquisition of extended-volume microvasculature images at a submicron resolution. The objective of this study was to extract information from this type of images by integrating a sequence of 3D image processing steps on Terabyte scale datasets. METHODS: We acquired coronary microvasculature images throughout an entire short-axis slice of a 3-month-old Wistar-Kyoto rat heart. This dataset covered 13 × 10 × 0.6 mm at a resolution of 0.933 × 0.933 × 1.866 µm and occupied 700 Gigabytes of disk space. We used chunk-based image segmentation, combined with an efficient graph generation technique, to quantify the microvasculature in the large-scale images. Specifically, we focused on the microvasculature with a vessel diameter up to 15 µm. RESULTS: Morphological data for the complete short-axis ring were extracted within 16 h using this pipeline. From the analyses, we identified that microvessel lengths in the rat coronary microvasculature varied from 6 to 300 µm. However, their distribution was heavily skewed toward shorter lengths, with a mode of 16.5 µm. In contrast, vessel diameters ranged from 3 to 15 µm and had an approximately normal distribution of 6.5 ± 2 µm. CONCLUSION: The tools and techniques from this study will serve other investigations into the microcirculation, and the wealth of data from this study will enable the analysis of biophysical mechanisms using computer models.

Reconstruction of coronary circulation networks: A review of methods.

Building anatomically accurate models of the coronary vascular system enables potentially deeper understandings of coronary circulation. To achieve this, (a) images at different levels of vascular network-arteries, arterioles, capillaries, venules, and veins-need to be obtained through suitable imaging modalities; and (b) from images, morphological and topological information needs to be extracted using image processing techniques. While there are several modalities that enable the imaging of large vessels, microcirculation imaging-capturing vessels having diameter lesser than 100 µm-has to date been typically confined to small regions of the heart. This spatially limited microcirculatory information has often been used within cardiac models, with the potentially erroneous assumption that it is representative of the whole organ. However, with the recent advancements in imaging and image processing, it is rapidly becoming feasible to acquire, process, and quantify microcirculation data at the scale of whole organ. In this review, we summarize the progress toward this goal followed through a presentation of the current state-of-the-art imaging and image processing techniques in the context of coronary microcirculation extraction, prominently but not exclusively, from small animals.