Years ago, our team at MIT created sensors and machine learning algorithms to detect changes in human emotion.As we shrunk the sensors, made them wearable, and started to collect data24/7, wediscovered several surprises. For example, 100 years of psychophysiology research had described the autonomic "sweat" response occurring with emotional arousal as a general phenomenon across the body-- activated with things such as hard cognitive effort or emotional stress (e.g., making your palms sweaty). To our surprise, we found this signal showing huge peaks on the wrist during non-REM sleep, and especially large and lateralized peaks on a single wrist during some kinds of seizures and some kinds of stress. In fact, it could even peak significantly when the EEG showed no cortical brain activity. This talk will highlight some of the most surprising findings along the journey of measuring emotion “in the wild" with implications for anxiety, depression, sleep-memory consolidation, epilepsy, autism, and more.Can these findings help us build non-invasive wearables to accurately forecast important changes in our physical and mental health? 

Hemorrhagic myocardial infarction (MI) has been reported in 41% and 54% of ST-elevated MI patients after primary percutaneous coronary intervention. These patients are at high risk for adverse left ventricle (LV) remodeling, impaired LV function and increased risk of fatal arrhythmias.  Relaxation time MRI such as T2*-maps are sensitive to hemorrhagic infarct iron content, but are also affected by myocardial edema and fibrosis. Quantitative Susceptibility Mapping (QSM), which uses the MR signal phase to quantify tissue magnetic susceptibility, may be a more specific and sensitive marker of hemorrhagic MI. The objective of this study was to develop and validate cardiac QSM in a large animal model of myocardial infarction, investigate the association of magnetic susceptibility with iron content and infarct pathophysiology, and compare QSM to relaxation time mapping, susceptibility-weighted imaging (SWI), and late-gadolinium enhanced (LGE) MRI. 

 Visceral myopathy is a debilitating chronic medical condition in which smooth muscle of the bowel, bladder, and uterus is weak or dysfunctional. When the bowel muscle is weak and unable to efficiently contract, the bowel becomes distended which causes pain, bilious vomiting, growth failure, and nutritional deficiencies. The abdominal distension can become life-threatening. Patients often become dependent on intravenous nutrition and undergo multiple rounds of abdominal surgery, which only partially alleviates symptoms. Recently, rare mutations in gamma smooth muscle actin (ACTG2) have been shown to be responsible for a large subset of visceral myopathies. ACTG2 is a critical protein in the smooth muscle contractile apparatus. However, we have only limited knowledge of how ACTG2 mutations may cause human disease. To improve our understanding of the pathophysiology of ACTG2 mutations, my work has the following specific aims:

1) Determine how pathogenic ACTG2 mutations affect actin structure and function in primary human intestinal smooth muscle cells (HISMCs).
2) Examine the effects of ACTG2 mutations on differentiation of human pluripotent stem cells into smooth muscle cells (SMCs).

These studies will allow us to elucidate the mechanisms through which ACTG2 mutations impair normal visceral smooth muscle development and function. I am examining the effects of ACTG2 mutations on actin filament organization in fixed cells and actin dynamics in live cells using fluorescent probes. I will investigate changes in contractile force generation using traction force microscopy. I am also developing a novel differentiation method to convert pluripotent stem cells into cells closely resembling visceral SMCs. We will use this method with CRISPR/Cas9 gene editing to study the effects of the mutations on SMC differentiation. We are currently generating pluripotent stem cell lines containing ACTG2 mutations and performing assays to investigate actin organization and dynamics. We are using these systems to identify robust phenotypes highlighting the mechanisms through which ACTG2 mutations cause disease. We hope to leverage this information in the selection of targets for high-throughput drug screening, which may eventually lead to novel treatment strategies for visceral myopathies. 

Two main sources of pathology have been identified in the behavioral variant of frontotemporal dementia (bvFTD): tau inclusions (FTLD-tau) and TDP-43 aggregates (FTLD-TDP). With therapies emerging that target these proteins, exploring distinct trajectories of degeneration can be extremely helpful for tracking progression in clinical trials and improve prognosis estimation. We hypothesized that longitudinal cortical thinning (CT) would identify areas of extant disease progression in bvFTD subgroups and longitudinal hypoperfusion would identify distinct regions of anticipated neurodegeneration. We included N=47 patients with probable or definite bvFTD and two MRI scanning sessions including T1-weighted and arterial spin labeling (ASL) scans, recruited through the Penn Frontotemporal Degeneration Center. Neuropathology, genetic mutations, or CSF protein markers (phosphorylated tau (p-tau)/Ab1-42<.09 for likely FTLD; p-tau<8.75 for FTLD-TDP) were used to identify bvFTD with likely FTLD-tau (n=28, mean age=63.1 years, mean disease duration=3.89 years) or likely FTLD-TDP (n=19, mean age=61.9, mean disease duration=3.06). Voxel-wise cortical thickness and cerebral blood flow estimates were generated for each T1 and ASL scan, respectively, using longitudinal pipelines in ANTs. We created annual change images by subtracting follow-up images from baseline and dividing by inter-scan interval. In whole brain voxel-wise comparisons, FTLD-tau showed significantly greater right orbitofrontal CT and longitudinal hypoperfusion in right middle temporal and angular cortex relative to FTLD-TDP. FTLD-TDP displayed greater progressive CT in left superior and middle frontal cortex, precentral gyrus, and right temporal cortex, and longitudinal hypoperfusion in medial prefrontal cortex relative to FTLD-tau. In conclusion, FTLD-tau and FTLD-TDP show distinct patterns of longitudinal CT and hypoperfusion. Structural and functional MRI contribute independent information potentially useful for characterizing disease progression in vivo for clinical trials.

Star pattern test objects are used to evaluate the high-contrast performance of imaging systems. These objects were used to investigate alternative scanning geometries for a prototype next-generation tomosynthesis (NGT) system. The NGT system has 2D planar source motion and linear detector motion, and is capable of myriad acquisition geometries. We designed a virtual star pattern with a voxel size of 5𝜇m, and used it to evaluate the spatial resolution performance of the NGT system for three different acquisition geometries. The Open Virtual Clinical Trials (OpenVCT) framework was used to simulate virtual star patterns for acquisition geometries of the NGT system. Simulated x-ray projections of the virtual phantom were used to create super-sampled 3D image reconstructions. Using the same acquisition geometries on the NGT system, a physical star pattern was imaged to create experimental 3D image reconstructions. The simulated and experimental data were compared qualitatively by visual inspection, and quantitatively using an in-house metric. This metric computes the Fourier transform radially for one quadrant of the star pattern to discern the limit of spatial resolution (LSR) and the existence of aliasing. The results exhibit the same characteristics in terms of super-resolution and moiré patterns (arising from aliasing) with visual inspection. The LSR for the 12 conditions analyzed are all within 3% for the experimental and simulated data. Aliasing was determined to be present in the same simulated image reconstructions as the experimental complements. Super-resolution is observed for two of the three NGT acquisition geometries in the experimental and simulated images.

Ischemic mitral regurgitation (IMR) is primarily a left ventricular disease in which the mitral valve is dysfunctional due to ventricular remodeling after myocardial infarction. Current automated methods have focused on analyzing the mitral valve and left ventricle independently. While these methods have allowed for valuable insights into mechanisms of IMR, they do not fully integrate pathological features of the left ventricle and mitral valve. Thus, there is an unmet need to develop an automated segmentation algorithm for the left ventricular mitral valve complex, in order to allow for a more comprehensive study of this disease. The objective of this study is to generate and evaluate segmentations of the left ventricular mitral valve complex in pre-operative 3D transesophageal echocardiography using multi-atlas label fusion. These patient-specific segmentations could enable future statistical shape analysis for clinical outcome prediction and surgical risk stratification. In this study, we demonstrate a preliminary segmentation pipeline that achieves an average Dice coefficient of 0.78 ± 0.06.

Within the neural engineering field, next-generation implantable neuroelectronic interfaces are being developed using biologically-inspired and/or biologically-derived materials to improve upon the stability and functional lifetime of current interfaces. These technologies are based around biomaterials, bioactive molecules, living cells, or some combination of these to promote host neuronal survival, reduce the foreign body response, and improve chronic device-tissue integration. In lab, we are applying a tissue- engineering/biomaterials approach to generate implantable micro-tissue featuring long- projecting axonal tracts encased within carrier hydrogel micro-columns. These so-called "living electrodes" have been engineered with carefully tailored material, mechanical, and biological properties designed to enable natural, synaptic-based modulation of specific host circuitry and/or monitoring of host cortical activity. To date, we have developed a fabrication method for reproducible fabrication, characterized these constructs’ growth, survival, and architecture, and demonstrated successful implantation, survival, and network-level activity of the constructs in a rat model. We have also shown that these constructs may be engineered for optical control (via optogenetics) and optical output (through calcium imaging).

Scientific claims are a foundation of scientific discourse. Extracting claims from scientific articles is primarily done by researchers during literature review and discussions. The enormous growth in scientific articles makes this ever more challenging and time-consuming. Here, we develop a deep neural network architecture to solve the problem. Our model an F1 score of 0.704 on a large corpus of expertly annotated claims within abstracts. Our results suggest that we can use a small dataset of annotated resources to achieve high-accuracy claim detection. We release a tool for discourse and claim detection, and a novel dataset annotated by experts. We discuss further applications beyond Biomedical literature.

 

 

A number of challenging questions routinely arise before heart valve surgery. Should a regurgitant valve be repaired or replaced? If repaired, which techniques and maneuvers should be employed? When is the optimal time to perform surgery? This is just the beginning of a series of questions whose answers rely on an individualized assessment of heart valve morphology and function. With advanced imaging technology like 3D echocardiography already in the operating room, we have a window into patient-specific disease characteristics that cannot otherwise be appreciated immediately before surgery. Unfortunately, imaging resources are often not tapped to their full potential, which contributes to delays in surgical innovation. This presentation introduces an image-based modeling approach to creating enhanced visualizations and quantification of heart valves from real-time 3D echocardiography. The goal of this approach is to fundamentally change the way that surgeons interact with and utilize images in the operating room. Pre-operative image-based heart valve models can be used to characterize mechanisms of disease prior to cardiopulmonary bypass and to identify connections between pre-operative image features and clinical outcomes. Image-based modeling provides a means for surgeons to innovate and master new surgical techniques like bicuspid aortic valve repair and to devise new standardized approaches for surgical treatment of ischemic mitral regurgitation. Image analysis and surgical innovation are linked within and beyond the cardiovascular domain, and this work aims to optimize the potential of imaging and shape modeling for pre-surgical planning.