Search Results
Evolutionary design of explainable algorithms for biomedical image segmentation
An unresolved issue in contemporary biomedicine is the overwhelming number and diversity of complex images that require annotation, analysis and interpretation. Recent advances in Deep Learning have revolutionized the field of computer vision, creating algorithms that compete with human experts in image segmentation tasks. However, these frameworks require large human-annotated datasets for training and the resulting “black box” models are difficult to interpret. In this study, we introduce Kartezio, a modular Cartesian Genetic Programming-based computational strategy that generates fully transparent and easily interpretable image processing pipelines by iteratively assembling and parameterizing computer vision functions. The pipelines thus generated exhibit comparable precision to state-of-the-art Deep Learning approaches on instance segmentation tasks, while requiring drastically smaller training datasets. This Few-Shot Learning method confers tremendous flexibility, speed, and functionality to this approach. We then deploy Kartezio to solve a series of semantic and instance segmentation problems, and demonstrate its utility across diverse images ranging from multiplexed tissue histopathology images to high resolution microscopy images. While the flexibility, robustness and practical utility of Kartezio make this fully explicable evolutionary designer a potential game-changer in the field of biomedical image processing, Kartezio remains complementary and potentially auxiliary to mainstream Deep Learning approaches.
Better insurance could effectively mitigate the increase in economic growth losses from U.S. hurricanes under global warming
Global warming is likely to increase the proportion of intense hurricanes in the North Atlantic. Here, we analyze how this may affect economic growth. To this end, we introduce an event-based macroeconomic growth model that temporally resolves how growth depends on the heterogeneity of hurricane shocks. For the United States, we find that economic growth losses scale superlinearly with shock heterogeneity. We explain this by a disproportional increase of indirect losses with the magnitude of direct damage, which can lead to an incomplete recovery of the economy between consecutive intense landfall events. On the basis of two different methods to estimate the future frequency increase of intense hurricanes, we project annual growth losses to increase between 10 and 146% in a 2°C world compared to the period 1980–2014. Our modeling suggests that higher insurance coverage can compensate for this climate change–induced increase in growth losses.
Breast Cancer Stem Cell–Derived Tumors Escape from γδ T-cell Immunosurveillance In Vivo by Modulating γδ T-cell Ligands
There are no targeted therapies for patients with triple-negative breast cancer (TNBC). TNBC is enriched in breast cancer stem cells (BCSC), which play a key role in metastasis, chemoresistance, relapse, and mortality. γδ T cells hold great potential in immunotherapy against cancer and might provide an approach to therapeutically target TNBC. γδ T cells are commonly observed to infiltrate solid tumors and have an extensive repertoire of tumor-sensing mechanisms, recognizing stress-induced molecules and phosphoantigens (pAgs) on transformed cells. Herein, we show that patient-derived triple-negative BCSCs are efficiently recognized and killed by ex vivo expanded γδ T cells from healthy donors. Orthotopically xenografted BCSCs, however, were refractory to γ δ T-cell immunotherapy. We unraveled concerted differentiation and immune escape mechanisms: xenografted BCSCs lost stemness, expression of γ δ T-cell ligands, adhesion molecules, and pAgs, thereby evading immune recognition by γ δ T cells. Indeed, neither promigratory engineered γ δ T cells, nor anti–PD-1 checkpoint blockade, significantly prolonged overall survival of tumor-bearing mice. BCSC immune escape was independent of the immune pressure exerted by the γ δ T cells and could be pharmacologically reverted by zoledronate or IFNα treatment. These results pave the way for novel combinatorial immunotherapies for TNBC.
Intestinal epithelial barrier integrity investigated by label-free techniques in ulcerative colitis patients
The intestinal epithelial barrier, among other compartments such as the mucosal immune system, contributes to the maintenance of intestinal homeostasis. Therefore, any disturbance within the epithelial layer could lead to intestinal permeability and promote mucosal inflammation. Considering that disintegration of the intestinal epithelial barrier is a key element in the etiology of ulcerative colitis, further assessment of barrier integrity could contribute to a better understanding of the role of epithelial barrier defects in ulcerative colitis (UC), one major form of chronic inflammatory bowel disease. Herein, we employ fast, non-destructive, and label-free non-linear methods, namely coherent anti-Stokes Raman scattering (CARS), second harmonic generation (SHG), two-photon excited fluorescence (TPEF), and two-photon fluorescence lifetime imaging (2P-FLIM), to assess the morpho-chemical contributions leading to the dysfunction of the epithelial barrier. For the first time, the formation of epithelial barrier gaps was directly visualized, without sophisticated data analysis procedures, by the 3D analysis of the colonic mucosa from severely inflamed UC patients. The results were compared with histopathological and immunofluorescence images and validated using transmission electron microscopy (TEM) to indicate structural alterations of the apical junction complex as the underlying cause for the formation of the epithelial barrier gaps. Our findings suggest the potential advantage of non-linear multimodal imaging is to give precise, detailed, and direct visualization of the epithelial barrier in the gastrointestinal tract, which can be combined with a fiber probe for future endomicroscopy measurements during real-time in vivo imaging.