Indra John Heckenbach

• Presentation
• Research outputs
• Most downloaded

My PhD was focused on applying machine learning to understand the mechanisms of aging.  I am using deep neural networks to analyze histology images of human tissue. With image data for skin, liver, and brain provided by Patobank, I’m training convolutional neural networks to recognize morphological changes that occur during aging in a feature-neutral manner.  One goal is to have neural networks discover the most significant characteristics of aging. As a secondary approach, I’m targeting particular features, such as nuclei and tissue types. Neural networks are trained to recognize regions of interest, and then other networks are used to associate patterns from those regions with chronological age or other properties.  These methods can provide insight into how features of cells and tissue change with age and disease.

In addition to my primary project, I’m applying similar techniques to several other studies.  I’ve used segmentation to recognize DAPI-stained nuclei and beta galactosidase from high-content microscopy, and then trained a neural network to detect cellular senescence from nuclear morphology alone.  I’ve also assisted other projects to quantify inflammation and senescence using immunofluorescent markers. After training a neural network to recognize Kupffer cells (liver macrophages), colocalization revealed an increase in inflammatory factors specific to this cell type.  I also quantified how adipose tissue recruits macrophages and their size expands with age. These effects contribute to an age-related decline in NAD and an increase in cellular senescence.