(See also the 4 PDF files appended below) My objectives have been to understand human diseases better using computer-aided tools and, in doing so, advance the field of personalized medicine. My background is in computing and mathematics. I received undergraduate training in electronics and electrical engineering. I went on to complete postgraduate training in statistical signal processing (EPF Lausanne) and biophysics and optical microscopy (ETH Zürich). During my dissertation work at Scripps Research, I focused on the development of novel algorithms for automated analysis of the motion of cytoskeletal proteins. I apply methods based on scale-space, information and graph theory, clustering algorithms, multi-objective optimization, spatial and Bayesian filtering, template matching, neural networks, deep learning, expectation-maximization, Monte Carlo simulations, stochastic processes, hypothesis testing, texture and wavelet analysis.
At Cornell Medicine, I led a team of three computational scientists and developed algorithms for computer vision analyses of images of patient-derived CTCs. I quantified the effects of six chemotherapy drugs on four breast cancer cell lines, which demonstrated a correlation between tumor type, microtubule dynamics, and drug efficacy. I evaluated the ability of NAD+ to restore altered microtubule dynamics in a breast cancer cell line and its effects during axonal degeneration in terms of protecting microtubules in a mouse model. Consulting for industry, I contributed to studies on drug discovery for Alzheimer’s disease based on the analysis of changes in mitochondria morphology and motion in patient-derived iPS cells (iPerian Inc.) and Parkinson’s disease based on motion analysis of lysosomes in mouse astrocytes (Pfizer Inc.). At UCSF, I stained for histology, GFP-transduced live cells, imaged at a high spatial and temporal resolution, treated with cytotoxic drugs and ferroptosis-inducing small molecules, performed cell viability assays and WGS analysis of organoid cultures I derived from metastatic and primary prostate and metastatic colorectal tumor tissues resected from hospital patients. With my startup company DataSet Analysis, I developed software for real-time motion tracking of vesicle movement. I also worked on the establishment of a biobank for body fluids and longitudinal analysis of urinary small RNA sequencing data to detect lung cancer and other degenerative diseases. At Aarhus University, my focus was the classification of circulating cell-free DNA in blood samples from cancer patients participating in clinical trials in Denmark and the UK. I analyzed DNA fragmentation length distributions in healthy individuals without and with comorbidities, patients with low-grade and high-grade adenoma, colon, rectal, lung, breast, gastric, ovarian, pancreatic, duodenal, and bile duct tumors in the context of early disease detection as well as the detection of residual disease and recurrence after curative-intent surgery. At BioSpyder Technologies Inc., I contributed to the development and bioinformatics analysis of a novel single-cell transcriptomics assay as well as to the analysis of templated oligo-sequencing data to identify gene expression profiles for diagnosis of Alzheimer’s and Parkinson’s disease in self-collected finger-stick blood samples spotted on filter paper. At Amydis Inc., I developed computer vision tools and led the efforts to analyze patterns of aberrant protein aggregates in ex vivo images of the patient retina for the detection of Alzheimer's, Parkinson's, amyotrophic lateral sclerosis, frontotemporal dementia, transthyretin amyloidosis, and glaucoma. I designed and prototyped a computer vision algorithm to facilitate medical diagnosis based on bag-of-visual-words and bag-of-bags-of-words image classification and retrieval in a surgically-induced non-human primate in vivo disease model. At Medentum Innovations Inc., I analyzed and classified data for the detection of pediatric ear (acute otitis media, chronic suppurative otitis media, myringosclerosis, otitis externa, otitis media with effusion, earwax plug, etc.) and throat (exudative pharyngitis, non-exudative pharyngitis, and aphthous ulcer) disease in images acquired with an otoscope as well as adventitious lung sounds (wheeze, rhonchi, stridor, and crackle) for respiratory and cardiovascular disease detection acquired with a stethoscope for AI-based telehealth diagnostics.
My objective has been to develop resources for functional interrogation of drug response in a physiologically relevant system amenable to molecular manipulations and investigate personalized drug response ex vivo. I have developed image analysis software for automated motion tracking of labeled microtubules and actin - ClusterTrack (for measurements of interphase cells), Instantaneous Flow Tracker (for measurements of interdigitated flows in dividing cells, contractile actin networks in migrating epithelial cells and growth cones), and Dataset Tracker (for real-time optical flow feature tracking), which can serve as the base module of an integrated platform of all existing and future algorithms for real-time cellular analysis. The computational assay I propose could successfully be applied to evaluate treatment strategies for any human organ.