Current Research Areas

Our lab develops new biomedical imaging instruments and image analysis tools to help fight disease. These projects include developing measures that can predict the progression of cancer, building instruments that can image living creatures more gently than traditional methods, using machine learning to classify medical images, and developing open-source software for use by the image analysis community around the world.

Image Analysis Software: ImageJ and Fiji

ImageJ ( is an open-source image processing program designed for analysis of scientific multidimensional images. It is highly extensive, with thousands of plugins and macros for performing a wide variety of tasks for a strong, established user base. Fiji Is Just ImageJ, with extras. It is a distribution of ImageJ with many plugins useful for scientific image analysis in fields such as life sciences. It is actively maintained, with updates released often.

Our development goals for ImageJ:

  • Design: Create the next generation of ImageJ driven by the needs of the community.
  • Collaborate: Work together across organizations, fostering open development through sharing and reuse.
  • Broaden: Make ImageJ useful and relevant across disciplines within the scientific community.
  • Maintain: Preserve backwards compatibility with existing ImageJ functionality.
  • Unify: Provide a central online resource for the ImageJ community.
  • Lead: Drive ImageJ development forward with a clear vision.

Machine Learning for PDAC Screening

Pancreatic ductal adenocarcinoma (PDAC) is projected to be the second leading cause of cancer death by 2030 due to the often advanced degree of disease progression at diagnosis and lack of effective treatment approaches thereafter. PDAC grade (G1-3) and stage (TMN1-4) are negatively correlated withpatient survival, and both factors are the major indicators for designing treatment plans. We are studying the effectiveness of machine learning (Convolutional Neural Networks, CNNs) to classify normal pancreatic tissue and different gradations of PDAC using a large dataset consisting of stained tissue microarrays (TMA). Early results show promise in CNN’s ability to derive prognostic information from a holistic assessment of the tumor tissue. Through the machine learning techniques we currently have in development, it may soon be possible to achieve 99% accuracy in PDAC tumor grading, increasing treatment efficacy for care providers and improving patient outcomes.

Glial Cell Imaging with FLIM

Microglia are the resident macrophages of the human brain and comprise up to 5-15% of all cells in the brain. Microglia play a crucial role in the neuroinflammatory response to neurodegenerative diseases such as Alzheimer’s disease, Parkinson’s disease, Multiple Sclerosis. To properly understand the role of microglia in these illnesses, it is important to be able to visualize these cells without external labeling and to be able to differentiate between resting and activated cell states. Fluorescence Lifetime Imaging Microscopy (FLIM) of endogenous metabolic coenzyme NADH/FAD allows for both monitoring of microglia without labeling and for functional states to be differentiated. With FLIM, LOCI has found that microglia undergo a lifetime shift in NADH/FAD levels when compared with other glia cells and brain tissue, indicating a shift in metabolism associated with cell activation (figure). LOCI will continue to focus on computational advancement in analyzing lifetime data with the ultimate goal of improving our basic understanding of microglia cell activity, opening new avenues for effective treatment of neurodegenerative disease.

Tumor Associated Collagen Signatures (TACS)

Changes in collagen organization have been linked to prognosis in breast, ovarian, and pancreatic cancers. Over the past decade, LOCI and collaborators have developed Tumor Associated Collagen Signatures (TACS), a model to characterize tumor progression. Initially designed for breast cancer, TACS was later adapted to other cancer types including pancreatic cancer. TACS falls into three categories based on the alignment of collagen fibers with respect to each other or with respect to associated tumor boundaries. Specifically, non-invasive regions (TACS-1 and TACS-2 regions) are contained by collagen fibers oriented parallel to the tumor boundary while regions of local invasion (TACS-3 region) possess areas where collagen has been realigned perpendicular to the tumor boundary to facilitate local invasion . TACS-3 presence is correlated with poor survival and may be used as a predictor of imminent invasion and metastasis. LOCI has been putting many efforts into collagen quantification and has developed specialized tools including CurveAlign and CT-FIRE for this purpose, that can measure not only the individual collagen fiber geometries like thickness, length, angle, and curvature but also bulk properties like global or local fiber density and alignment. Figure1 demonstrates CurveAlign software to find TACS-3 regions.