Characterization of collagen image biomarkers in tissue microenvironment via deep learning approaches

Machine learning

Instrumentation

Biology/Pathology

Weak label and multi-instance learning

Uncertainty and confidence

Discover new image biomarkers in diseases

Interpretability

Representation learning

Explanation models

Computational imaging for pathology

Characterize the relationship between ECM and cells.

Correlation between image biomarkers and non-image vectors

3D pathology, extend the view of ECM, CurveAlign in 3D histopathological images

Cross-modality image synthesis for imaging collagen in histopathological slides

Single image super-resolution for whole slide imaging

Distinguishing pancreatitis and pancreatic cancer using deep weakly supervised learning

Learn with weak labels

Distributed representations

Representation learning and transfer learning

Characterizing Syndecan-1 regulated collagen image biomarkers of breast cancer in histopathological images

Explanation models

Use labels from genetic analysis

Watch dog for deep classifiers

Separate features learned for cells and ECMs

Deep multi-instance learning, weak pathological labels

Extend local classifiers to bigger scales for whole slide images

Extendable open-source software for hardware controlling

Polarization microscopy

Smart microscopy

Visualization of collagen fibers using computational optics

Run-time analysis

Hardware acceleration

Image-to image translation networks

Generative adversarial networks

Curriculum learning

Self-supervised color normalization

Alignment analysis of collagen fibers in volumetric images