Software developed by CSB and referenced in papers.
1. HNSCC spatial analysis
This repository contains the code used to generate the data shown in the manuscript "Spatial analysis identifies PD-L1 expressing DC niches are predictors of immunotherapy outcomes in head and neck squamous cell cancer".
The pipeline is split in four parts:
1. Illumination correction: Correcting raw images for uneven illumination
2. Alignment, quench subtraction, segmentation and data extraction: Extracting cell level quantitative information for further analysis
3. Cell phenotyping: Assigning cellular phenotypes and perform subtyping
4. Spatial processing: Extract spatial metrics
Download Source Code from GitHub
2. Cell Data Loader
Cell Data Loader is a simple AI support tool in Python that can take in images of cells (or other image types) and output them with minimal effort to formats that can be read by Pytorch (Tensor) or Tensorflow (Numpy) format. With Cell Data Loader, users have the option to output their cell images as whole images, sliced images, or, with the support of CellPose, segment their images by cell and output those individually.
It can also be used for normal computer vision research, which is why CellPose is not a strict dependency.
Download Source Code from GitHub
3. Multi-Input Medical Image Machine Learning Toolkit
The Multi-Input Medical Image Machine Learning Toolkit (MultiMedImageML) is a library of Pytorch functions that can encode multiple 3D images (designed specifically for brain images) and offer a single- or multi-label output, such as a disease detection.
Thus, with a dataset of brain images and labels, you can train a model to predict dementia or multiple sclerosis from multiple input brain images.
Download Source Code from GitHub
4. Software for Automated Cell Profiling from Cyclic Imaging
Paper: "Deep Learning Pipeline for Automated Cell Profiling from Cyclic Imaging”
This package pre-processes raw image files from cyclic immunofluorescence microscopy, corrects for translation errors between imaging cycles, segments individual cells and generates single-cell molecular profiles.
5. Three-stage classification of Alzheimer’s disease in MRI
Paper: “Three-stage classification of Alzheimer’s disease in MRI in clinical records of the Mass General Brigham Health Care System.”
Download Source Code from GitHub
6. Deep Learning Systems For Lymphoma Diagnostics
Paper: “Design and clinical validation of a point-of-care device for the diagnosis of lymphoma via contrast-enhanced microholography and machine learning.” PMID: 30555750
This code is part of a project to leverage deep learning systems for lymphoma diagnostics. We use the Maximally Stable Extrema Regions (MSER) blob detection algorithm to find regions of interest in a hologram image of cells and use a trained Convolutional Neural Network (CNN) to filter out non-cell regions.
Region_Detection.py:
– Detects cells in a hologram using the MSER algorithm and a pre-trained CNN.
- Contains several helper functions for processing multiple images from a single folder, or multiple folders that each contain multiple images.
Region_Filter.py:
- Contains the code to construct a CNN using Keras.
- Can load a pre-trained CNN or build an empty architecture and train it.
- Contains helper function to properly format and classify an input region with a pre-trained CNN.
Data_Loader.py:
- Contains numerous functions to load labelled training images from a directory into (x,y) training tensors.
- Can pre-process data to make for easier training.
- Designed to return data in a format that the region filter CNN can be trained with.
7. Cell Segmentation for Breast Cancer Cells
Paper: "CytoPAN – Portable cellular analyses for rapid point-of-care cancer diagnosis.” PMID: 32759277
This Matlab code is part of a project to identify breast cancer cells in bright-field and measure the mean fluorescence intensity of respective cells in fluorescence imaging for breast cancer biomarkers (QUAD, HER2, ERPR).
CellAnalysis.m:
- Sobel edge detection method was used to determine cellular edges based on high contrast in the bright-field image.
- Binary gradient mask underwent dilation, hole filling, and smoothing
- Mask regions correlated with DAPI signals are identified as biological cells.
- Mean intensities of biomarkers (QUAD, HER2 and ERPR) of individual biological cells are measured.
8. R-Script and Data Files for EV Protein Search
Paper: "A high-throughput magneto-electrochemical array for the integrated isolation and profiling of extracellular vesicles from plasma”
The following files are used to search EV protein markers for colorectal cancer diagnosis.
Marker_selection.Rmd:
– R script to load and filter datasets from public databases, and to compare protein expression levels between normal and tumor tissues.
Marker_selection.nb.html:
– HTML output generated by Marker_selection.Rmd.
evpedia.csv
– List of EV proteins retrieved from EVpedia.
9. RNAseq data for CANDI wafer
To determine the effects of the CANDI wafer on phagocytic cells, bulk RNAseq was performed. Bone marrow-derived macrophages were isolated and differentiated as previously described. Cells were then stimulated for 24 h with drug loaded wafer material to induce activation. RNA was isolated using the RNeasy Plus Mini Kit (Qiagen). Final RNA concentration was determined by absorbance (Nanodrop), and samples were stored at -80 ºC until shipment for sequencing (NovoGene).
Download RNASeq data
CANDIwafer_RNAseq_CANDIwafer_vs_Empty_wafer.xlsx
CANDIwafer_RNAseq_Empty_wafer_vs_Control.xlsx
Python script to create synthetic images from IHC data
This repository provides a Python script to generate synthetic dot-based images of cell identities using single-cell spatial data and segmentation masks. The tool was developed for visualizing cell-type distributions in multiplexed tissue imaging data