Skip to content

Installation

Overview

There are four components that can be installed. You may install all of them or just the ones you need, depending on your requirements. Those components will allow you to use/depoly an existing model that is published or build your own models with: MES Tools or with the Python API

Releases

A prebuilt package (excluding MES Tools) is available for direct download of the latest version, eliminating the need to compile from source. The binaries are built on Ubuntu 24.04, and will run on any Linux distribution with glibc ≥ 2.39. You will also need to install OpenMP

Preliminary Steps to Build the Tools

These are the preliminary installation steps required:

1. Install Compiler and Build Tools (Ubuntu)

To install the essential compiler and build tools, run:

1
sudo apt install binutils gcc g++ cmake make -y

2. Install OpenMP Support (Ubuntu)

To enable OpenMP (used for parallel processing), install the following package:

1
sudo apt install libgomp1 -y

NOTE: This is required step.

3. Install Boost Libraries (Ubuntu)

To install the required Boost components, on Ubuntu 24.04, use:

1
sudo apt install libboost-system1.83-dev libboost-filesystem1.83-dev libboost-regex1.83-dev libboost-program-options1.83-dev -y

Note: On Ubuntu 22.04, Boost version 1.74 is available and is also compatible.

If you want to compile the AlgoMarker library or compile it against a different Boost library, please download and compile Boost manually and follow the Boost Compilation Steps. This project has been tested with Boost versions 1.67 through 1.85 and should work with other versions as well.

Boost Compilation Steps

Example installation steps for version 1.85.0:

Boost Compilation
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
# Ensure you have wget to download the file and bzip2 to extract the "bz2" file. Setup in Ubuntu:
sudo apt install bzip2 wget -y

# Download the Boost Library
wget https://archives.boost.io/release/1.85.0/source/boost_1_85_0.tar.bz2

# Extract
tar -xjf boost_1_85_0.tar.bz2
rm -f boost_1_85_0.tar.bz2

# Setup Boost Install directory to current directory
WORK_BUILD_FOLDER=$(realpath .)
cd boost_1_85_0

# Configure to current system
./bootstrap.sh 
./b2 --clean

# Generate static libs
./b2 cxxflags="-march=x86-64" link=static variant=release linkflags=-static-libstdc++ -j8 cxxflags="-fPIC" --stagedir="${WORK_BUILD_FOLDER}/Boost" --with-program_options --with-system --with-regex --with-filesystem

mkdir -p ${WORK_BUILD_FOLDER}/Boost/include

# Generate symbolic link to headers inside Boost/include path
ln -sf ${WORK_BUILD_FOLDER}/boost_1_85_0/boost  ${WORK_BUILD_FOLDER}/Boost/include

# Generate shared/dynamic libs for tools, not needed for AlgoMarker
./b2 cxxflags="-march=x86-64" link=shared variant=release linkflags=-static-libstdc++ -j8 cxxflags="-fPIC" --stagedir="${WORK_BUILD_FOLDER}/Boost" --with-program_options --with-system --with-regex --with-filesystem

1. AlgoMarker Library

Description

The AlgoMarker shared library is used for deploying the AlgoMarker model. This library works with your final binary model output to access the model, apply it, and retrieve results. It provides a C-level API.

The Git repository is available at MR_LIBS Git Repository.

Installation

  1. Clone the Git repository: 
    1
    git clone git@github.com:Medial-EarlySign/MR_LIBS.git
  2. Compile the Boost library. Refer to Install Boost Libraries. You must compile the Boost library since the -fPIC flag is needed, and it is not included in Ubuntu packages.
  3. Edit Internal/AlgoMarker/CMakeLists.txt to include the following line: 
    1
    set(BOOST_ROOT "$ENV{HOME}/boost-pic-install")
    This should point to your Boost compiled home directory (WORK_BUILD_FOLDER) from the compilation step. Ensure the compiled libraries are in /libs and the headers are in /include.
  4. Execute: 
    1
    Internal/AlgoMarker/full_build.sh

2. AlgoMarker Wrapper

Description

The AlgoMarker Wrapper provides a REST API for the AlgoMarker C++ Library. There are two wrappers available:

  1. C++ Native Wrapper: Minimal dependencies, very fast, and efficient. Uses Boost Beast. Can be installed in a minimal Ubuntu Chiselled Docker image with just glibc.
  2. Python Wrapper: Built with FastAPI, more flexible for changes, and supports the old AlgoMarker API. It is slower and has more dependencies but is friendlier for testing.

Installation

C++ Native Wrapper:

  1. Set up Boost Libraries. No need to compile.
  2. Clone the repository: 
    1
    git clone git@github.com:Medial-EarlySign/MR_Tools.git
  3. If you compiled the Boost library, edit AlgoMarker_python_API/ServerHandler/CMakeLists.txt to include the following line: 
    1
    set(BOOST_ROOT "$ENV{HOME}/boost-pic-install")
    This should point to your Boost compiled home directory (WORK_BUILD_FOLDER) from the compilation step. Ensure the compiled libraries are in /libs and the headers are in /include.
  4. Compile the wrapper: 
    1
    AlgoMarker_python_API/ServerHandler/compile.sh

  5. Execute the server: 

    1
    AlgoMarker_python_API/ServerHandler/Linux/Release/AlgoMarker_Server --algomarker_path $AM_CONFIG --library_path $AM_LIB --port 1234

    • AM_CONFIG: Path to the AlgoMarker configuration file.
    • AM_LIB: Path to the AlgoMarker shared library. Refer to AlgoMarker Library for compilation steps.

Python Wrapper:

  1. Clone the repository: 
    1
    git clone git@github.com:Medial-EarlySign/MR_Tools.git
  2. Edit AlgoMarker_python_API/run_server.sh and update the following:
  • AM_CONFIG: Path to the AlgoMarker configuration file.
  • AM_LIB: Path to the AlgoMarker shared library. Refer to AlgoMarker Library for compilation steps.
  • If using the old ColonFlag, follow the steps in the ColonFlag setup page to compile the ICU library. Add the ICU library path to LD_LIBRARY_PATH in the script before calling uvicorn. 3. Make sure you have all python depenencies installed (The AlgoMarker.py itself has no dependencies if you want to use it directly with FastAPI): 
    1
    python -m pip install fastapi
    4. Run the Server AlgoMarker_python_API/run_server.sh

For more details follow: Python AlgoMarker API Wrapper

3. MES Tools to Train and Test Models

Description

Executables for training and testing models, along with other tools developed by MES for command-line use.

Installation

  1. Set up Boost Libraries. No need to compile.
  2. Clone the repositories: 
    1
2
    git clone git@github.com:Medial-EarlySign/MR_Tools.git
git clone git@github.com:Medial-EarlySign/MR_LIBS.git
  3. Navigate to the MR_Tools directory: 
    1
    cd MR_Tools
  4. Edit All_Tools/CMakeLists.txt to set LIBS_PATH to the MR_LIBS cloned directory. The default structure is: 
    1
2
3
    Root Directory
├── MR_LIBS
└── MR_Tools
    With this structure, no edits are needed. If you chose to compile Boost Libraries, don't forget to set BOOST_ROOT in the CMakeList.txt file
  5. Execute: 
    1
    AllTools/full_build.sh

4. Python API for MES Infrastructure

Description

A Python API Wrapper for the MES Infrastructure.

Installation

  1. Install the required libraries: 
    1
    sudo apt install python3-dev swig -y
  2. Compile the Boost library.
  3. Edit Internal/MedPyExport/generate_binding/CMakeLists.txt to include the following line: 
    1
    set(BOOST_ROOT "$ENV{HOME}/boost-pic-install")
    This should point to your Boost compiled home directory (WORK_BUILD_FOLDER) from the compilation step. Ensure the compiled libraries are in /libs and the headers are in /include.
  4. Make sure you have NumPy installed: 
    1
    python -m pip install numpy
    This library is compatible with both NumPy 1.x and 2.x. For the broadest compatibility, you should compile with NumPy 2.x, as this will also work for clients who have NumPy 1.x. However, compiling with NumPy 1.x will not be compatible with clients using 2.x.
  5. Execute: 
    1
    Internal/MedPyExport/generate_binding/make-simple.sh

Code to Load All Tools into the Environment

Start-Up Script
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
#!/bin/bash

# Path to Boost library
LD_PATH=${HOME}/Documents/MES/Boost/lib
# Path to Git Repo clones
MR_LIBS=${HOME}/Documents/MES/MR_LIBS
MR_TOOLS=${HOME}/Documents/MES/MR_Tools
MR_SCRIPTS=${HOME}/Documents/MES/MR_Scripts

export PATH=$PATH:${MR_TOOLS}/AllTools/Linux/Release:${MR_SCRIPTS}/Python-scripts:${MR_SCRIPTS}/Bash-Scripts:${MR_SCRIPTS}/Perl-scripts
if [ ! -z "$LD_LIBRARY_PATH" ]; then
    export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:${LD_PATH}
else
    export LD_LIBRARY_PATH=${LD_PATH}
fi
export PYTHONPATH=${MR_LIBS}/Internal/MedPyExport/generate_binding/Release/medial-python312:${MR_TOOLS}/RepoLoadUtils/common

Please edit LD_PATH, MR_LIBS, MR_TOOLS, and MR_SCRIPTS paths as needed.