Monte Carlo eXtreme (MCX) - CUDA Edition

  • Author: Qianqian Fang <q.fang at>
  • License: GNU General Public License version 3 (GPLv3)
  • Version: 1.8 (v2020, Furious Fermion)
  • Website:

Table of Content:

1. What's New
2. Introduction
3. Requirement and Installation
4. Running Simulations
5. Using JSON-formatted input files
6. Using JSON-formatted shape description files
7. Output data formats
8. Using MCXLAB in MATLAB and Octave
9. Using MCX Studio GUI
10. Interpreting the Output
11. Best practices guide
12. Acknowledgement
13. Reference

1. What's New

MCX v2020 represents a new milestone towards the development of a fast, versatile and feature-rich open-source Monte Carlo 3D photon simulator. It is packed with numerous improvements in both functionality and stability. We want to specifically highlight the below major additions:

  • Built-in benchmarks for easy testing and adoption by new users
  • Transition to JSON/JNIfTI input/output files for easy data sharing
  • Exporting simulation as JSON with binary volume data
  • All-in-one Windows installer for MCXStudio/MCX/MMC/MCXCL
  • Automated code building, testing and continuous integration via Travis-CI
  • CMake based compilation and Visual Studio support

A detailed list of updates is summarized below (key features marked with "*"):

  • 2020-08-20 [e8e6b58] print an explicit messgae if error 30 is found
  • 2020-08-19*[883f61b] restore windows progress bar support, disabled in @ae2d60e45
  • 2020-08-17 [c47de01] allow running testing script on machines without nvidia gpu
  • 2020-08-16 [0c25958] add more tests for various mcx options
  • 2020-08-16 [ff2f68f] add sphshell benchmark - see GPU MMC paper
  • 2020-08-15 [2afab4a] test if media prop count is less than max label
  • 2020-08-15 [2d71eb7] accept array as Domain.Media json input
  • 2020-08-15*[433df1f] accept json modifier via --json for easy testing
  • 2020-08-14 [09adbd0] support --bc or cfg.bc to set an entire bounding face as detector
  • 2020-08-11*[e095dbb] speed up by 1.6x on 1080Ti by restoring source template for pencil beam only
  • 2020-08-11 [a220cc2] autoblock size no less than 64, speed up on Turing GPU by doubling threads
  • 2020-08-04 [71d4196] fix incorrect detpt column when savedetflag/issaveexit are both set
  • 2020-08-04 [20c596a] retrieve energy tot and abs regardless of isnormalized
  • 2020-08-03*[30e01a1] add standalone script to submit to mcx speed contest
  • 2020-07-31 [d9a5953] avoid autoblock is 0 when driver fails, close #99
  • 2020-07-28 [daa9d56] fix inaccurate core count on Volta, Turing and Pascal
  • 2020-07-25 [37793ae] fix -b 0 -B rrrrrr crash, thanks to @ShijieYan
  • 2020-07-22*[f844fe8] add automated building script via travis-ci
  • 2020-07-22*[cbf0225] add unit testing script
  • 2020-07-09 [5b038a7] add winget manifest
  • 2020-07-04*[4bda593] inno windows installer
  • 2020-07-02 [38529a7] accept -f in mcxshow and mcxviewer
  • 2020-07-02 [34ecf5f] add glscene directly to the source code
  • 2020-07-01*[f1828d3] add manpage for mcx
  • 2020-06-29 [cd4acb8] visual studio project file updated
  • 2020-06-29 [b22025d] update vs project file
  • 2020-06-28 [71bedc5] add benchmark options, add jnii/bnii output formats
  • 2020-05-02 [e7ce8f7] compiles lzma on windows, #94
  • 2020-05-02*[e56fa2b] add UBJ support, output .bnii files, close #95
  • 2020-05-01 [ed80ad5] now support lzma and lzip compression
  • 2020-05-01 [f136bc9] upgrade all built-in binary files to JData formatted JSON files #94
  • 2020-04-29 [0d9c162] save detected photon data in JSON/JData format, close #94
  • 2020-04-25 [045a3de] update json schema
  • 2020-04-25 [e8aae66] initialize gsrcpattern
  • 2020-04-23*[8086175] add built-in colini27 data,add --dumpjson, add -F jnii output format
  • 2020-04-19*[da73b8d] add mcx built in benchmarks
  • 2020-03-21*[b8fb79a] plot data in x/y/z slices,add axis labels and grid
  • 2020-03-20 [d7e6203] add axis lable and scaling to volume viewer
  • 2020-03-14*[b91dc5a] update mcxstudio gui to support gpu mmc
  • 2020-02-18*[8c37911] adding pymcx written by Maxime Baillot as submodule
  • 2020-02-08 [ba78df5] add template to disable continuous medium support, close #89
  • 2020-01-28 [b7c1982] speed up cone beam photon launch, fix accuracy, close #86
  • 2020-01-25*[984b2a0] initial support for hybrid optical properties: tissue type label combined with continous optical properties
  • 2019-11-19 [1c07b16] scale partial-path when getting det photon time and weight, close #83
  • 2019-08-08 [0bdbef6] allow to browse file folder on windows
  • 2019-07-26*[8a341ee] update mcxstudio to add the new flags
  • 2019-07-23 [e3b53dc] add 2d sample script
  • 2019-07-22*[c4baa84] output fluence/flux in replay, backport changes from mcxcl
  • 2019-05-24 [02efc62] bug fix for continuous varying media patch

Between 2019 and 2020, four new journal papers have been published as the result of this project. Please see the full list at

2. Introduction

Monte Carlo eXtreme (MCX) is a fast photon transport simulation software for 3D heterogeneous turbid media. By taking advantage of the massively parallel threads and extremely low memory latency in a modern graphics processing unit (GPU), MCX is capable of performing Monte Carlo (MC) photon simulations at a blazing speed, typically hundreds to a thousand times faster than a fully optimized CPU-based MC implementation.

The algorithm of this software is detailed in the References [Fang2009,Yu2018]. A short summary of the main features includes:

  • 3D heterogeneous media represented by voxelated array
  • support over a dozen source forms, including wide-field and pattern illuminations
  • boundary reflection support
  • time-resolved photon transport simulations
  • saving photon partial path lengths and trajectories
  • optimized random number generators
  • build-in flux/fluence normalization to output Green's functions
  • user adjustable voxel resolution
  • improved accuracy with atomic operations
  • cross-platform graphical user interface
  • native Matlab/Octave support for high usability
  • flexible JSON interface for future extensions
  • multi-GPU support

This software can be used on Windows, Linux and Mac OS. MCX is written in C/CUDA and requires an NVIDIA GPU (support for AMD/Intel CPUs/GPUs via ROCm is still under development). A more portable OpenCL implementation of MCX, i.e. MCXCL, was announced on July, 2012 and supports almost all NVIDIA/AMD/Intel CPU and GPU models. If your hardware does not support CUDA, please download MCXCL from the below URL:

3. Requirement and Installation

Please read this section carefully. The majority of failures using MCX were found related to incorrect installation of NVIDIA GPU driver.

Please browse for step-by-step instructions.

For MCX-CUDA, the requirements for using this software include

  • a CUDA capable NVIDIA graphics card
  • pre-installed NVIDIA graphics driver

You must install a CUDA capable NVIDIA graphics card in order to use MCX. A list of CUDA capable cards can be found at [2]. The oldest graphics card that MCX supports is the Fermi series (circa 2010). Using the latest NVIDIA card is expected to produce the best speed. You must have a fermi (GTX 4xx) or newer (5xx/6xx/7xx/9xx/10xx/20xx series) graphics card. The default release of MCX supports atomic operations and photon detection. In the below webpage, we summarized the speed differences between different generations of NVIDIA GPUs

For simulations with large volumes, sufficient graphics memory is also required to perform the simulation. The minimum amount of graphics memory required for a MC simulation is Nx*Ny*Nz bytes for the input tissue data plus Nx*Ny*Nz*Ng*4 bytes for the output flux/fluence data - where Nx,Ny,Nz are the dimensions of the tissue volume, Ng is the number of concurrent time gates, 4 is the size of a single-precision floating-point number. MCX does not require double-precision capability in your hardware.

To install MCX, you need to download the binary executable compiled for your computer architecture (32 or 64bit) and platform, extract the package and run the executable under the <mcx root>/bin directory.

For Windows users, you must make sure you have installed the appropriate NVIDIA driver for your GPU. You should also configure your OS to run CUDA simulations. This requires you to open the mcx/setup/win64 folder using your file explorer, right-click on the "apply_timeout_registry_fix.bat" file and select "Run as administrator". After confirmation, you should see a windows command window with message

  Patching your registry
  Press any key to continue ...

You MUST REBOOT your Windows computer to make this setting effective. The above patch modifies your driver settings so that you can run MCX simulations for longer than a few seconds. Otherwise, when running MCX for over a few seconds, you will get a CUDA error: "unspecified error".

Please see the below link for details

If you use Linux, you may enable Intel integrated GPU (iGPU) for display while leaving your NVIDIA GPU dedicated for computing using nvidia-prime, see

or choose one of the 4 other approaches in this blog post

4. Running Simulations

To run a simulation, the minimum input is a configuration (text) file, and, if the input file does not contain built-in domain shape descriptions, an external volume file (a binary file with a specified voxel format via -K/--mediabyte). Typing mcx without any parameters prints the help information and a list of supported parameters, as shown below:

#                      Monte Carlo eXtreme (MCX) -- CUDA                      #
#          Copyright (c) 2009-2020 Qianqian Fang <q.fang at>          #
#                                                  #
#                                                                             #
# Computational Optics & Translational Imaging (COTI) Lab- #
#            Department of Bioengineering, Northeastern University            #
#    The MCX Project is funded by the NIH/NIGMS under grant R01-GM114365      #
$Rev::041e38$ v2020 $Date::2020-08-24 14:25:23 -04$ by $Author::Qianqian Fang $

usage: mcx <param1> <param2> ...
where possible parameters include (the first value in [*|*] is the default)

== Required option ==
 -f config     (--input)       read an input file in .json or .inp format
                               if the string starts with '{', it is parsed as
			       an inline JSON input file
 --bench ['cube60','skinvessel',..] run a buint-in benchmark specified by name
                               run --bench without parameter to get a list

== MC options ==
 -n [0|int]    (--photon)      total photon number (exponential form accepted)
                               max accepted value:9.2234e+18 on 64bit systems
 -r [1|+/-int] (--repeat)      if positive, repeat by r times,total= #photon*r
                               if negative, divide #photon into r subsets
 -b [1|0]      (--reflect)     1 to reflect photons at ext. boundary;0 to exit
 -B '______'   (--bc)          per-face boundary condition (BC), 6 letters for
    /case insensitive/         bounding box faces at -x,-y,-z,+x,+y,+z axes;
			       overwrite -b if given. 
			       each letter can be one of the following:
			       '_': undefined, fallback to -b
			       'r': like -b 1, Fresnel reflection BC
			       'a': like -b 0, total absorption BC
			       'm': mirror or total reflection BC
			       'c': cyclic BC, enter from opposite face

			       if input contains additional 6 letters,
			       the 7th-12th letters can be:
			       '0': do not use this face to detect photon, or
			       '1': use this face for photon detection (-d 1)
			       the order of the faces for letters 7-12 is 
			       the same as the first 6 letters
			       eg: --bc ______010 saves photons exiting at y=0
 -u [1.|float] (--unitinmm)    defines the length unit for the grid edge
 -U [1|0]      (--normalize)   1 to normalize flux to unitary; 0 save raw
 -E [0|int|mch](--seed)        set random-number-generator seed, -1 to generate
                               if an mch file is followed, MCX "replays" 
                               the detected photon; the replay mode can be used
                               to calculate the mua/mus Jacobian matrices
 -z [0|1]      (--srcfrom0)    1 volume origin is [0 0 0]; 0: origin at [1 1 1]
 -k [1|0]      (--voidtime)    when src is outside, 1 enables timer inside void
 -Y [0|int]    (--replaydet)   replay only the detected photons from a given 
                               detector (det ID starts from 1), used with -E 
			       if 0, replay all detectors and sum all Jacobians
			       if -1, replay all detectors and save separately
 -V [0|1]      (--specular)    1 source located in the background,0 inside mesh
 -e [0.|float] (--minenergy)   minimum energy level to trigger Russian roulette
 -g [1|int]    (--gategroup)   number of maximum time gates per run

== GPU options ==
 -L            (--listgpu)     print GPU information only
 -t [16384|int](--thread)      total thread number
 -T [64|int]   (--blocksize)   thread number per block
 -A [1|int]    (--autopilot)   1 let mcx decide thread/block size, 0 use -T/-t
 -G [0|int]    (--gpu)         specify which GPU to use, list GPU by -L; 0 auto
 -G '1101'     (--gpu)         using multiple devices (1 enable, 0 disable)
 -W '50,30,20' (--workload)    workload for active devices; normalized by sum
 -I            (--printgpu)    print GPU information and run program
 --atomic [1|0]                1: use atomic operations to avoid thread racing
                               0: do not use atomic operation (not recommended)

== Input options ==
 -P '{...}'    (--shapes)      a JSON string for additional shapes in the grid.
                               only the root object named 'Shapes' is parsed 
			       and added to the existing domain defined via -f 
			       or --bench
 -j '{...}'    (--json)        a JSON string for modifying all input settings.
                               this input can be used to modify all existing 
			       settings defined by -f or --bench
 -K [1|int|str](--mediabyte)   volume data format, use either a number or a str
                               1 or byte: 0-128 tissue labels
			       2 or short: 0-65535 (max to 4000) tissue labels
			       4 or integer: integer tissue labels 
			      99 or labelplus: 32bit composite voxel format
                             100 or muamus_float: 2x 32bit floats for mua/mus
                             101 or mua_float: 1 float per voxel for mua
			     102 or muamus_half: 2x 16bit float for mua/mus
			     103 or asgn_byte: 4x byte gray-levels for mua/s/g/n
			     104 or muamus_short: 2x short gray-levels for mua/s
 -a [0|1]      (--array)       1 for C array (row-major); 0 for Matlab array

== Output options ==
 -s sessionid  (--session)     a string to label all output file names
 -O [X|XFEJPM] (--outputtype)  X - output flux, F - fluence, E - energy density
    /case insensitive/         J - Jacobian (replay mode),   P - scattering, 
			       event counts at each voxel (replay mode only)
                               M - momentum transfer; 
 -d [1|0]      (--savedet)     1 to save photon info at detectors; 0 not save
 -w [DP|DSPMXVW](--savedetflag)a string controlling detected photon data fields
    /case insensitive/         1 D  output detector ID (1)
                               2 S  output partial scat. even counts (#media)
                               4 P  output partial path-lengths (#media)
			       8 M  output momentum transfer (#media)
			      16 X  output exit position (3)
			      32 V  output exit direction (3)
			      64 W  output initial weight (1)
      combine multiple items by using a string, or add selected numbers together
      by default, mcx only saves detector ID and partial-path data
 -x [0|1]      (--saveexit)    1 to save photon exit positions and directions
                               setting -x to 1 also implies setting '-d' to 1.
			       same as adding 'XV' to -w.
 -X [0|1]      (--saveref)     1 to save diffuse reflectance at the air-voxels
                               right outside of the domain; if non-zero voxels
			       appear at the boundary, pad 0s before using -X
 -m [0|1]      (--momentum)    1 to save photon momentum transfer,0 not to save.
                               same as adding 'M' to the -w flag
 -q [0|1]      (--saveseed)    1 to save photon RNG seed for replay; 0 not save
 -M [0|1]      (--dumpmask)    1 to dump detector volume masks; 0 do not save
 -H [1000000] (--maxdetphoton) max number of detected photons
 -S [1|0]      (--save2pt)     1 to save the flux field; 0 do not save
 -F [mc2|...] (--outputformat) fluence data output format:
                               mc2 - MCX mc2 format (binary 32bit float)
                               jnii - JNIfTI format (
                               bnii - Binary JNIfTI (
                               nii - NIfTI format
                               hdr - Analyze 7.5 hdr/img format
                               tx3 - GL texture data for rendering (GL_RGBA32F)
	the bnii/jnii formats support compression (-Z) and generate small files
	load jnii (JSON) and bnii (UBJSON) files using below lightweight libs:
	  MATLAB/Octave: JNIfTI toolbox, 
	  MATLAB/Octave: JSONLab toolbox, 
	  Python:        PyJData:
	  JavaScript:    JSData:
 -Z [zlib|...] (--zip)         set compression method if -F jnii or --dumpjson
                               is used (when saving data to JSON/JNIfTI format)
			       0 zlib: zip format (moderate compression,fast) 
			       1 gzip: gzip format (compatible with *.gz)
			       2 base64: base64 encoding with no compression
			       3 lzip: lzip format (high compression,very slow)
			       4 lzma: lzma format (high compression,very slow)
			       5 lz4: LZ4 format (low compression,extrem. fast)
			       6 lz4hc: LZ4HC format (moderate compression,fast)
 --dumpjson [-,0,1,'file.json']  export all settings, including volume data using
                               JSON/JData ( format for 
			       easy sharing; can be reused using -f
			       if followed by nothing or '-', mcx will print
			       the JSON to the console; write to a file if file
			       name is specified; by default, prints settings
			       after pre-processing; '--dumpjson 2' prints 
			       raw inputs before pre-processing

== User IO options ==
 -h            (--help)        print this message
 -v            (--version)     print MCX revision number
 -l            (--log)         print messages to a log file instead
 -i 	       (--interactive) interactive mode

== Debug options ==
 -D [0|int]    (--debug)       print debug information (you can use an integer
  or                           or a string by combining the following flags)
 -D [''|RMP]                   1 R  debug RNG
    /case insensitive/         2 M  store photon trajectory info
                               4 P  print progress bar
      combine multiple items by using a string, or add selected numbers together

== Additional options ==
 --root         [''|string]    full path to the folder storing the input files
 --gscatter     [1e9|int]      after a photon completes the specified number of
                               scattering events, mcx then ignores anisotropy g
                               and only performs isotropic scattering for speed
 --internalsrc  [0|1]          set to 1 to skip entry search to speedup launch
 --maxvoidstep  [1000|int]     maximum distance (in voxel unit) of a photon that
                               can travel before entering the domain, if 
                               launched outside (i.e. a widefield source)
 --maxjumpdebug [10000000|int] when trajectory is requested (i.e. -D M),
                               use this parameter to set the maximum positions
                               stored (default: 1e7)

== Example ==
example: (list built-in benchmarks)
       mcx --bench
or (list supported GPUs on the system)
       mcx -L
or (use multiple devices - 1st,2nd and 4th GPUs - together with equal load)
       mcx --bench cube60b -n 1e7 -G 1101 -W 10,10,10
or (use inline domain definition)
       mcx -f input.json -P '{"Shapes":[{"ZLayers":[[1,10,1],[11,30,2],[31,60,3]]}]}'
or (use inline json setting modifier)
       mcx -f input.json -j '{"Optode":{"Source":{"Type":"isotropic"}}}'
or (dump simulation in a single json file)
       mcx --bench cube60planar --dumpjson

To further illustrate the command line options, below one can find a sample command

  mcx -A 0 -t 16384 -T 64 -n 1e7 -G 1 -f input.json -r 2 -s test -g 10 -d 1 -w dpx -b 1

the command above asks mcx to manually (-A 0) set GPU threads, and launch 16384 GPU threads (-t) with every 64 threads a block (-T); a total of 1e7 photons (-n) are simulated by the first GPU (-G 1) and repeat twice (-r) - i.e. total 2e7 photons; the media/source configuration will be read from a JSON file named input.json (-f) and the output will be labeled with the session id “test” (-s); the simulation will run 10 concurrent time gates (-g) if the GPU memory can not simulate all desired time gates at once. Photons passing through the defined detector positions are saved for later rescaling (-d), and the saved photon data include detector id (letter 'd' in -w), partial path (letter 'p' in -w) and exit position (letter 'x' in -w); refractive index mismatch is considered at media boundaries (-b).

Historically, MCX supports an extended version of the input file format (.inp) used by tMCimg. However, we are phasing out the .inp support and strongly encourage users to adopt JSON formatted (.json) input files. Many of the advanced MCX options are only supported in the JSON input format.

A legacy .inp MCX input file looks like this:

1000000              # total photon, use -n to overwrite in the command line
29012392             # RNG seed, negative to generate, use -E to overwrite
30.0 30.0 0.0 1      # source position (in grid unit), the last num (optional) sets --srcfrom0 (-z)
0 0 1 0              # initial directional vector, 4th number is the focal-length, 0 for collimated beam, nan for isotropic
0.e+00 1.e-09 1.e-10 # time-gates(s): start, end, step
semi60x60x60.bin     # volume ('unsigned char' binary format, or specified by -K/--mediabyte)
1 60 1 60            # x voxel size in mm (isotropic only), dim, start/end indices
1 60 1 60            # y voxel size, must be same as x, dim, start/end indices 
1 60 1 60            # y voxel size, must be same as x, dim, start/end indices
1                    # num of media
1.010101 0.01 0.005 1.37  # scat. mus (1/mm), g, mua (1/mm), n
4       1.0          # detector number and default radius (in grid unit)
30.0  20.0  0.0  2.0 # detector 1 position (real numbers in grid unit) and individual radius (optional)
30.0  40.0  0.0      # ..., if individual radius is ignored, MCX will use the default radius
20.0  30.0  0.0      #
40.0  30.0  0.0      # 
pencil               # source type (optional)
0 0 0 0              # parameters (4 floats) for the selected source
0 0 0 0              # additional source parameters

Note that the scattering coefficient mus=musp/(1-g).

The volume file (semi60x60x60.bin in the above example), can be read in two ways by MCX: row-major[3] or column-major depending on the value of the user parameter "-a". If the volume file was saved using matlab or fortran, the byte order is column-major, and you should use "-a 0" or leave it out of the command line. If it was saved using the fwrite() in C, the order is row-major, and you can either use "-a 1".

You may replace the binary volume file by a JSON-formatted shape file. Please refer to Section V for details.

The time gate parameter is specified by three numbers: start time, end time and time step size (in seconds). In the above example, the configuration specifies a total time window of [0 1] ns, with a 0.1 ns resolution. That means the total number of time gates is 10.

MCX provides an advanced option, -g, to run simulations when the GPU memory is limited. It specifies how many time gates to simulate concurrently (when the GPU does not have sufficient memory to simulate all desired time gates all together). Users may want to limit that number to less than the total number specified in the input file - and by default it runs one gate at a time in a single simulation. But if there's enough memory based on the memory requirement in Section II, you can simulate all 10 time gates (from the above example) concurrently by using "-g 10" in which case you have to make sure the video card has at least 60*60*60*10*5=10MB of free memory. If you do not include the -g, MCX will assume you want to simulate just 1 time gate at a time.. If you specify a time-gate number greater than the total number in the input file, (e.g, "-g 20") MCX will stop when the 10 time-gates are completed. If you use the autopilot mode (-A), then the time-gates are automatically estimated for you.

5. Using JSON-formatted input files

Starting from version 0.7.9, MCX accepts a JSON-formatted input file in addition to the legacy .inp input files. JSON (JavaScript Object Notation) is a portable, human-readable and "fat-free" text format to represent complex and hierarchical data. Using the JSON format makes a input file self-explanatory, extensible and easy-to-interface with other applications (like MATLAB and Python).

A sample JSON input file can be found under the examples/quicktest folder. The same file, qtest.json, is also shown below:

    "Help": {
      "[en]": {
        "Domain::VolumeFile": "file full path to the volume description file, can be a binary or JSON file",
        "Domain::Dim": "dimension of the data array stored in the volume file",
        "Domain::OriginType": "similar to --srcfrom0, 1 if the origin is [0 0 0], 0 if it is [1.0,1.0,1.0]",
	"Domain::LengthUnit": "define the voxel length in mm, similar to --unitinmm",
        "Domain::Media": "the first medium is always assigned to voxels with a value of 0 or outside of
                         the volume, the second row is for medium type 1, and so on. mua and mus must 
                         be in 1/mm unit",
        "Session::Photons": "if -n is not specified in the command line, this defines the total photon number",
        "Session::ID": "if -s is not specified in the command line, this defines the output file name stub",
        "Forward::T0": "the start time of the simulation, in seconds",
        "Forward::T1": "the end time of the simulation, in seconds",
        "Forward::Dt": "the width of each time window, in seconds",
        "Optode::Source::Pos": "the grid position of the source, can be non-integers, in grid unit",
        "Optode::Detector::Pos": "the grid position of a detector, can be non-integers, in grid unit",
        "Optode::Source::Dir": "the unitary directional vector of the photon at launch",
        "Optode::Source::Type": "source types, must be one of the following: 
        "Optode::Source::Param1": "source parameters, 4 floating-point numbers",
        "Optode::Source::Param2": "additional source parameters, 4 floating-point numbers"
    "Domain": {
	"VolumeFile": "semi60x60x60.bin",
        "Dim":    [60,60,60],
        "OriginType": 1,
	"LengthUnit": 1,
        "Media": [
             {"mua": 0.00, "mus": 0.0, "g": 1.00, "n": 1.0},
             {"mua": 0.005,"mus": 1.0, "g": 0.01, "n": 1.0}
    "Session": {
	"Photons":  1000000,
	"RNGSeed":  29012392,
	"ID":       "qtest"
    "Forward": {
	"T0": 0.0e+00,
	"T1": 5.0e-09,
	"Dt": 5.0e-09
    "Optode": {
	"Source": {
	    "Pos": [29.0, 29.0, 0.0],
	    "Dir": [0.0, 0.0, 1.0],
	    "Type": "pencil",
	    "Param1": [0.0, 0.0, 0.0, 0.0],
	    "Param2": [0.0, 0.0, 0.0, 0.0]
	"Detector": [
		"Pos": [29.0,  19.0,  0.0],
		"R": 1.0
                "Pos": [29.0,  39.0,  0.0],
                "R": 1.0
                "Pos": [19.0,  29.0,  0.0],
                "R": 1.0
                "Pos": [39.0,  29.0,  0.0],
                "R": 1.0

A JSON input file requiers several root objects, namely "Domain", "Session", "Forward" and "Optode". Other root sections, like "Help", will be ignored. Each object is a data structure providing information indicated by its name. Each object can contain various sub-fields. The orders of the fields in the same level are flexible. For each field, you can always find the equivalent fields in the *.inp input files. For example, The "VolumeFile" field under the "Domain" object is the same as Line#6 in qtest.inp; the "RNGSeed" under "Session" is the same as Line#2; the "Optode.Source.Pos" is the same as the triplet in Line#3; the "Forward.T0" is the same as the first number in Line#5, etc.

An MCX JSON input file must be a valid JSON text file. You can validate your input file by running a JSON validator, for example You should always use "" to quote a "name" and separate parallel items by ",".

MCX accepts an alternative form of JSON input, but using it is not recommended. In the alternative format, you can use "rootobj_name.field_name": value to represent any parameter directly in the root level. For example

    "Domain.VolumeFile": "semi60x60x60.json",
    "Session.Photons": 10000000,

You can even mix the alternative format with the standard format. If any input parameter has values in both formats in a single input file, the standard-formatted value has higher priority.

To invoke the JSON-formatted input file in your simulations, you can use the "-f" command line option with MCX, just like using an .inp file. For example:

  mcx -A 1 -n 20 -f onecube.json -s onecubejson

The input file must have a ".json" suffix in order for MCX to recognize. If the input information is set in both command line, and input file, the command line value has higher priority (this is the same for .inp input files). For example, when using "-n 20", the value set in "Session"/"Photons" is overwritten to 20; when using "-s onecubejson", the "Session"/"ID" value is modified. If your JSON input file is invalid, MCX will quit and point out where the format is incorrect.

6. Using JSON-formatted shape description files

Starting from v0.7.9, MCX can also use a shape description file in the place of the volume file. Using a shape-description file can save you from making a binary .bin volume. A shape file uses more descriptive syntax and can be easily understood and shared with others.

Samples on how to use the shape files are included under the example/shapetest folder.

The sample shape file, shapes.json, is shown below:

     "Shapes::Common Rules": "Shapes is an array object. The Tag field sets the voxel value for each
         region; if Tag is missing, use 0. Tag must be smaller than the maximum media number in the
         input file.Most parameters are in floating-point (FP). If a parameter is a coordinate, it
         assumes the origin is defined at the lowest corner of the first voxel, unless user overwrite
         with an Origin object. The default origin of all shapes is initialized by user's --srcfrom0
         setting: if srcfrom0=1, the lowest corner of the 1st voxel is [0,0,0]; otherwise, it is [1,1,1]",
     "Shapes::Name": "Just for documentation purposes, not parsed in MCX",
     "Shapes::Origin": "A floating-point (FP) triplet, set coordinate origin for the subsequent objects",
     "Shapes::Grid": "Recreate the background grid with the given dimension (Size) and fill-value (Tag)",
     "Shapes::Sphere": "A 3D sphere, centered at C0 with radius R, both have FP values",
     "Shapes::Box": "A 3D box, with lower corner O and edge length Size, both have FP values",
     "Shapes::SubGrid": "A sub-section of the grid, integer O- and Size-triplet, inclusive of both ends",
     "Shapes::XLayers/YLayers/ZLayers": "Layered structures, defined by an array of integer triples:
          [start,end,tag]. Ends are inclusive in MATLAB array indices. XLayers are perpendicular to x-axis, and so on",
     "Shapes::XSlabs/YSlabs/ZSlabs": "Slab structures, consisted of a list of FP pairs [start,end]
          both ends are inclusive in MATLAB array indices, all XSlabs are perpendicular to x-axis, and so on",
     "Shapes::Cylinder": "A finite cylinder, defined by the two ends, C0 and C1, along the axis and a radius R",
     "Shapes::UpperSpace": "A semi-space defined by inequality A*x+B*y+C*z>D, Coef is required, but not Equ"
  "Shapes": [
     {"Name":     "Test"},
     {"Origin":   [0,0,0]},
     {"Grid":     {"Tag":1, "Size":[40,60,50]}},
     {"Sphere":   {"Tag":2, "O":[30,30,30],"R":20}},
     {"Box":      {"Tag":0, "O":[10,10,10],"Size":[10,10,10]}},
     {"Subgrid":  {"Tag":1, "O":[13,13,13],"Size":[5,5,5]}},
     {"XSlabs":   {"Tag":4, "Bound":[[5,15],[35,40]]}},
     {"Cylinder": {"Tag":2, "C0": [0.0,0.0,0.0], "C1": [15.0,8.0,10.0], "R": 4.0}},
     {"ZLayers":  [[1,10,1],[11,30,2],[31,50,3]]}

A shape file must contain a "Shapes" object in the root level. Other root-level fields are ignored. The "Shapes" object is a JSON array, with each element representing a 3D object or setting. The object-class commands include "Grid", "Sphere", "Box" etc. Each of these object include a number of sub-fields to specify the parameters of the object. For example, the "Sphere" object has 3 subfields, "O", "R" and "Tag". Field "O" has a value of 1x3 array, representing the center of the sphere; "R" is a scalar for the radius; "Tag" is the voxel values. The most useful command is "[XYZ]Layers". It contains a series of integer triplets, specifying the starting index, ending index and voxel value of a layered structure. If multiple objects are included, the subsequent objects always overwrite the overlapping regions covered by the previous objects.

There are a few ways for you to use shape description records in your MCX simulations. You can save it to a JSON shape file, and put the file name in Line#6 of your .inp file, or set as the value for Domain.VolumeFile field in a .json input file. In these cases, a shape file must have a suffix of .json.

You can also merge the Shapes section with a .json input file by simply appending the Shapes section to the root-level object. You can find an example, jsonshape_allinone.json, under examples/shapetest. In this case, you no longer need to define the "VolumeFile" field in the input.

Another way to use Shapes is to specify it using the -P (or --shapes) command line flag. For example:

 mcx -f input.json -P '{"Shapes":[{"ZLayers":[[1,10,1],[11,30,2],[31,60,3]]}]}'

This will first initialize a volume based on the settings in the input .json file, and then rasterize new objects to the domain and overwrite regions that are overlapping.

For both JSON-formatted input and shape files, you can use the JSONlab toolbox [4] to load and process in MATLAB.

7. Output data formats

MCX may produces several output files depending user's simulation settings. Overall, MCX produces two types of outputs, 1) data accummulated within the 3D volume of the domain (volumetric output), and 2) data stored for each detected photon (detected photon data).

Volumetric output

By default, MCX stores a 4D array denoting the fluence-rate at each voxel in the volume, with a dimension of Nx*Ny*Nz*Ng, where Nx/Ny/Nz are the voxel dimension of the domain, and Ng is the total number of time gates. The output data are stored in the format of single-precision floating point numbers. One may choose to output different physical quantities by setting the -O option. When the flag -X/--saveref is used, the output volume may contain the total diffuse reflectance only along the background-voxels adjacent to non-zero voxels. A negative sign is added for the diffuse reflectance raw output to distinguish it from the fuence data in the interior voxels.

When photon-sharing (simultaneous simulations of multiple patterns) or photon-replay (the Jacobian of all source/detector pairs) is used, the output array may be extended to a 5D array, with the left-most/fastest index being the number of patterns Ns (in the case of photon-sharing) or src/det pairs (in replay), denoted as Ns.

Several data formats can be used to store the 3D/4D/5D volumetric output.

mc2 files

The .mc2 format is simply a binary dump of the entire volumetric data output, consisted of the voxel values (single-precision floating-point) of all voxels and time gates. The file contains a continuous buffer of a single-precision (4-byte) 5D array of dimension Ns*Nx*Ny*Nz*Ng, with the fastest index being the left-most dimension (i.e. column-major, similar to MATLAB/FORTRAN).

To load the mc2 file, one should call loadmc2.m and must provide explicitly the dimensions of the data. This is because mc2 file does not contain the data dimension information.

Saving to .mc2 volumetric file is depreciated as we are transitioning towards JNIfTI/JData formatted outputs (.jnii).

nii files

The NIfTI-1 (.nii) format is widely used in neuroimaging and MRI community to store and exchange ND numerical arrays. It contains a 352 byte header, followed by the raw binary stream of the output data. In the header, the data dimension information as well as other metadata is stored.

A .nii output file can be generated by using -F nii in the command line.

The .nii file is widely supported among data processing platforms, including MATLAB and Python. For example

jnii files

The JNIfTI format represents the next-generation scientific data storage and exchange standard and is part of the OpenJData initiative ( led by the MCX author Dr. Qianqian Fang. The OpenJData project aims at developing easy-to-parse, human-readable and easy-to-reuse data storage formats based on the ubiquitously supported JSON/binary JSON formats and portable JData data annotation keywords. In short, .jnii file is simply a JSON file with capability of storing binary strongly-typed data with internal compression and built in metadata.

The format standard (Draft 1) of the JNIfTI file can be found at

A .jnii output file can be generated by using -F jnii in the command line.

The .jnii file can be potentially read in nearly all programming languages because it is 100% comaptible to the JSON format. However, to properly decode the ND array with built-in compression, one should call JData compatible libraries, which can be found at

Specifically, to parse/save .jnii files in MATLAB, you should use

To parse/save .jnii files in Python, you should use

In Python, the volumetric data is loaded as a dict object where data['NIFTIData'] is a NumPy ndarray object storing the volumetric data.

bnii files

The binary JNIfTI file is also part of the JNIfTI specification and the OpenJData project. In comparison to text-based JSON format, .bnii files can be much smaller and faster to parse. The .bnii format is also defined in the BJData specification

and is the binary interface to .jnii. A .bnii output file can be generated by using -F bnii in the command line.

The .bnii file can be potentially read in nearly all programming languages because it was based on UBJSON (Universal Binary JSON). However, to properly decode the ND array with built-in compression, one should call JData compatible libraries, which can be found at

Specifically, to parse/save .bnii files in MATLAB, you should use

To parse/save .jnii files in Python, you should use

In Python, the volumetric data is loaded as a dict object where data['NIFTIData'] is a NumPy ndarray object storing the volumetric data.

Detected photon data

If one defines detectors, MCX is able to store a variety of photon data when a photon is captured by these detectors. One can selectively store various supported data fields, including partial pathlengths, exit position and direction, by using the -w/--savedetflag flag. The storage of detected photon information is enabled by default, and can be disabled using the -d flag.

The detected photon data are stored in a separate file from the volumetric output. The supported data file formats are explained below.

mch files

The .mch file, or MC history file, is stored by default, but we strongly encourage users to adpot the newly implemented JSON/.jdat format for easy data sharing.

The .mch file contains a 256 byte binary header, followed by a 2-D numerical array of dimensions #savedphoton * #colcount as recorded in the header.

 typedef struct MCXHistoryHeader{
	char magic[4];                 /**< magic bits= 'M','C','X','H' */
	unsigned int  version;         /**< version of the mch file format */
	unsigned int  maxmedia;        /**< number of media in the simulation */
	unsigned int  detnum;          /**< number of detectors in the simulation */
	unsigned int  colcount;        /**< how many output files per detected photon */
	unsigned int  totalphoton;     /**< how many total photon simulated */
	unsigned int  detected;        /**< how many photons are detected (not necessarily all saved) */
	unsigned int  savedphoton;     /**< how many detected photons are saved in this file */
	float unitinmm;                /**< what is the voxel size of the simulation */
	unsigned int  seedbyte;        /**< how many bytes per RNG seed */
        float normalizer;              /**< what is the normalization factor */
	int respin;                    /**< if positive, repeat count so total photon=totalphoton*respin; if negative, total number is processed in respin subset */
	unsigned int  srcnum;          /**< number of sources for simultaneous pattern sources */
	unsigned int  savedetflag;     /**< number of sources for simultaneous pattern sources */
	int reserved[2];               /**< reserved fields for future extension */
 } History;

When the -q flag is set to 1, the detected photon initial seeds are also stored following the detected photon data, consisting of a 2-D byte array of #savedphoton * #seedbyte.

To load the mch file, one should call loadmch.m in MATLAB/Octave.

Saving to .mch history file is depreciated as we are transitioning towards JSON/JData formatted outputs (.jdat).

jdat files

When -F jnii is specified, instead of saving the detected photon into the legacy .mch format, a .jdat file is written, which is a pure JSON file. This file contains a hierachical data record of the following JSON structure

   "MCXData": {

where "Info" is required, and other subfields are optional depends on users' input. Each subfield in this file may contain JData 1-D or 2-D array constructs to allow storing binary and compressed data.

Although .jdat and .jnii have different suffix, they are both JSON/JData files and can be opened/written by the same JData compatible libraries mentioned above, i.e.


For Python

In Python, the volumetric data is loaded as a dict object where data['MCXData']['PhotonData'] stores the photon data, data['MCXData']['Trajectory'] stores the trajectory data etc.

Photon trajectory data

For debugging and plotting purposes, MCX can output photon trajectories, as polylines, when -D M flag is attached, or mcxlab is asked for the 5th output. Such information can be stored in one of the following formats.

mct files

By default, MCX stores the photon trajectory data in to a .mct file MC trajectory, which uses the same binary format as .mch but renamed as .mct. This file can be loaded to MATLAB using the same loadmch.m function.

Using .mct file is depreciated and users are encouraged to migrate to .jdat file as described below.

jdat files

When -F jnii is used, MCX merges the trajectory data with the detected photon and seed data and saved as a JSON-compatible .jdat file. The overall structure of the .jdat file as well as the relevant parsers can be found in the above section.

8. Using MCXLAB in MATLAB and Octave

MCXLAB is the native MEX version of MCX for Matlab and GNU Octave. It includes the entire MCX code in a MEX function which can be called directly inside Matlab or Octave. The input and output files in MCX are replaced by convenient in-memory struct variables in MCXLAB, thus, making it much easier to use and interact. Matlab/Octave also provides convenient plotting and data analysis functions. With MCXLAB, your analysis can be streamlined and speed- up without involving disk files.

Please read the mcxlab/README.txt file for more details on how to install and use MCXLAB.

9. Using MCX Studio GUI

MCX Studio is a graphics user interface (GUI) for MCX. It gives users a straightforward way to set the command line options and simulation parameters. It also allows users to create different simulation tasks and organize them into a project and save for later use. MCX Studio can be run on many platforms such as Windows, GNU Linux and Mac OS.

To use MCX Studio, it is suggested to put the mcxstudio binary in the same directory as the mcx command; alternatively, you can also add the path to mcx command to your PATH environment variable.

Once launched, MCX Studio will automatically check if mcx binary is in the search path, if so, the "GPU" button in the toolbar will be enabled. It is suggested to click on this button once, and see if you can see a list of GPUs and their parameters printed in the output field at the bottom part of the window. If you are able to see this information, your system is ready to run MCX simulations. If you get error messages or not able to see any usable GPU, please check the following:

  • are you running MCX Studio/MCX on a computer with a supported card?
  • have you installed the CUDA/NVIDIA drivers correctly?
  • did you put mcx in the same folder as mcxstudio or add its path to PATH?

If your system has been properly configured, you can now add new simulations by clicking the "New" button. MCX Studio will ask you to give a session ID string for this new simulation. Then you are allowed to adjust the parameters based on your needs. Once you finish the adjustment, you should click the "Verify" button to see if there are missing settings. If everything looks fine, the "Run" button will be activated. Click on it once will start your simulation. If you want to abort the current simulation, you can click the "Stop" button.

You can create multiple tasks with MCX Studio by hitting the "New" button again. The information for all session configurations can be saved as a project file (with .mcxp extension) by clicking the "Save" button. You can load a previously saved project file back to MCX Studio by clicking the "Load" button.

10. Interpreting the Output

MCX's output consists of two parts, the fluence volume file (.mc2, .nii, .jnii etc) and the detected photon data (.mch, .jdat etc).

Output files

An mc2/nii/jnii file contains the fluence-rate distributions from the simulation in the given medium. By default, this fluence-rate is a normalized solution (as opposed to the raw probability) therefore, one can compare this directly to analytical solutions (i.e. Green's function) of RTE/DE. The dimensions of the volume contained in this file are Nx, Ny, Nz, and Ng where Ng is the total number of time gates.

By default, MCX produces the Green's function of the fluence rate for the given domain and source. Sometime it is also known as the time-domain "two-point" function. If you run MCX with the following command

  mcx -f input.json -s output ....

the fluence-rate data will be saved in a file named "output.mc2" under the current folder. If you run MCX without "-s output", the output file will be named as "input.json.dat".

To understand this further, you need to know that a fluence-rate (Phi(r,t)) is measured by number of particles passing through an infinitesimal spherical surface per unit time at a given location regardless of directions. The unit of the MCX output is "W/mm2 = J/(mm<sup>2s)", if it is interpreted as the "energy fluence-rate" [6], or "1/(mm2s)", if the output is interpreted as the "particle fluence-rate" [6].

The Green's function of the fluence-rate means that it is produced by a unitary source. In simple terms, this represents the fraction of particles/energy that arrives a location per second under '''the radiation of 1 unit (packet or J) of particle or energy at time t=0'''. The Green's function is calculated by a process referred to as the "normalization" in the MCX code and is detailed in the MCX paper [6] (MCX and MMC outputs share the same meanings).

Please be aware that the output flux is calculated at each time-window defined in the input file. For example, if you type

  0.e+00 5.e-09 1e-10  # time-gates(s): start, end, step

in the 5th row in the input file, MCX will produce 50 fluence-rate snapshots, corresponding to the time-windows at [0 0.1] ns, [0.1 0.2]ns ... and [4.9,5.0] ns. To convert the fluence rate to the fluence for each time-window, you just need to multiply each solution by the width of the window, 0.1 ns in this case.

To convert the time-dependent fluence-rate to continuous-wave (CW) fluence (fluence in short), you need to integrate the fluence-rate along the time dimension. Assuming the fluence-rate after 5 ns is negligible, then the CW fluence is simply sum(flux_i*0.1 ns, i=1,50). You can read mcx/examples/validation/plotsimudata.m and mcx/examples/sphbox/plotresults.m for examples to compare an MCX output with the analytical fluence-rate/fluence solutions.

One can load an mc2 output file into Matlab or Octave using the loadmc2 function in the <mcx root>/utils folder.

To get a continuous-wave solution, run a simulation with a sufficiently long time window, and sum the flux along the time dimension, for example

   mcx=loadmc2('output.mc2',[60 60 60 10],'float');

Note that for time-resolved simulations, the corresponding solution in the results approximates the flux at the center point of each time window. For example, if the simulation time window setting is [t0,t0+dt,t0+2dt,t0+3dt...,t1], the time points for the snapshots stored in the solution file is located at [t0+dt/2, t0+3*dt/2, t0+5*dt/2, ... ,t1-dt/2]

A more detailed interpretation of the output data can be found at

MCX can also output "current density" (J(r,t), unit W/m^2, same as Phi(r,t)) - referring to the expected number of photons or Joule of energy flowing through a unit area pointing towards a particular direction per unit time. The current density can be calculated at the boundary of the domain by two means:

  1. using the detected photon partial path output (i.e. the second output of mcxlab.m),
one can compute the total energy E received by a detector, then one can divide E by the area/aperture of the detector to obtain the J(r) at a detector (E should be calculated as a function of t by using the time-of-fly of detected photons, the E(t)/A gives J(r,t); if you integrate all time gates, the total E/A gives the current I(r), instead of the current density).

  1. use -X 1 or --saveref/cfg.issaveref option in mcx to enable the
diffuse reflectance recordings on the boundary. the diffuse reflectance is represented by the current density J(r) flowing outward from the domain.

The current density has, as mentioned, the same unit as fluence rate, but the difference is that J(r,t) is a vector, and Phi(r,t) is a scalar. Both measuring the energy flow across a small area (the are has direction in the case of J) per unit time.

You can find more rigorous definitions of these quantities in Lihong Wang's Biomedical Optics book, Chapter 5.

Console print messages

Timing information is printed on the screen (stdout). The clock starts (at time T0) right before the initialization data is copied from CPU to GPU. For each simulation, the elapsed time from T0 is printed (in ms). Also the accumulated elapsed time is printed for all memory transaction from GPU to CPU.

When a user specifies "-D P" in the command line, or set cfg.debuglevel='P', MCX or MCXLAB prints a progress bar showing the percentage of completition.

11. Best practices guide

To maximize MCX's performance on your hardware, you should follow the best practices guide listed below:

Use dedicated GPUs

A dedicated GPU is a GPU that is not connected to a monitor. If you use a non-dedicated GPU, any kernel (GPU function) can not run more than a few seconds. This greatly limits the efficiency of MCX. To set up a dedicated GPU, it is suggested to install two graphics cards on your computer, one is set up for displays, the other one is used for GPU computation only. If you have a dual-GPU card, you can also connect one GPU to a single monitor, and use the other GPU for computation (selected by -G in mcx). If you have to use a non-dedicated GPU, you can either use the pure command-line mode (for Linux, you need to stop X server), or use the "-r" flag to divide the total simulation into a set of simulations with less photons, so that each simulation only lasts a few seconds.

Launch as many threads as possible

It has been shown that MCX's speed is related to the thread number (-t). Generally, the more threads, the better speed, until all GPU resources are fully occupied. For higher-end GPUs, a thread number over 10,000 is recommended. Please use the autopilot mode, "-A", to let MCX determine the "optimal" thread number when you are not sure what to use.

12. Acknowledgement

cJSON library by Dave Gamble

GLScene library for Lazarus by GLScene developers

Texture3D sample project by Jürgen Abel

Synapse communication library for Lazarus

  • Files: mcxstudio/synapse/*
  • Copyright (c) 1999-2017, Lukas Gebauer
  • URL:
  • License: MIT License or LGPL version 2 or later or GPL version 2 or later
  • Comment: A subset of the Synapse units is included as part of the MCX source code tree to allow compilation of the MCX Studio binary on various platforms without needing to install the full package.

ZMat data compression unit

LZ4 data compression library

LZMA/Easylzma data compression library

  • Files: src/zmat/easylzma/*
  • Copyright: 2009, Lloyd Hilaiel, 2008, Igor Pavlov
  • License: public-domain
  • Comment: All the cruft you find here is public domain. You don't have to credit anyone to use this code, but my personal request is that you mention Igor Pavlov for his hard, high quality work.

myslicer toolbox by Anders Brun

MCX Filter submodule

pymcx Python module

13. Reference

  • [Fang2009] Qianqian Fang and David A. Boas, "Monte Carlo Simulation of Photon Migration in 3D Turbid Media Accelerated by Graphics Processing Units,"
Optics Express, vol. 17, issue 22, pp. 20178-20190 (2009).

  • [Yu2018] Leiming Yu, Fanny Nina-Paravecino, David Kaeli, Qianqian Fang, "Scalable and massively parallel Monte Carlo photon transport simulations for heterogeneous computing platforms," J. Biomed. Opt. 23(1), 010504 (2018).

If you use MCX in your research, the author of this software would like you to cite the above papers in your related publications.


Powered by Habitat