SimulateMicroscopy.py

#!/usr/bin/env python
'''
Author: David Ladd

Brief: Convert finite element FCa field results to simulated confocal
       microscopy data to be processed by CaCLEAN.

Copyright 2019 David Ladd, University of Melbourne

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
'''

import inputs.utilities as util
import numpy as np
import math
import pandas as pd
import naturalneighbor
import skimage.io
import trimesh
import scipy
import scipy.io
from scipy import signal
import os
#import matplotlib.pyplot as plt

#=================================================================
# C l a s s e s
#=================================================================

class fieldInfo(object):
    'base class for info about fields'
    def __init__(self):
        self.region = ''
        self.group = ''
        self.numberOfFields = 0
        self.fieldNames = []
        self.numberOfFieldComponents = []

# =====================================
# C o n t r o l   P a n e l
# =====================================
inputsDir = "./inputs/"
outputsDir = "./simulatedMicroscopy_outputs/"
testCaCLEANOutputsDir = "./TestCaCLEAN_outputs/"
FE_modelDir = '../cardiaccalcium_finiteelement/'
# -------------------------------------
# Input nodal FE data info
# -------------------------------------
stopTime = 30.                             # finish time for the FEM simulation (ms)
startTime = 0.0                            # start time for the FEM simulation (ms)  
timeIncrement = 0.1                        # timestep increment (ms)
numberOfProcessors = 8                     # number of processors if run as an MPI job
species = [3]                              # Ca(1), F(2), FCa(3), CaM(4), CaMCa(5), ATP(6), ATPCa(7)
speciesLabel = ['FCa']
dependentFieldNumber = 2                   # The dependent field of interest
outputFrequency = 50                       # the output frequency of the FEM solver (or desired frequency to read the node data files at)
regSpace =  0.05375 #1075 #0.0215 0.05375  # regular grid spacing to interpolate to (in um)
outputSpacingStep = 4  
psfFile = 'psf_50x50x140_53nmPx_fwhm_410xy_1800z.tif'
roundTo = regSpace
initFCa = 2.08
padding = 0
SNR = 100.0

# -------------------------------------
# Identify FEM I/O files
# -------------------------------------
meshDir = FE_modelDir + 'input/'
nodeFile = 'Combined_8Sarc_1319kNodes_node.h5'  #'Combined_8Sarc_1436kNodes_node.h5'
FE_resultsDir = FE_modelDir + "output/"
surfaceMaskFile =  "Combined_8Sarc_503kNodes_surfaceFixed.stl"
boundaryTol = 2.1*regSpace

# -------------------------------------
# Identify RyR I/O files
# -------------------------------------
spacingType = "1umSpacing"
mitoModel = 'NoMito'
numSarc = 8
numRyrPerSarc = 51 #123
maxClustersPerSlice = 400
numberOfKNodes = 1319 #1436 #1319
ryrClusterCentersDir = inputsDir + 'ryr_locations/N' + str(numRyrPerSarc) + '/'
detectRyrTol = 0.2
detectRyrTolInc = 0.01
detectRyrTolMax = 1.5000001
# -------------------------------------

# Read in the node and element maps generated by tetgen
print('Checking for node file...')
if not os.path.exists(meshDir + nodeFile):
    print('Downloading node file to directory: ' + meshDir)
    util.download_file(nodeFile, meshDir)
store = pd.HDFStore(meshDir + nodeFile)
df = store['Node_Coordinates']
totalNumberOfNodes = df.shape[0]
nodeX = np.ascontiguousarray(df.iloc[:, 0].values)
nodeY = np.ascontiguousarray(df.iloc[:, 1].values)
nodeZ = np.ascontiguousarray(df.iloc[:, 2].values)
nodeGeom = np.asarray(df)
store.close()

# -------------------------------------
# Output regular data info
# -------------------------------------
xGridMax = (math.ceil(np.max(nodeX)/(roundTo*outputSpacingStep))  + padding)*(roundTo*outputSpacingStep)
xGridMin = (math.floor(np.min(nodeX)/(roundTo*outputSpacingStep)) - padding)*(roundTo*outputSpacingStep)
yGridMax = (math.ceil(np.max(nodeY)/(roundTo*outputSpacingStep))  + padding)*(roundTo*outputSpacingStep)
yGridMin = (math.floor(np.min(nodeY)/(roundTo*outputSpacingStep)) - padding)*(roundTo*outputSpacingStep)
zGridMax = (math.ceil(np.max(nodeZ)/(roundTo*outputSpacingStep))  + padding)*(roundTo*outputSpacingStep)
zGridMin = (math.floor(np.min(nodeZ)/(roundTo*outputSpacingStep)) - padding)*(roundTo*outputSpacingStep)

gridDimensions = [(int((xGridMax - xGridMin)/regSpace + roundTo/100.0)+1),
                  (int((yGridMax - yGridMin)/regSpace + roundTo/100.0)+1),
                  (int((zGridMax - zGridMin)/regSpace + roundTo/100.0)+1)]

xGrid = [xGridMin, xGridMax, gridDimensions[0]*1j]
yGrid = [yGridMin, yGridMax, gridDimensions[1]*1j]
zGrid = [zGridMin, zGridMax, gridDimensions[2]*1j]
grid_ranges = [xGrid, yGrid, zGrid]

x = np.linspace(xGrid[0], xGrid[1], gridDimensions[0])[0:gridDimensions[0]:outputSpacingStep]
y = np.linspace(yGrid[0], yGrid[1], gridDimensions[1])[0:gridDimensions[1]:outputSpacingStep]
z = np.linspace(zGrid[0], zGrid[1], gridDimensions[2])[0:gridDimensions[2]:outputSpacingStep]

print(x)
print(y)
print(z)
xLen = len(x)
yLen = len(y)
zLen = len(z)
totalNumberOfGridPoints = xLen*yLen*zLen
# Create a list of points from the regular grid
grid = np.zeros((totalNumberOfGridPoints, 3))
c = 0
for i in range(xLen):
    for j in range(yLen):
        for k in range(zLen):
            grid[c] = [x[i], y[j], z[k]]
            c += 1

# -------------------------------------
# Read in FEM nodal data
# -------------------------------------
# Get number of timesteps
totalSimTime = stopTime - startTime
numberOfTimesteps = 1
if totalSimTime > timeIncrement:
    numberOfTimesteps += int((totalSimTime/timeIncrement)/outputFrequency)

print('Number of timsteps: ' + str(numberOfTimesteps))

nn_interp = np.zeros((numberOfTimesteps, gridDimensions[0], gridDimensions[1], gridDimensions[2]))
interpDataFile = outputsDir + 'FCa_' + spacingType  + '_' + mitoModel + '_interp' + str(int(regSpace*1000)) + '_' + str(numberOfTimesteps) + 'timesteps.npz'
try:
    os.makedirs(outputsDir)
except OSError as e:
    if e.errno != 17:
        raise
# Check whether we have already read and stored the interpolated data
if os.path.isfile(interpDataFile):
    with open(interpDataFile):
        print('Reading interpolated data from: ' + interpDataFile)
        loaded = np.load(interpDataFile)
        nn_interp = loaded['nn_interp']
else:
    specData = np.zeros((numberOfTimesteps, totalNumberOfNodes))
    specGeomDataFile = outputsDir + 'FCa_' + spacingType  + '_' + mitoModel + '_NodalData_' +  str(numberOfTimesteps) + 'timesteps.npz'
    # Check whether we have already read and stored the nodal data
    if os.path.isfile(specGeomDataFile):
        with open(specGeomDataFile):
            print('Reading FEM data from: ' + specGeomDataFile)
            loaded = np.load(specGeomDataFile)
            specData = loaded['spec']
            nodeGeom = loaded['geom']
    else:
        # Collect geometry and species data from OpenCMISS .exnode files
        fields = fieldInfo()
        for timestep in range(numberOfTimesteps):
            nodeData = np.zeros((totalNumberOfNodes,4))
            outputStep = timestep * outputFrequency
            s = 0
            for speciesNumber in species:
                filenameRoot = FE_resultsDir + 'TIME_STEP_SPEC_'+ str(speciesNumber) + '.part'
                for proc in range(numberOfProcessors):
                    filename = filenameRoot + str(proc).zfill(2) + '.' + str(outputStep).zfill(3) + '.exnode'
                    #filename += str(proc).zfill(2) + '.300.exnode'
                    importNodeData = np.zeros((totalNumberOfNodes,4))
                    print('Reading: ' + filename)
                    util.readExnode(filename,fields,importNodeData,totalNumberOfNodes,dependentFieldNumber)
                    nodeData += importNodeData
                specData[timestep] = nodeData[:,3]
                s+=1
        print('Saving nodal data to: ' + specGeomDataFile)
        # Save nodal data
        np.savez_compressed(specGeomDataFile, spec=specData, geom=nodeGeom)

    for timestep in range(numberOfTimesteps):
        print('   Performing discrete Sibson interpolation, Time (ms): ' + str(timestep*timeIncrement*outputFrequency))
        specData[timestep, specData[timestep] < 0.01] = initFCa
        nn_interp[timestep] = naturalneighbor.griddata(nodeGeom, specData[timestep], grid_ranges)

    # Save interpolated data
    np.savez_compressed(interpDataFile, nn_interp=nn_interp)
    print('Saving interpolated data to: ' + interpDataFile)

# -------------------------------------
# Convolve image against selected PSF
# -------------------------------------
convolve = True
if convolve:
    # download the psf file if needed
    print('Checking for PSF file...')
    if not os.path.exists(inputsDir + psfFile):
        print('Downloading PSF file to directory: ' + inputsDir)
        util.download_file(psfFile, inputsDir)
    convolved = np.zeros_like(nn_interp)
    psfOrig = skimage.io.imread(inputsDir + psfFile)
    # in our case, the y-axis is the z depth for the simulated microscopy
    psf = np.moveaxis(psfOrig, 2, 1)
    magPsf = sum(sum(sum(psf)))
    for timestep in range(numberOfTimesteps):
        print('Convolving interpolated image with PSF, timestep: ' + str(timestep))
        convolved[timestep] = signal.convolve(nn_interp[timestep], psf, mode='same') # / magPsf
        #convolved[timestep] = ndimage.convolve(nn_interp[timestep], psf, mode='constant') / magPsf

# -------------------------------------
# Select 2D slice and export
# CaCLEAN fields to a MatLab .mat file
# -------------------------------------
ryrCentersOrig = np.zeros((numRyrPerSarc*numSarc, 3))
zOffset = 1.0
for s in range(numSarc):
    sarcRyr = np.loadtxt(ryrClusterCentersDir+'simPP'+str(s+1)+'.txt')
    sarcRyr[:, 2] += zOffset
    ryrCentersOrig[s*numRyrPerSarc:(s+1)*numRyrPerSarc, :] = sarcRyr
    zOffset += 2.0

# x/y seems to be flipped in the ryrCenters file?
ryrCenters = np.zeros_like(ryrCentersOrig)
ryrCenters[:, 0] = ryrCentersOrig[:, 1]
ryrCenters[:, 1] = ryrCentersOrig[:, 0]
ryrCenters[:, 2] = ryrCentersOrig[:, 2]
print(ryrCenters)


# tolerances to check for clusters
tolerances = np.arange(detectRyrTolInc, detectRyrTolMax, detectRyrTolInc)
numTol = len(tolerances)
# Choose slices 
multiSliceData = []
multiBgrData = []
multiMask = []

sliceList = list(np.arange(5, yLen-4, 2, dtype="int"))
numSlices = len(sliceList)

print('Checking for surface mask file...')
if not os.path.exists(inputsDir + surfaceMaskFile):
    print('Downloading surface mask file to directory: ' + inputsDir)
    util.download_file(surfaceMaskFile, inputsDir)
# Load outer surface mask
print('surface mask used: ' + inputsDir + surfaceMaskFile)
mesh = trimesh.load(inputsDir + surfaceMaskFile)

# For each slice, construct a list of RyRs with centers within
# a range of distance tolerances (admissible windows) and a 
# mask file to indicate which pixels represent the intracellular space.
# Downsample the convolved data to match typical microscopy resolution
# (215 nm xy, 5 ms in time by default). Also create a background image
# set to the initial FCa conditions. Add light noise (SNR=100) to the 
# simulated microscopy data and the background image.
multiRyrClusterCenters = np.full((numSlices, numTol, maxClustersPerSlice, 2), np.nan)
for s in range(numSlices):
    ySlice = sliceList[s]
    yLocation = y[ySlice]
    numDetect = 0
    ryrSliceCenters = []
    for tol in range(numTol):
        detectRyrTol = tolerances[tol]
        centersTemp = ryrCenters[np.where(np.isclose(ryrCenters[:, 1], yLocation, atol=detectRyrTol))]
        numDetect = len(centersTemp)
        scaled = (np.delete(centersTemp, 1, axis = 1) / (regSpace*outputSpacingStep)
                  + padding*outputSpacingStep + 1.)  # pixel space conversions 
        multiRyrClusterCenters[s, tol,:numDetect] = scaled
    # mask out the points outside of the mesh
    print('checking if pixels in mesh...')
    ySliceGrid = grid[np.where(np.isclose(grid[:, 1], yLocation, atol=detectRyrTolInc))]
    mask = mesh.contains(ySliceGrid)
    print('Calculating distance from triangulated surfaces')
    (closest_points, distances, triangle_id) = mesh.nearest.on_surface(ySliceGrid)
    maskOut = np.full((xLen, zLen), True, dtype=bool)
    c = 0
    for i in range(xLen):
        for k in range(zLen):
            if not mask[c]:
                maskOut[i, k] = False
            if distances[c] < boundaryTol:
                maskOut[i, k] = False
            c += 1
    mask_out_slice = maskOut
    # Downsample the convolved data in x, y, and t
    convolvedReduced = convolved[:, 0:gridDimensions[0]:outputSpacingStep,
                                 0:gridDimensions[1]:outputSpacingStep,
                                 0:gridDimensions[2]:outputSpacingStep]
    sliceData = np.moveaxis(np.squeeze(convolvedReduced[:, :, ySlice, :]), 0, 2)
    # Add light noise to the slice data and a background image
    noisySliceData = np.zeros_like(sliceData)
    bgr = np.zeros_like(sliceData)
    stdDev = np.mean(sliceData[:, :, 0]) / SNR
    for timestep in range(numberOfTimesteps):
        noise1 = np.random.normal(0, stdDev, [xLen, zLen])
        noise2 = np.random.normal(0, stdDev, [xLen, zLen])
        noisySliceData[:, :, timestep] = sliceData[:, :, timestep] + noise1
        bgr[:, :, timestep] = sliceData[:, :, 0] + noise2
    multiSliceData.append(noisySliceData)
    multiBgrData.append(bgr)
    multiMask.append(mask_out_slice)

xyt = [x[1]-x[0], z[1]-z[0], timeIncrement*outputFrequency]
outString = outputsDir + "simulatedMicroscopyResults_interp" + str(int(regSpace*1000)) + "_SNR" + str(int(SNR)) + '_' + spacingType + '_' + mitoModel + '_' + str(xLen) + 'x' + str(zLen) + 'x' + str(numberOfTimesteps)

outputFile = outString + '.mat'
print('Writing MatLab file: ' + outputFile)
scipy.io.savemat(outputFile, mdict={'MultiMask': multiMask,
                                    'MultiRyrClusterCenters': multiRyrClusterCenters,
                                    'RyrTolerances': tolerances,
                                    'MultiBgr': multiBgrData,
                                    'MultiIdenoised': multiSliceData,
                                    'xyt_dim': xyt})

# Make outputs dir for TestCaCLEAN results (easier to do here than in Matlab)
try:
    os.makedirs(testCaCLEANOutputsDir)
except OSError as e:
    if e.errno != 17:
        raise


# # timestep = 4
# # temp_nn_interp= naturalneighbor.griddata(nodeGeom, specData[timestep], grid_ranges)
# fig = plt.figure()
# ax = fig.add_subplot(111)
# #im = ax.imshow(temp_nn_interp[:,50,:])
# #im = ax.imshow(nn_interp[4,:,50,:])
# im = ax.imshow(multiSliceData[5][:,:,4])
# plt.colorbar(im)
# plt.show()
# plt.close(fig)