Skip to content

Commit

Permalink
Add new models
Browse files Browse the repository at this point in the history
  • Loading branch information
@Mmasoud1
Mmasoud1 committed Mar 27, 2024
1 parent 27917be commit 2b061eb
Show file tree
Hide file tree
Showing 7 changed files with 1,105 additions and 24 deletions.
35 changes: 34 additions & 1 deletion index.html
View file
Original file line number Diff line number Diff line change
Expand Up @@ -991,6 +991,38 @@
},


{ cols: [
{ view: "label", label: "Auto Threshold Input",
align: "left"
},

{
view: "select",
id: "autoThresholdStatus",
value: 1,
options: ["0", "0.1", "0.2"]
},

{
view: "label",
label: '<img src="css/svg/info-circle-solid.svg"/>',
css: {"color":"black !important", "font-weight": "bold", "cursor": "pointer"},
width: 25,
height: 25,
click: function() {
let info = "For speed-up the inference with limited browser memory, auto thresholding the brain from noisy voxels around it before cropping it and feeding the result to the inference model can lower memory use.";
$$("modelTooltip").show();
document.getElementById("tooltipDiv").innerHTML =
"<i style='font-size:1.4vw' class='fa fa-info-circle' ></i> <font style='font-size:0.77vw' >&nbsp&nbsp"+ info +" </font>"
}
}

]
},





{ cols: [
{ view: "label", label: "Crop Input",
Expand Down Expand Up @@ -1280,6 +1312,7 @@
enableTranspose:($$("transposeStatus").getValue() === 'true'),
enableCrop: ($$("cropStatus").getValue() === 'true'),
cropPadding: parseInt($$("cropPadStatus").getValue()),
autoThreshold: parseInt($$("autoThresholdStatus").getValue()),
enableQuantileNorm: ($$("quantileNormStatus").getValue() === 'true'),
filterOutWithPreMask: ($$("filterByMaskStatus").getValue() === 'true'),
enableSeqConv: ($$("seqConvStatus").getValue() === 'true'),
Expand Down Expand Up @@ -1320,7 +1353,7 @@
webix.ui({
view:"window",
id: "modelBrowsingWindow",
height:750,
height:770,
width:400,
head:{
view:"toolbar", css: "toolbarclass", elements:[
Expand Down
181 changes: 158 additions & 23 deletions js/brainchop/mainMeshNetFunctions.js
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1257,6 +1257,123 @@ rgbToHex = (rgbObj) => {



/**
* Get MRI after threshold noisy voxels around the brain for better cropping later
* @since 3.0.0
* @param {tf.Tensor} tensor - Tensor3d, e.g. Tensor3d of all MRI volume data
* @param {number} percentage - Threshold percentage is just a number between 0 and 1
* @returns {tf.Tensor}
*
*/


applyMriThreshold = async(tensor, percentage) => {
// Perform asynchronous operations outside of tf.tidy
const maxTensor = tensor.max();
const thresholdTensor = maxTensor.mul(percentage);
const threshold = await thresholdTensor.data(); // Extracts the threshold value

// Dispose tensors not needed anymore
maxTensor.dispose();
thresholdTensor.dispose();

// Use tf.tidy for synchronous operations
const denoisedMriData = tf.tidy(() => {
const dataForProcessing = tensor.clone();

// Thresholding (assuming background has very low values compared to the head)
const mask = dataForProcessing.greater(threshold[0]);
const denoisedMriData = dataForProcessing.mul(mask);

// No need to manually dispose dataForProcessing and mask, as tf.tidy() will dispose them auto.
return denoisedMriData;
});

return denoisedMriData;
}




/**
* Get MRI copping coordinates after threshold
* @since 3.0.0
* @param {tf.Tensor} tensor - Tensor3d, e.g. Tensor3d of all MRI volume data
* @param {number} percentage - Threshold percentage is just a number between 0 and 1
* @returns {Array}
* @example
*
* arr = Array.from({length: 27}, (x, i) => i/10)
* => Array(27) [ 0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, … , 2.6]
*
* cropped = await cropTensorWithThreshold ( tf.tensor( Array.from({length: 27}, (x, i) => i/10) , [3, 3, 3]), 0.2 )
*
* => Array [ {…}, {…} ]
*
* cropped[0].print()
*
*/


cropTensorWithThreshold = async(tensor, percentage) => {

// Find the maximum value of the tensor
const maxTensor = tensor.max();

// Multiply the maximum value by the thresholdRatio to get % of the max
const thresholdTensor = maxTensor.mul(percentage);

// Extract the value from the tensor
const threshold = await thresholdTensor.data();

const dataForProcessing = tensor.clone();

// Thresholding (assuming background has very low values compared to the head)
const mask = dataForProcessing.greater(threshold[0]);
const masked_data = dataForProcessing.mul(mask);

// Find the bounding box around the head (non-zero region) in the filtered data
const indices = await tf.whereAsync(masked_data.greater(0));
dataForProcessing.dispose();
mask.dispose();
masked_data.dispose();

// Extract z, y, x coordinates from the indices
const zs = indices.slice([0, 0], [indices.shape[0], 1]); // z coordinates
const ys = indices.slice([0, 1], [indices.shape[0], 1]); // y coordinates
const xs = indices.slice([0, 2], [indices.shape[0], 1]); // x coordinates

// Compute min and max indices for each dimension
const min_z = zs.min().arraySync();
const max_z = zs.max().arraySync();
const min_y = ys.min().arraySync();
const max_y = ys.max().arraySync();
const min_x = xs.min().arraySync();
const max_x = xs.max().arraySync();

// Crop the original tensor using the bounding box from the filtered data
const cropped_tensor = tensor.slice([min_z, min_y, min_x], [max_z - min_z + 1, max_y - min_y + 1, max_x - min_x + 1]);

// Clean up tensors to free memory
indices.dispose();
zs.dispose();
ys.dispose();
xs.dispose();

// Return the cropped tensor along with the min and max indices
return [cropped_tensor, {
minZ: min_z,
maxZ: max_z,
minY: min_y,
maxY: max_y,
minX: min_x,
maxX: max_x
}];

}



/**
* load pre-trained model from local drive
*
Expand Down Expand Up @@ -4751,16 +4868,28 @@ function convByOutputChannelAndInputSlicing(input, filter, biases, stride, pad,

let mask_3d;

if(pipeline1_out == null) {
// binarize original image
mask_3d = slices_3d.greater([0]).asType('bool');
if(pipeline1_out == null) { // preModel is null

// Check if thresholding the MRI to remove noisy voxels for better cropping is needed.
let autoThresholdValue = inferenceModelsList[$$("selectModel").getValue() - 1]["autoThreshold"];

if( (autoThresholdValue > 0) && (autoThresholdValue <= 1) ) {

// Filtered MRI from noisy voxel below autoThresholdValue
slices_3d = await applyMriThreshold(slices_3d, autoThresholdValue);
} else {
console.log("No valid crop threshold value");
}

// binarize original image
mask_3d = slices_3d.greater([0]).asType('bool');

} else {

mask_3d = pipeline1_out.greater([0]).asType('bool');
//-- pipeline1_out.dispose();
mask_3d = pipeline1_out.greater([0]).asType('bool');
//-- pipeline1_out.dispose();

}
}

console.log(" mask_3d shape : ", mask_3d.shape);

Expand Down Expand Up @@ -5975,9 +6104,22 @@ get3dObjectBoundingVolume = async(slices_3d) => {
slices_3d = minMaxNormalizeVolumeData(slices_3d);
}


let mask_3d;

if(pipeline1_out == null) {
if(pipeline1_out == null) { // preModel is null

// Check if thresholding the MRI to remove noisy voxels for better cropping is needed.
let autoThresholdValue = inferenceModelsList[$$("selectModel").getValue() - 1]["autoThreshold"];

if( (autoThresholdValue > 0) && (autoThresholdValue <= 1) ) {

// Filtered MRI from noisy voxel below autoThresholdValue
slices_3d = await applyMriThreshold(slices_3d, autoThresholdValue);
} else {
console.log("No valid crop threshold value");
}

// binarize original image
mask_3d = slices_3d.greater([0]).asType('bool');

Expand Down Expand Up @@ -6448,26 +6590,21 @@ checkInferenceModelList = () => {

statData["No_SubVolumes"] = 1;

// check for pre model to load e.g. Brain_Extraction or Brain_Masking
let preModel = null;
let modelEntry = inferenceModelsList[$$("selectModel").getValue() - 1];
console.log("modelEntry ", modelEntry)

//-- If pre-model is not null then slices_3d mask will be generated..
//-- The mask is needed to remove the skull and set noise in background to 0, and get the brain bounding volume properly
let slices_3d_mask = null;


// load pre-model for inference first, can be null if no pre-model such as GWM models
if(modelEntry["preModelId"]) {

preModel = load_model(inferenceModelsList[ modelEntry["preModelId"] - 1]['path'] );
let preModel = load_model(inferenceModelsList[ modelEntry["preModelId"] - 1]['path'] );
let transpose = inferenceModelsList[ modelEntry["preModelId"] - 1]["enableTranspose"];

let quantileNorm = inferenceModelsList[ modelEntry["preModelId"] - 1]["enableQuantileNorm"];

let preModel_slices_3d = null;

//-- If pre-model is not null then slices_3d mask will be generated..
//-- The mask is needed to remove the skull and set noise in background to 0, and get the brain bounding volume properly
let slices_3d_mask = null;

if(quantileNorm) {
// Quantile normalize function needs specific models to be used
console.log("preModel Quantile normalization enabled");
Expand All @@ -6479,20 +6616,19 @@ checkInferenceModelList = () => {
}




//-- Transpose MRI data to be match pytorch/keras input output
//-- Check if pre-model needs transpose..
if(transpose) {
preModel_slices_3d = preModel_slices_3d.transpose()

preModel_slices_3d = preModel_slices_3d.transpose();
console.log("Input transposed for pre-model");

} else {
console.log("Transpose not enabled for pre-model");
}

statData["Brainchop_Ver"] = "PreModel_FV" ; // e.g. "PreModel_FV"


preModel.then(function (res) {

try {
Expand Down Expand Up @@ -6552,7 +6688,7 @@ checkInferenceModelList = () => {
statData["Model"] = inferenceModelsList[ modelEntry["preModelId"] - 1]["modelName"];
statData["Extra_Info"] = inferenceModelsList[$$("selectModel").getValue() - 1]["modelName"];


// maxLabelPredicted in whole volume of the brain
let maxLabelPredicted = 0;
let delay = inferenceModelsList[ modelEntry["preModelId"] - 1]["inferenceDelay"];
Expand All @@ -6566,7 +6702,6 @@ checkInferenceModelList = () => {

//Dispose the volume
tf.dispose(preModel_slices_3d);


let timer = window.setInterval(function() {

Expand Down
Loading

0 comments on commit 2b061eb

Please sign in to comment.