File size: 6,317 Bytes
12d2e9e |
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 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 |
{
"cells": [
{
"cell_type": "markdown",
"id": "75884451-1ef9-49ab-8561-4eebe036cb3f",
"metadata": {},
"source": [
"# Supplementary Vignette 2\n",
"\n",
"## Example workflow for H&E images\n",
"\n",
"Here we demonstrate a typical workflow for preprocessing of H&E images, consisting of the following steps:\n",
"\n",
"1. Loading the raw image\n",
"2. Defining a simple preprocessing pipeline for tissue detection\n",
"3. Creating a PyTorch DataLoader for interfacing with any downstream machine learning model\n",
"\n",
"The image used in this example is publicly avilalable for download: http://openslide.cs.cmu.edu/download/openslide-testdata/Aperio/\n",
"\n",
"**a. Load the image**"
]
},
{
"cell_type": "code",
"execution_count": 3,
"id": "4fe03d4f-1e07-4afe-9ed2-dc027977b782",
"metadata": {},
"outputs": [],
"source": [
"from pathml.core import SlideData, types\n",
"\n",
"# load the image\n",
"wsi = SlideData(\"../../data/CMU-1.svs\", name=\"example\", slide_type=types.HE)"
]
},
{
"cell_type": "markdown",
"id": "cd6e65ee-13b7-486b-85bf-b011a851f255",
"metadata": {
"jp-MarkdownHeadingCollapsed": true,
"tags": []
},
"source": [
"**b. Define a preprocessing pipeline**\n",
"\n",
"Pipelines are created by composing a sequence of modular transformations; in this example we apply a blur to reduce noise in the image followed by tissue detection"
]
},
{
"cell_type": "code",
"execution_count": 5,
"id": "b94114e2-4384-4f6e-b0cd-14e4ebcb205c",
"metadata": {},
"outputs": [],
"source": [
"from pathml.preprocessing import Pipeline, BoxBlur, TissueDetectionHE\n",
"\n",
"pipeline = Pipeline(\n",
" [\n",
" BoxBlur(kernel_size=15),\n",
" TissueDetectionHE(\n",
" mask_name=\"tissue\",\n",
" min_region_size=500,\n",
" threshold=30,\n",
" outer_contours_only=True,\n",
" ),\n",
" ]\n",
")"
]
},
{
"cell_type": "markdown",
"id": "4c5177e6-d7e5-4218-a150-362c05e35988",
"metadata": {},
"source": [
"**c. Run preprocessing**\n",
"\n",
"Now that we have constructed our pipeline, we are ready to run it on our WSI.\n",
"PathML supports distributed computing, speeding up processing by running tiles in parallel among many workers rather than processing each tile sequentially on a single worker. This is supported by [Dask.distributed](https://distributed.dask.org/en/latest/index.html) on the backend, and is highly scalable for very large datasets. \n",
"\n",
"The first step is to create a `Client` object. In this case, we will use a simple cluster running locally; however, Dask supports other setups including Kubernetes, SLURM, etc. See the [PathML documentation](https://pathml.readthedocs.io/en/latest/running_pipelines.html#distributed-processing) for more information."
]
},
{
"cell_type": "code",
"execution_count": 6,
"id": "c7eb71bf-2189-41ee-9e51-194f9c655ab2",
"metadata": {},
"outputs": [],
"source": [
"from dask.distributed import Client, LocalCluster\n",
"\n",
"cluster = LocalCluster(n_workers=6)\n",
"client = Client(cluster)\n",
"\n",
"wsi.run(pipeline, distributed=True, client=client);"
]
},
{
"cell_type": "code",
"execution_count": 7,
"id": "46aaf3f2-7ccf-4bcd-be64-888dafc6c096",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Total number of tiles extracted: 150\n"
]
}
],
"source": [
"print(f\"Total number of tiles extracted: {len(wsi.tiles)}\")"
]
},
{
"cell_type": "markdown",
"id": "4959af73-db3e-4c38-a7f3-1e6f854e50ac",
"metadata": {},
"source": [
"**e. Save results to disk**\n",
"\n",
"The resulting preprocessed data is written to disk, leveraging the HDF5 data specification optimized for efficiently manipulating larger-than-memory data."
]
},
{
"cell_type": "code",
"execution_count": 8,
"id": "0abf4264-13f5-4264-b8f3-93bf97589b77",
"metadata": {},
"outputs": [],
"source": [
"wsi.write(\"./data/CMU-1-preprocessed.h5path\")"
]
},
{
"cell_type": "markdown",
"id": "ad585896-b456-40fb-821e-dc45894519af",
"metadata": {},
"source": [
"**f. Create PyTorch DataLoader**\n",
"\n",
"The `DataLoader` provides an interface with any machine learning model built on the PyTorch ecosystem"
]
},
{
"cell_type": "code",
"execution_count": 9,
"id": "324b6286-c3d7-49ad-8a86-4f70998c4358",
"metadata": {},
"outputs": [],
"source": [
"from pathml.ml import TileDataset\n",
"from torch.utils.data import DataLoader\n",
"\n",
"dataset = TileDataset(\"./data/CMU-1-preprocessed.h5path\")\n",
"dataloader = DataLoader(dataset, batch_size=16, num_workers=4)"
]
},
{
"cell_type": "markdown",
"id": "f17c90a3-feed-48a4-97df-35c7ecd9a7e1",
"metadata": {
"tags": []
},
"source": [
"### Summary\n",
"\n",
"Here we demonstrate a complete `PathML` workflow for analyzing brightfield images:\n",
"\n",
"1. Loading the raw image\n",
"2. Define a simple preprocessing pipeline for tissue detection\n",
"3. Create a PyTorch DataLoader for interfacing with any downstream machine learning model\n",
"\n",
"Full documentation of the `PathML` API is available at https://pathml.org. \n",
"\n",
"Full code for this vignette is available at https://github.com/Dana-Farber-AIOS/pathml/tree/master/examples/manuscript_vignettes_stable"
]
}
],
"metadata": {
"environment": {
"kernel": "pathml",
"name": ".m115",
"type": "gcloud",
"uri": "gcr.io/deeplearning-platform-release/:m115"
},
"kernelspec": {
"display_name": "pathml",
"language": "python",
"name": "pathml"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.9.18"
}
},
"nbformat": 4,
"nbformat_minor": 5
}
|