Integration with slide level-labels

Integration with slide level-labels#

In this tutorial we will demonstrate how to integrate whole slide images (WSIs) with slide-level labels and derive quantitative scores for each slide via top-K scoring.

We will also demonstrate how to run tasks in a distributed fashion using dask.

For this, we will be using a pre-processed dataset of artery tissue from GTEx, whihc contains healthy and calcified samples.

import pooch
import pandas as pd

table = pooch.retrieve(
    url="https://lazyslide.blob.core.windows.net/lazyslide-data/GTEx_artery_dataset.csv.gz",
    fname="GTEx_artery_dataset.csv.gz",
    path=".",
    known_hash=None,
)

dataset = pd.read_csv(table)
dataset.head()
Tissue Sample Id Sex Age Bracket Pathology Categories
0 GTEX-111YS-2226 male 60-69 calcification
1 GTEX-11GSP-2926 female 60-69 calcification
2 GTEX-11LCK-1426 male 30-39 clean_specimens
3 GTEX-11ONC-2726 male 60-69 calcification
4 GTEX-12126-0726 male 20-29 clean_specimens

Distributed processing with dask#

When running GPU-intensive works like feature extraction for multiple WSIs, we recommanded to run one task on one GPU every time. To accelarate the processing speed, either distributed across multiple GPU cards or multiple machines.

Here, we showcase how to parallel the jobs with dask on a SLURM cluster.

For different hardward availabilities, alternatives are:

  1. dask-jobqueue: For PBS, Slurm, MOAB, SGE, LSF, and HTCondor.

  2. coiled: AWS, GCP, Azure etc.

  3. dask-cuda: If you have multiple GPU cards locally.

from dask_jobqueue import SLURMCluster
from dask.distributed import Client

cluster = SLURMCluster(
    queue="gpu",
    cores=8,
    processes=1,
    memory="20 GB",
    job_extra_directives=["-q gpu", "--gres=gpu:h100pcie:1", "--time=02:00:00"],
    worker_extra_args=["--resources GPU=1"],
    log_directory="./dask-logs",
)

client = Client(cluster)
cluster.scale(3)

client

Client

Client-4214ed8a-0f9e-11f0-88a8-9199a2d1e4fe

Connection method: Cluster object Cluster type: dask_jobqueue.SLURMCluster
Dashboard: http://10.110.81.4:8787/status

Cluster Info

Let’s define the task that we want to do for each slide.

from wsidata import open_wsi
import lazyslide as zs
import numpy as np

terms = ["calcification", "atherosclerosis"]


def wsi_feature_extraction(slide):
    s = pooch.retrieve(
        url=f"https://brd.nci.nih.gov/brd/imagedownload/{slide}",
        fname=f"{slide}.svs",
        path="gtex_artery_slides",
        known_hash=None,
    )
    wsi = open_wsi(s)
    zs.pp.find_tissues(wsi)
    zs.pp.tile_tissues(wsi, 256, mpp=0.5, background_fraction=0.5)

    # uni2 features
    zs.tl.feature_extraction(wsi, "uni2")
    zs.tl.feature_aggregation(wsi, "uni2")

    # conch feature
    zs.tl.feature_extraction(wsi, "conch", pbar=False)
    embed = zs.tl.text_embedding(terms)
    zs.tl.text_image_similarity(wsi, embed, "conch")
    wsi.write()

    # score the slide with Top-K max pooling
    scores = zs.metrics.topk_scores(wsi["conch_tiles_text_similarity"], k=100)

    return scores

futures = [
    client.submit(wsi_feature_extraction, slide, resources={"GPU": 1})
    for slide in dataset["Tissue Sample Id"]
]

If you want to monitor the process, you can either go to the dask dashboard or use a simple progress bar

from dask.distributed import as_completed
from tqdm.auto import tqdm

for _ in tqdm(as_completed(futures), total=len(futures)):
    pass

slide_scores = pd.DataFrame(
    np.vstack([f.result() for f in futures]),
    columns=terms,
    index=dataset["Tissue Sample Id"],
)
slide_scores.to_csv("slide_scores.csv")

Examination of Top-K scoring by text query#

import seaborn as sns

sns.boxplot(
    data=slide_scores.join(dataset.set_index("Tissue Sample Id")),
    x="Pathology Categories",
    y="calcification",
)

<Axes: xlabel='Pathology Categories', ylabel='calcification'>
../_images/2cf3bde38b6b84ffc5a2a90f78fda344ad6be9c0f6a77bb34801c2644ebce2f1.png 
client.shutdown()

Slide aggregation#

from wsidata import agg_wsi

dataset["slide"] = [f"gtex_artery_slides/{s}.svs" for s in dataset["Tissue Sample Id"]]
agg_data = agg_wsi(dataset, "uni2", wsi_col="slide", agg_key="agg_slide")
agg_data.obs = agg_data.obs.merge(slide_scores, on="Tissue Sample Id").set_index(
    "Tissue Sample Id"
)
agg_data

AnnData object with n_obs × n_vars = 45 × 1536
    obs: 'Sex', 'Age Bracket', 'Pathology Categories', 'slide', 'calcification', 'atherosclerosis'

agg_data.write_h5ad("agg_uni2_features.h5ad")