Evaluate nuclei segmentation

from PIL import Image
import sys
import pyvips
import numpy as np
from openslide import OpenSlide
import tifffile

Image.MAX_IMAGE_PIXELS = None
img_path = sys.argv[1]
img = np.array(Image.open(img_path))

if img.shape[0] == 3:
    img = np.transpose(img, axes=(1, 2, 0))
elif img.shape[2] == 4: # RGBA to RGB
    img = img[:,:,:3]

pyvips_img = pyvips.Image.new_from_array(img)
save_path=sys.argv[2]

pixel_size = float(sys.argv[3]) # 1 pixel = ? mu m

pyvips_img.tiffsave(
    save_path, 
    bigtiff=False, 
    pyramid=True, 
    tile=True, 
    tile_width=256, 
    tile_height=256, 
    compression='jpeg', 
    resunit=pyvips.enums.ForeignTiffResunit.CM,
    xres=1. / (pixel_size * 1e-4),
    yres=1. / (pixel_size * 1e-4))

import json
from pathlib import Path
from typing import List, Tuple, Union

import argparse
import inspect
import os
import sys

import numpy as np
import torch
import torch.nn.functional as F
import tqdm
import yaml
from matplotlib import pyplot as plt
from PIL import Image, ImageDraw
from skimage.color import rgba2rgb
from sklearn.metrics import accuracy_score
from torch.utils.data import DataLoader
from torchmetrics.functional import dice
from torchmetrics.functional.classification import binary_jaccard_index
from torchvision import transforms

import pandas as pd
import json


with open('./preprocessing/Human_Breast_Andersson_10142021_ST_H1/cell_detection/cell_detection.json') as json_data:
    data = json.load(json_data)

# load image
image_full = Image.open('./Human_Breast_Andersson_10142021_ST_H1.tif')
image_full = image_full.convert('RGB')

# get the cell types of the cells and plot the contour of the cells
with open('./preprocessing0/Human_Breast_Andersson_10142021_ST_H1/cell_detection/cells.geojson', encoding='utf-8') as f:
    gj = json.load(f)

gj = pd.DataFrame(gj) 

# plot the cells with cell type
fig, ax = plt.subplots(1, 1, figsize=(100, 100))
ax.imshow(image_full)

for i in range(5):
    contour = gj.iloc[i,]['geometry']['coordinates']
    color = gj.iloc[i,]['properties']['classification']['color']
    color = (color[0]/255, color[1]/255, color[2]/255)
    cell_type = gj.iloc[i,]['properties']['classification']['name']
    for j in range(len(contour)):
        contour_cell = np.array(contour[j])[0]
        ax.plot(contour_cell[:,0], contour_cell[:,1], color=color,label = cell_type)

fig.savefig('.segmentation/cell.png')


# get the cell types of the cells
with open('./preprocessing0/Human_Breast_Andersson_10142021_ST_H1/cell_detection/cell_detection.geojson', encoding='utf-8') as f:
    gj = json.load(f)

gj = pd.DataFrame(gj)
data_organized = pd.DataFrame()

for i in range(5):
    temp = pd.DataFrame(gj.iloc[i,]['geometry']['coordinates'])
    temp['cell_type'] = gj.iloc[i,]['properties']['classification']['name']
    data_organized = pd.concat([data_organized,temp])

# organize data
data_organized.columns = ['x','y','cell_type']

coord = pd.read_csv('./ST/coord/Human_Breast_Andersson_10142021_ST_H1_coord.csv', index_col=0)

# only keep the spot within the Cell-ViT inferred region
coord_keep = coord[coord['xaxis']-coord['r'] >= data_organized.x.min()]
coord_keep = coord_keep[coord_keep['xaxis']+coord_keep['r'] <= data_organized.x.max()]
coord_keep = coord_keep[coord_keep['yaxis']-coord_keep['r'] >= data_organized.y.min()]
coord_keep = coord_keep[coord_keep['yaxis']+coord_keep['r'] <= data_organized.y.max()]

index_celltype = ['Neoplastic', 'Inflammatory', 'Connective', 'Dead', 'Epithelial']
inferred = np.zeros((coord_keep.shape[0],5))

# aggregate cell number in each spot
for i in range(coord_keep.shape[0]):
    xaxis = coord_keep.iloc[i,]['xaxis']
    yaxis = coord_keep.iloc[i,]['yaxis']
    radius = coord_keep.iloc[i,]['r']
    cell_in = data_organized[(data_organized['x']-xaxis)**2 + (data_organized['y']-yaxis)**2 <= radius**2]
    if cell_in.shape[0] == 0:
        continue
    else:
        for j in range(5):
            inferred[i,j] = cell_in[cell_in['cell_type'] == index_celltype[j]].shape[0]
    print(i)


coord_keep['Neoplastic'] = inferred[:,0]
coord_keep['Inflammatory'] = inferred[:,1]
coord_keep['Connective'] = inferred[:,2]
coord_keep['Dead'] = inferred[:,3]
coord_keep['Epithelial'] = inferred[:,4]

coord_keep.to_csv('.segmentation/cell_type_256.csv')

# Then repeat the same code for SAM based model.