It's alive

[mzur]

As it happens Quimbi still seems to run fine in a current browser. I'll see if I can reproduce Jan's input file format using the msiproc HDF5 some time. Then people might actually use this tool.

[mzur]

Works alright with one of Annika's full datasets (360x269x20000) and in realtime on my laptop.

Here is the draft of my script to convert HDF5 to the Quimbi format. The matrix box seems to get lost and the overlay image is not working yet:

# Parses a HDF5 MSI image to the Quimbi format
import sys
import os
import pandas as pd
import numpy as np
from PIL import Image

# in_path = 'data/20180322_AF_DHB_RP_mus_K_109_1_converted/20180322_AF_DHB_RP_mus_K_109_1_processed_norm.h5'
in_path = sys.argv[1]
out_path = sys.argv[2]

file = pd.HDFStore(in_path)
key = file.keys()[0]
data = file.get(key)

coordinates = np.array(data.index.tolist())[:, 0:2]
minimum = coordinates.min(axis=0)
maximum = coordinates.max(axis=0)
dimensions = maximum - minimum + 1

dataset_name = key.strip('/')
# Dataset information
print('{},data/{}/,.png,{},{},{}'.format(dataset_name, dataset_name, data.shape[1], dimensions[0], dimensions[1]))
# Preprocessing information
print('unknown')
# Brightfield overlay information
print('TODO')

global_max = data.max().max()
global_min = data.min().min()

chunk_size = 4
current_channel_index = 0
last_channel_index = data.shape[1]
image = np.zeros((dimensions[0], dimensions[1], 4), dtype=np.float32)
filename = []

for channel in data:
   image_channel_index = current_channel_index % chunk_size
   for coords, pixel in zip(coordinates, data[channel]):
      image[coords[0] - minimum[0], coords[1] - minimum[1], image_channel_index] = pixel

   current_channel_index += 1
   filename.append(str(channel))

   if image_channel_index == 3 or current_channel_index == last_channel_index:
      png_image = (image - global_min) / global_max * 255
      name = '{}/{}.png'.format(out_path, '-'.join(filename))
      Image.fromarray(png_image.astype(np.uint8)).save(name)
      print('-'.join(filename))
      filename = []

file.close()