mat_convert.py

import scipy.io
import numpy as np
import h5py
import os
import errno
from random import shuffle

def setup_data():

	#np.set_printoptions()

	#print(sorted(os.listdir('braintumors_1')))
		
	csvFileName = "brain_scan_data.csv"#csv that will hold each brain scan in one row
	all_scan_array = []
	count = 0
	skip_count = 0
	count_1 = 0
	count_2 = 0
	count_3 = 0

	#files = sorted(os.listdir('braintumors_1'), key=lambda f: int(os.path.splitext(f)[0]))
	files = os.listdir('braintumors_1')

	shuffle(files) #COMMENT THIS OUT IF YOU WANT TO TEST ACCURACY BETWEEN THE DIFFERENT MODELS; KEEP IT TO TEST ACCURACY BETWEEN EACH INDIVIDUAL MODEL WITH THEMSELVES (don't compare different models), via MULTIPLE RUNS OF EACH MODEL 

	print("Converting all mat files to arrays, please wait... (may take a few minutes)")
	for file in files: #iterate through all .mat files in braintumors_1 directory
		#if count > 400:
		#	break
		currMatFileNum = file.split('.')[0] #current file in braintumors_1 directory; splits 3064.mat into ['3064', 'mat']
		#print("Attemping to convert mat file number: "+ currMatFileNum) #e.g. "3064" from 3064.mat

		#labels: 1 for meningioma, 2 for glioma, 3 for pituitary tumor
		try:
			with h5py.File("braintumors_1/{}".format(file), 'r') as f: #open .mat file to save to new .csv file corresponding to that .mat file

				label = f['cjdata/label'].value[0][0] #label is of type numpy.float64

				scan_array_2d = f['cjdata/tumorMask'].value
				#print(scan_array_2d)
				if (len(scan_array_2d)) != 512:
					print("Failed to convert mat file number {}, incorrect format!".format(currMatFileNum))
					skip_count += 1
					pass
					#print(len(scan_array_2d))
					#print(file)
				else:
					scan_array_1d = []
					for row in scan_array_2d:#convert the current brain scan array into one long array
						scan_array_1d.extend(row)
					
					if label == 1: #meningioma
						scan_array_1d.append(1)
						count_1 += 1
					elif label == 2: #glioma
						scan_array_1d.append(2)
						count_2 += 1
					elif label == 3: #pituitary tumor
						scan_array_1d.append(3)
						count_3 += 1

					all_scan_array.append(scan_array_1d)#append the scan_array_1d as one entry in the array of all scans
					#print("Success! Appending mat file number: "+ currMatFileNum)
		except:
			print("h5py failed!")
			raise
		count += 1

	print("Converted "+ str(count - skip_count) + " scans into {0} x {1} array.".format(len(all_scan_array), len(all_scan_array[0])))
	print("Skipped " + str(skip_count) + " images due to incorrect size (not 512x512).")
	print("Number of 1's:{0}\nNumber of 2's:{1}\nNumber of 3's:{2}".format(count_1, count_2, count_3))

	new_arr = np.array(all_scan_array, dtype=float)
	#print(new_arr)
	return new_arr