def export_database(database_file_folder):
manifest_db = manifest.Manifest()
manifest_csv = open(database_file_folder + '/manifest.txt', 'w')
manifest_csv.write("id\tfilename\tmd5\tsize\tstate\n")
for wsi in manifest_db.get_projects():
for temp in wsi:
manifest_csv.write(str(temp) + '\t')
manifest_csv.write("\n")
mission_db = mission.Mission()
mission_csv = open(database_file_folder + '/predict_mask.txt', 'w')
mission_csv.write("id\tfilename\tmd5\tsize\tstate\n")
for wsi in mission_db.get_predict_masks():
for temp in wsi:
mission_csv.write(str(temp) + '\t')
mission_csv.write("\n")
def get_table(slide_uuid):
mission_db = mission.Mission()
if slide_uuid == "":
jobs = mission_db.get_predict_masks()
else:
jobs = mission_db.get_predict_masks_by_uuid(slide_uuid)
data = []
for job in jobs:
temp = []
temp.append(job[0]) # ID
# slide_id = manifest_db.get_project_by_uuid(job[2])[0]
# print('static/data/analysis_data/' + str(job[2]) + '/mission' + str(job[0]) + '_result.png')
if os.path.exists('static/data/analysis_data/' + str(job[2]) + '/mission' + str(job[0]) + '_result.png'):
temp.append('<a href="/static/data/analysis_data/' + str(job[2]) + '/mission' + str(
job[0]) + '_result.png"target="_blank">' + str(job[1]) + '</a>')
else:
temp.append(str(job[1]))
temp.append('<button type="button" onclick="btn_click(\'remove_mission\',' +
str(job[0]) + ')">remove</button>')
temp.append('<a href="/slide?slide_id=' + str(job[1]) + '"target="_blank">' + str(job[2]) + '</a>')
temp.append(str(job[3]))
temp.append(bar(str(job[4] / job[5] * 100)))
temp.append(str(job[4]) + '/' + str(job[5]))
if job[6] is None:
temp.append("Not Finished")
elif "mission" in job[6]:
temp.append('<a href="/slide?slide_id=' + str(job[1]) +
'&mask_url=' + job[6] + '"target="_blank">' + job[6] + '</a>'
)
else:
temp.append('<a href="/slide?slide_id=' + str(job[1]) +
'&mask_url=' + 'mission' + str(job[0]) + '_' + str(job[3]) + '.png' +
'"target="_blank">' + job[6] + '</a>'
)
temp.append(str(job[8]))
temp.append(str(job[7]))
data.append(temp)
return data
def predict_mask_with_job_id(slide_id, model_name="0"):
# if str(model_name) == '0':
# model_path = "Model/resnet_34_crd_model_59.pth"
# elif str(model_name) == '1':
# model_path = "Model/resnet_34_transfer_predicted_crd_model_best.pth"
# else:
model_path = "models/" + model_name
make_bg(slide_id)
manifest_db = manifest.Manifest()
info = manifest_db.get_project_by_id(slide_id)
mission_db = mission.Mission()
job_id = mission_db.insert(slide_uuid=info[1],
slide_id=info[0],
job_type=model_name,
total=-1)
original_data_folder = original_data_root + info[1] + '/'
analysis_data_folder = analysis_data_root + info[1] + '/'
svs_file = original_data_folder + info[2]
# if not os.path.exists(analysis_data_folder + 'patch/'):
# os.mkdir(analysis_data_folder + 'patch/')
mask = cv2.imread(analysis_data_folder + info[4], cv2.IMREAD_GRAYSCALE)
oslide = openslide.OpenSlide(svs_file)
w, h = oslide.dimensions
pre_result = numpy.zeros((mask.shape[0], mask.shape[1], 4),
dtype=numpy.uint8)
times = mask.shape[1] / w * 2000
mission_db.update_total_by_id(job_id=job_id, total=w // 2000 * h // 2000)
if "subtype" not in str(model_name):
myModule = predict_module.ResNetClassification(model_path=model_path,
num_classes=2,
batch_size=64,
num_workers=0)
for x in range(w // 2000):
for y in range(h // 2000):
if available_region(mask[int(y * times):int(y * times + times),
int(x * times):int(x * times +
times)]):
# if os.path.exists(analysis_data_folder + 'patch/' + str(x) + '_' + str(y) + '.png'):
# patch = Image.open(analysis_data_folder + 'patch/' + str(x) + '_' + str(y) + '.png')
# else:
# patch = oslide.read_region((x * 2000, y * 2000), 0, (2000, 2000))
# patch.save(analysis_data_folder + 'patch/' + str(x) + '_' + str(y) + '.png')
patch = oslide.read_region((x * 2000, y * 2000), 0,
(2000, 2000))
patch = numpy.array(patch.convert('RGB'))
result = myModule.predict(
numpy.resize(patch, tuple([1, 2000, 2000, 3])))
# result = myModule.predict(total_image)
# probablity_list.append((x, y, result[0]))
pre_result[int(y * times):int(y * times + times),
int(x * times):int(x * times + times),
numpy.argmax(result[0])] = result[
0, numpy.argmax(result[0])] * 255
mission_db.update_finished_by_id(job_id=job_id,
finished=x * h // 2000 + y +
1)
pre_result = post_processing(pre_result)
mask_file_name = 'mission' + str(job_id) + '_' + str(
model_name) + '.png'
cv2.imwrite(analysis_data_folder + mask_file_name, pre_result)
mission_db.update_predict_mask_by_id(job_id=job_id,
predict_mask=mask_file_name)
mask_file = cv2.imread(analysis_data_folder + mask_file_name)
original_file_name = "smaller_image.png"
original_file = cv2.imread(analysis_data_folder + original_file_name)
alpha = numpy.zeros(mask_file.shape)
alpha[:, :, 0] = mask_file[:, :, 0]
alpha[:, :, 1] = mask_file[:, :, 0]
alpha[:, :, 2] = mask_file[:, :, 0]
alpha = alpha.astype(float) / 128
mask_file = mask_file.astype(float)
original_file = original_file.astype(float)
mask_file = cv2.multiply((1 - alpha) * 0.7, mask_file)
original_file = cv2.multiply(0.3 + alpha * 0.7, original_file)
out_image = original_file + mask_file
result_file_name = 'mission' + str(job_id) + '_result.png'
cv2.imwrite(analysis_data_folder + result_file_name, out_image)
elif "subtype" in str(model_name) and "hybrid" in str(model_name):
myModule = predict_module.ResNetClassification(model_path=model_path,
num_classes=4,
batch_size=64,
num_workers=0)
for x in range(w // 2000):
for y in range(h // 2000):
if available_region(mask[int(y * times):int(y * times + times),
int(x * times):int(x * times +
times)]):
patch = oslide.read_region((x * 2000, y * 2000), 0,
(2000, 2000))
patch = numpy.array(patch.convert('RGB'))
result = myModule.predict(
numpy.resize(patch, tuple([1, 2000, 2000, 3])))
pre_result[int(y * times):int(y * times + times),
int(x * times):int(x * times + times),
numpy.argmax(result[0])] = result[
0, numpy.argmax(result[0])] * 255
mission_db.update_finished_by_id(job_id=job_id,
finished=x * h // 2000 + y +
1)
sub = ["health", "ccRCC", "pRCC", "chRCC"]
result_sub = [0, 0, 0, 0]
for i in range(4):
result_sub[i] = numpy.sum(pre_result[:, :, i] != 0)
result_sub_sum = numpy.sum(result_sub)
summary = ""
for i in range(4):
result_sub[i] = round(result_sub[i] / result_sub_sum * 100, 2)
summary = summary + sub[i] + ": " + str(result_sub[i]) + "%, "
result = numpy.zeros((pre_result.shape[0], pre_result.shape[1], 3))
for i in range(3):
result[:, :, i] = pre_result[:, :, i + 1]
mask_file_name = 'result_mission' + str(job_id) + '_' + str(
model_name) + '.png'
cv2.imwrite(analysis_data_folder + mask_file_name, result)
result = numpy.zeros((pre_result.shape[0], pre_result.shape[1], 3))
i = 1
result[:, :, 0] = pre_result[:, :, i + 1] * [15] / 255
result[:, :, 1] = pre_result[:, :, i + 1] * [147] / 255
result[:, :, 2] = pre_result[:, :, i + 1] * [254] / 255
i = 0
result[:, :, i] += pre_result[:, :, i + 1]
i = 2
result[:, :, i] += pre_result[:, :, i + 1]
mask_file_name = 'mission' + str(job_id) + '_' + str(
model_name) + '.png'
cv2.imwrite(analysis_data_folder + mask_file_name, result)
mission_db.update_predict_mask_by_id(job_id=job_id,
predict_mask=summary)
elif "_subtype" in str(model_name) and "hybrid" not in str(model_name):
myModule = predict_module.ResNetClassification(model_path=model_path,
num_classes=3,
batch_size=64,
num_workers=0)
for x in range(w // 2000):
for y in range(h // 2000):
if available_region(mask[int(y * times):int(y * times + times),
int(x * times):int(x * times +
times)]):
patch = oslide.read_region((x * 2000, y * 2000), 0,
(2000, 2000))
patch = numpy.array(patch.convert('RGB'))
result = myModule.predict(
numpy.resize(patch, tuple([1, 2000, 2000, 3])))
pre_result[int(y * times):int(y * times + times),
int(x * times):int(x * times + times),
numpy.argmax(result[0])] = result[
0, numpy.argmax(result[0])] * 255
mission_db.update_finished_by_id(job_id=job_id,
finished=x * h // 2000 + y +
1)
sub = ["ccRCC", "pRCC", "chRCC"]
result_sub = [0, 0, 0]
for i in range(3):
result_sub[i] = numpy.sum(pre_result[:, :, i] != 0)
result_sub_sum = numpy.sum(result_sub)
summary = ""
for i in range(3):
result_sub[i] = round(result_sub[i] / result_sub_sum * 100, 2)
summary = summary + sub[i] + ": " + str(result_sub[i]) + "%, "
result = numpy.zeros((pre_result.shape[0], pre_result.shape[1], 3))
for i in range(3):
result[:, :, i] = pre_result[:, :, i]
mask_file_name = 'mission' + str(job_id) + '_' + str(
model_name) + '.png'
cv2.imwrite(analysis_data_folder + mask_file_name, result)
mission_db.update_predict_mask_by_id(job_id=job_id,
predict_mask=summary)
elif "MixMatch-subtype" in str(model_name):
myModule = predict_module2.ResNetClassification(model_path=model_path,
batch_size=64,
num_workers=0)
for x in range(w // 2000):
for y in range(h // 2000):
if available_region(mask[int(y * times):int(y * times + times),
int(x * times):int(x * times +
times)]):
patch = oslide.read_region((x * 2000, y * 2000), 0,
(2000, 2000))
patch = numpy.array(patch.convert('RGB'))
result1, result2 = myModule.predict(
numpy.resize(patch, tuple([1, 2000, 2000, 3])))
if result1[0, 1] > 0.5:
pre_result[int(y * times):int(y * times + times),
int(x * times):int(x * times + times),
0] = result1[0, 1] * 255
elif result1[0, 0] > 0.5:
pre_result[int(y * times):int(y * times + times),
int(x * times):int(x * times + times),
numpy.argmax(result2[0]) +
1] = result2[0,
numpy.argmax(result2[0])] * 255
mission_db.update_finished_by_id(job_id=job_id,
finished=x * h // 2000 + y +
1)
sub = ["health", "ccRCC", "pRCC", "chRCC"]
result_sub = [0, 0, 0, 0]
for i in range(4):
result_sub[i] = numpy.sum(pre_result[:, :, i] != 0)
result_sub_sum = numpy.sum(result_sub)
summary = ""
for i in range(4):
result_sub[i] = round(result_sub[i] / result_sub_sum * 100, 2)
summary = summary + sub[i] + ": " + str(result_sub[i]) + "%, "
result = numpy.zeros((pre_result.shape[0], pre_result.shape[1], 3))
for i in range(3):
result[:, :, i] = pre_result[:, :, i + 1]
mask_file_name = 'mission' + str(job_id) + '_' + str(
model_name) + '.png'
cv2.imwrite(analysis_data_folder + mask_file_name, result)
mission_db.update_predict_mask_by_id(job_id=job_id,
predict_mask=summary)
def get_total_number():
mission_db = mission.Mission()
return len(mission_db.get_predict_masks())
def remove_mission_by_id(job_id):
mission_db = mission.Mission()
mission_db.delete_predict_mask_by_id(job_id)
def clear_database():
manifest_db = manifest.Manifest()
manifest_db.delete_all_projects()
mission_db = mission.Mission()
mission_db.delete_all_predict()