Healthwatcher fit

spo2evaluation/evaluation/blant_altman.py

@@ -130,7 +130,8 @@ def BlandAltmanPlot(Xdata, Ydata, fig_num=None):
 
     if fig_num != None:
 
-        plt.figure(fig_num)
+
+        plt.figure()
 
         ##Set axis limits, this will account for the maximum average difference and the upper and lower
         ##limits of agreement, and all data points.
@@ -151,12 +152,14 @@ def BlandAltmanPlot(Xdata, Ydata, fig_num=None):
             ms="8",
             lw="2",
         )
-        # Add the mean, upper and lower levels of agreement to the Bland-Altman plot.
+        # Add the mean, upper and lower levels of agrespo2evaluation/evaluation/blant_altman.pyement to the Bland-Altman plot.
         plt.axhline(y=mean_difference, lw="2", color="k", ls="dashed")
         plt.axhline(y=upper_limit, lw="2", color="k", ls="dashed")
         plt.axhline(y=lower_limit, lw="2", color="k", ls="dashed")
         # Horizontal axis label
-        plt.xlabel("Average difference", fontsize="16")
+        plt.xlabel("Original SPO2 value", fontsize="16")
         # Vertical axis label
         plt.ylabel("Difference", fontsize="16")
+        plt.xlim(min(Average), max(Average))
+        plt.show()
     return (mean_difference, std_difference, upper_limit, lower_limit)

spo2evaluation/modelling/lamonaca_and_nemcova/lamonaca_2015.py

@@ -8,7 +8,7 @@
 import numpy as np
 import math
 
-import utils
+from . import utils
 
 
 class Camera_specs(object):

spo2evaluation/modelling/lamonaca_and_nemcova/nemcova_2020.py

@@ -3,7 +3,7 @@
 import numpy as np
 import math
 
-import utils
+from . import utils
 import matplotlib.pyplot as plt
 
 

spo2evaluation/modelling/wang_2017/wang_2017.py

@@ -3,7 +3,7 @@
 import numpy as np
 import math
 
-import utils
+from . import utils
 import matplotlib.pyplot as plt
 import scipy
 

spo2evaluation/preprocessing/data_loader_pandas.py

@@ -155,7 +155,7 @@ def __init__(self, data_path):
 
             self.meta = pd.DataFrame([data_sample_label_id, meta], ["sample_id", "fps"]).T
 
-            self.number_of_videos = self.sample_data['sample_id'].max()
+            self.number_of_videos = len(self.meta.index)
 
             # self.data = self.data.T
             print(self.sample_data)
@@ -222,7 +222,7 @@ def load_pickle(self):
         self.ground_truths_sample = pd.read_pickle(self.path_gt)
         self.meta = pd.read_pickle(self.path_meta)
         # print("yo\n", self.sample_data.loc[self.sample_data['sample_id'] == 1])
-        self.number_of_videos = self.sample_data['sample_id'].max()
+        self.number_of_videos = len(self.meta.index)
         print("Loaded", self.number_of_videos)
 
     def is_pickled(self) -> bool:

src/evaluation/Spo2_evaluation.py

@@ -7,6 +7,8 @@
 sys.path.append('../..')
 from spo2evaluation.evaluation import blant_altman
 from argparse import ArgumentParser
+import pandas as pd
+import matplotlib.pyplot as plt
 
 
 def main():
@@ -21,18 +23,21 @@ def main():
     args = parser.parse_args()
     print("Launching the evaluatio for", args.data_file)
 
+    results_df = pd.read_csv(args.data_file)
+
+
     # Extract Camera data
-    Data_camera = []
+    Data_camera = results_df['spo2_pred']
 
     # Extract GT data
-    Data_gt = []
+    Data_gt = results_df['spo2_gt']
 
     # Do statistical estimation
     least_square_slope_1, least_square_y_intercept_1, R_and_P_1 = blant_altman.Correlation(
         Data_camera, Data_gt
     )
     mean_difference_1, std_difference_1, upper_limit_1, lower_limit_1 = blant_altman.BlandAltmanPlot(
-        Data_camera, Data_gt
+        Data_camera, Data_gt, fig_num=1
     )
 
 

src/fit_healthwatcher.py

@@ -0,0 +1,54 @@
+import numpy as np
+from sklearn.linear_model import LinearRegression
+import pandas as pd
+import os
+
+DATA_DIR = '../data/0430/community'
+
+gts = pd.read_pickle(os.path.join(DATA_DIR, 'ground_truths_sample'))
+meta = pd.read_pickle(os.path.join(DATA_DIR, 'meta'))
+df = pd.read_pickle(os.path.join(DATA_DIR, 'sample_data'))
+
+df[['mean_red', 'std_red', 'mean_green', 'std_green', 'mean_blue', 'std_blue']] = df[['mean_red', 'std_red', 'mean_green', 'std_green', 'mean_blue', 'std_blue']].apply(pd.to_numeric)
+
+df = df.groupby(['sample_id']).mean()
+
+y = gts['SpO2'].to_numpy().reshape(-1,1)
+
+mean_red = df['mean_red'].to_numpy()
+std_red = df['std_red'].to_numpy()
+mean_blue = df['mean_blue'].to_numpy()
+std_blue = df['std_blue'].to_numpy()
+mean_green = df['mean_green'].to_numpy()
+std_green = df['std_green'].to_numpy()
+
+X = np.array([mean_red, std_red, mean_blue, std_blue, mean_green]).T # <-- 'new' method, linear regression on all features
+# X = (std_red / mean_red) / (std_blue / mean_blue).reshape(-1,1) <-- healthwatcher's method, aggregates all into 1
+
+split=int(len(X)*0.75)
+
+X_train=X[:split]
+X_test=X[split:]
+
+y_train=y[:split]
+y_test=y[split:]
+
+reg = LinearRegression().fit(X_train, y_train)
+print(reg.score(X, y))
+print(reg.coef_)
+print(reg.intercept_)
+
+y_pred = reg.predict(X_test)
+
+import matplotlib.pyplot as plt
+
+plt.scatter(x=y_pred, y=y_test)
+plt.xlabel('predictions')
+plt.ylabel('ground truth values')
+plt.show()
+import time
+
+pd.DataFrame(data={'spo2_gt': y_test.reshape(1,-1)[0], 'spo2_pred': y_pred.reshape(1,-1)[0]}).to_csv(f'../results/hw-fit-{int(time.time())}')
+
+
+

src/healthwatcher.py

@@ -5,6 +5,8 @@
 import spo2evaluation
 from spo2evaluation.modelling.healthwatcher import healthwatcher
 from spo2evaluation.preprocessing import data_loader_pandas
+import pandas as pd
+import time
 
 
 if __name__ == "__main__":
@@ -35,6 +37,12 @@
         print("Picling the data")
         dataset.pickle_data()
 
+    predictions = {'id': [],
+                 'spo2_gt': [],
+                 'spo2_pred': [],
+                 }
+
+
     for i in range(dataset.number_of_videos):
         sample, gt, meta = dataset.get_video(i)
 
@@ -48,4 +56,16 @@
 
         spo2 = healthwatcher.health_watcher(blue, blue_std, red, red_std, fps,
                                             smooth=False)
-        print(spo2)
+
+        print(gt['path'])
+        survey_id = gt['path'].iloc[0].split('/')[-1].split('=')[-1]
+
+
+        predictions['id'].append(survey_id)
+        predictions['spo2_gt'].append(gt['SpO2'].iloc[0])
+        predictions['spo2_pred'].append(spo2)
+
+    results_df = pd.DataFrame(predictions)
+
+    results_df.to_csv(f'../results/healthwatcher_{int(time.time())}.csv')
+

src/wang.py

@@ -1,5 +1,7 @@
 import cv2
-from spo2evaluation.modelling import wang_2017
+import sys
+sys.path.append('..') # expected to be run from inside src/
+from spo2evaluation.modelling.wang_2017 import wang_2017
 
 
 def main():