Menu
Index

Sample run 3

   
In this sample run, a sample application was performed to analyze full PSG record. Although it is based on real data, the purpose of this example is simply to explain how PSGMiner could be used for testing a hypothesis, and it is not meant to suggest a clinical procedure. In this sample study it is aimed to investigate relation between SpO2 signal  and other PSG signals. In other words, this was done to see whether the SpO2 value could be auto-marked by using signals from All PSG channels (without SpO2 Signal). The detailed statistical verification of the study results for the sample run is shown in the user interface of Machine Learning and below. According to the results for the sample run, the max accuracy rate is 71 %. A careful study of the results indicates that there is no important distinction between SpO2 and other signals However, this is by no means a demanding study, and it was only carried out to illustrate how our software can be used in clinical events.
 
PSG data
PSG data that are used at the test run 3 were taken from one real patients, whose identification information is kept as secret. (Patient age: 55 weight: 96kg, height: 173 cm, neck circumference: 43 cm, sex male).The patient was diagnosed with periodic leg movement disorder and underwent a one-night PSG for diagnosis. These PSG records are recorded with 44-channel polygraph (Compumedics 44E series, Avustralia) and these archive records are obtained through the use of complete PSG techniques in the Sleep Laboratory of the Faculty of Medicine, Trakya University.
 
Epoch length is  5 sn.
Record duration 7:21:00 (26460 sn)
Epoch count is 5292
 
 
Feature selection
Channels and features: All available channels and features in PSGMiner
Evaluator: InfoGain
Search method: Ranker
 
After feature selection, selected channels and features for classification.
_________________________
ROC Alpha 
C4A1 Nonlinearenergy
LOC Theta 
LOC Alpha 
LOC Linelength 
LOC Nonlinearenergy 
C4A1 ARmodellingerror8 
O2A1 ARmodellingerror8 
C4A1 Linelength 
F4A1 ARmodellingerror8 
SleepStages 
O2A1 Linelength 
ROC Nonlinearenergy 
ROC Theta 
LOC HiguchiFD 
C3A2 Waveletenergy 
O2A1 Nonlinearenergy 
ROC Meanpowerspectrum 
ROC Totalspectrum 
O1A2 Waveletenergy 
ROC Linelength 
ROC Waveletenergy 
F4A1 Alpha 
HR Heartrate 
LOC Totalspectrum 
LOC Meanpowerspectrum 
C3A2 Maxval 
C4A1 Alpha 
F4A1 Nonlinearenergy 
F3A2 Waveletenergy 
C4A1 Totalspectrum 
C4A1 Meanpowerspectrum 
ROC HiguchiFD 
F4A1 Linelength 
C3A2 RMSamplitude 
SpO2 class {6,7,8,9}                  
class 6: 60<= Mean (SpO2) <70
class 7: 70<= Mean (SpO2) <80
class 8: 80<= Mean (SpO2) <90
class 9: 90<= Mean (SpO2) <100
 
Classification Results (PSGMiner and WEKA)
 
PSGMiner
 
Neural Network (Back Propagation)
 
Neuron Type : Sigmoid
Hidenlayer Count : 1
Hidenlayer 0 Neuron Count : 3
Learning Rate : 0.3
Maximum Iteration : 500
Epsilon : 1E-8
Moment : 0.2
Scale : 0.1
 
---------------
 
Instances :980
Attributes :4
   
--Summary-- 
   
Accuracy :%61.429
Root mean squared error :0.543
Kappa Value :0.015
6==>
TP :0 TN :602 FP :2 FN :19
Sensitivity :0
Specificity :0.997
Precision :0
Negative Predictive Value :0.969404186795491
False Negative Rate :1
F1 score :0
Matthews correlation coefficient :-0.0100653186611269
ROC Area :0.498344370860927
7==>
TP :1 TN :601 FP :26 FN :64
Sensitivity :0.015
Specificity :0.959
Precision :0.037
Negative Predictive Value :0.903759398496241
False Negative Rate :0.985
F1 score :0.022
Matthews correlation coefficient :-0.039296159785918
ROC Area :0.486958655379708
8==>
TP :4 TN :598 FP :14 FN :261
Sensitivity :0.015
Specificity :0.977
Precision :0.222
Negative Predictive Value :0.69615832363213
False Negative Rate :0.985
F1 score :0.028
Matthews correlation coefficient :NAN
ROC Area :0.496109261314589
9==>
TP :597 TN :5 FP :336 FN :34
Sensitivity :0.946
Specificity :0.015
Precision :0.64
Negative Predictive Value :0.128205128205128
False Negative Rate :0.054
F1 score :0.763
Matthews correlation coefficient :NAN
ROC Area :0.480390015383114
Average==>
Sensitivity :0.614285714285714
Specificity :0.356563080818673
Precision :0.474546059938802
Negative Predictive Value :0.349533093991143
False Negative Rate :0.385714285714286
F1 score :0.500639506071287
Matthews correlation coefficient :NAN
ROC Area :0.485424397552193
Confusion Matrix 6 (Classified as)  7 (Classified as)  8 (Classified as)  9 (Classified as)
            6                  0                       0                       1                       18
            7                  0                       1                       2                       62
            8                  0                       5                       4                       256
            9                  2                       21                     11                     597
 
____________________________________________
 
WEKA Results
 
Neural Network (MultilayerPerceptron )
 
 
Options: -L 0.3 -M 0.2 -N 500 -V 0 -S 0 -E 20 -H a
 
Correctly Classified Instances         674               67.6707 %
Incorrectly Classified Instances       322               32.3293 %
Kappa statistic                          0.3275
Mean absolute error                      0.1767
Root mean squared error                  0.3396
Relative absolute error                 69.8616 %
Root relative squared error             95.5774 %
Total Number of Instances              996    
 
 
=== Detailed Accuracy By Class ===
 
              TP Rate   FP Rate   Precision   Recall  F-Measure   ROC Area  Class
                0.105     0.005      0.286     0.105     0.154      0.882    6
                0.569     0.048      0.451     0.569     0.503      0.931    7
                0.347     0.144      0.47      0.347     0.399      0.702    8
                0.842     0.474      0.764     0.842     0.801      0.785    9
Weighted Avg.    0.677     0.349      0.656     0.677     0.661      0.774
 
 
=== Confusion Matrix ===
 
  a   b   c   d   <-- classified as
  2  13   0   4 |   a = 6
  4  37  14  10 |   b = 7
  0  22  93 153 |   c = 8
  1  10  91 542 |   d = 9
 
____________________________________________
 
K-Nearest Neighbor
 
Options: -K 1 -W 0 -A "weka.core.neighboursearch.LinearNNSearch -A "weka.core.EuclideanDistance -R first-last""
 
IB1 instance-based classifier
using 1 nearest neighbour(s) for classification
 
Correctly Classified Instances         708               71.0843 %
Incorrectly Classified Instances       288               28.9157 %
Kappa statistic                          0.4271
Mean absolute error                      0.1456
Root mean squared error                  0.3794
Relative absolute error                 57.5354 %
Root relative squared error            106.7928 %
Total Number of Instances              996    
 
 
=== Detailed Accuracy By Class ===
 
              TP Rate   FP Rate   Precision   Recall  F-Measure   ROC Area  Class
                0.368     0.007      0.5       0.368     0.424      0.699    6
                0.538     0.03       0.556     0.538     0.547      0.787    7
                0.537     0.185      0.516     0.537     0.527      0.676    8
                0.811     0.335      0.816     0.811     0.813      0.738    9
Weighted Avg.    0.711     0.269      0.712     0.711     0.711      0.724
 
 
=== Confusion Matrix ===
 
  a   b   c   d   <-- classified as
  7   7   4   1 |   a = 6
  4  35  16  10 |   b = 7
  3  14 144 107 |   c = 8
  0   7 115 522 |   d = 9
 
____________________________________________
 
Naive Bayes
 
 
Naive Bayes Classifier
 
Correctly Classified Instances         384               38.5542 %
Incorrectly Classified Instances       612               61.4458 %
Kappa statistic                          0.116
Mean absolute error                      0.3029
Root mean squared error                  0.5264
Relative absolute error                119.736  %
Root relative squared error            148.1582 %
Total Number of Instances              996    
 
 
=== Detailed Accuracy By Class ===
 
              TP Rate   FP Rate   Precision   Recall  F-Measure   ROC Area  Class
                0.158     0.025      0.111     0.158     0.13       0.809    6
                0.862     0.226      0.211     0.862     0.338      0.871    7
                0.451     0.448      0.271     0.451     0.338      0.513    8
                0.317     0.148      0.797     0.317     0.453      0.712    9
Weighted Avg.    0.386     0.231      0.604     0.386     0.409      0.671
 
 
=== Confusion Matrix ===
 
  a   b   c   d   <-- classified as
  3  14   2   0 |   a = 6
  3  56   5   1 |   b = 7
  11  85 121  51 |   c = 8
  10 111 319 204 |   d = 9
 
____________________________________________
 
Decision Trees
 
 
 
Options: -C 0.25 -M 2
 
J48 pruned tree
 
Correctly Classified Instances         672               67.4699 %
Incorrectly Classified Instances       324               32.5301 %
Kappa statistic                          0.3511
Mean absolute error                      0.1728
Root mean squared error                  0.3828
Relative absolute error                 68.3046 %
Root relative squared error            107.7557 %
Total Number of Instances              996    
 
 
=== Detailed Accuracy By Class ===
 
              TP Rate   FP Rate   Precision   Recall  F-Measure   ROC Area  Class
                0.211     0.013      0.235     0.211     0.222      0.74     6
                0.462     0.033      0.492     0.462     0.476      0.76     7
                0.466     0.196      0.466     0.466     0.466      0.636    8
                0.797     0.389      0.789     0.797     0.793      0.713    9
Weighted Avg.    0.675     0.307      0.672     0.675     0.673      0.696
 
 
=== Confusion Matrix ===
 
  a   b   c   d   <-- classified as
  4   6   8   1 |   a = 6
  7  30  15  13 |   b = 7
  3  17 125 123 |   c = 8
  3   8 120 513 |   d = 9
 
 
 
The online help was made with Dr.Explain