Results & Discussion

Based on the result, it have been discovered that this proposed method for improvement does not significantly improve the previous method. Out of sixty sample images, only three images show that the accuracy is actually improved, one from Stare database (image 03) and two from DRIVE database(image 23 and 31).The average value for accuracy at 0.89 and 0.71 for sensitivity. Even though this proposed method appeared to be extracted more vessel than previous results, it is actually only referring to how human eyes see the picture without any numerical calculation. Since this is a medical matter, it is important to determine how good the extracted result was. However, the result does show some improvement for sensitivity. The bold number in the tables indicates improvement by using this proposed method the sensitivity and the accuracy for this proposed method is obtained by comparing the proposed result with the hand-labelled manual1 from DRIVE database and a hand-labelled vessel network provided by Valentina Kouznetsov for STARE database. Sensitivity indicates how well the test predicts the vessel pixels while accuracy is used to measure how well the test predicts both vessel and non vessel pixels. In this paper, the result obtained from Adam Hoover and their proposed match spatial filter probing algorithm is compared with the hand-labelled results provided by Valentina Kouznetsov. Furthermore, there is a limitation in this proposed method. This method cannot differentiate between lesions and vessels, therefore lesions might be extracted instead of the vessel itself

PERFORMANCE OF PROPOSED METHOD

On Window &,CPU 2.13GHZ, using MATLAB R2011b, the computational time for the whole process is less than 1minute which is basically faster compared to manual that took almost 2hours to executed and other method.  

The performance for this method is calculated by sensitivity and accuracy. The true positive (TP), true negatives (TN), false positives(FP), and false negatives(FN) is basically  four outcomes of single prediction for a two-class with classes “1” as “YES” while “0” as “No”.

  

• Sensitivity will tell how well the test predicts the vessel pixels.
• Accuracy used to measure how well the tet predict both vessel and non vessel pixels.