accuracy = (correctly predicted class / total testing class) × 100%

OR,

The accuracy can be defined as the percentage of correctly classified instances (TP + TN)/(TP + TN + FP + FN). where TP, FN, FP and TN represent the number of true positives, false negatives, false positives and true negatives, respectively.

also you can use standard performance measures:

Sensitivity = TP / TP + FN

Specificity = TN / TN + FP

Precision = TP / TP + FP

True-Positive Rate = TP / TP + FN

False-Positive Rate = FP / FP + TN

True-Negative Rate = TN / TN + FP

False-Negative Rate = FN / FN + TP

For good classifiers, TPR and TNR both should be nearer to 100%. Similar is the case with precision and accuracy parameters. On the contrary, FPR and FNR both should be as close to 0% as possible.

have a look at the following terms;

Precision, Recall, and Confusion table.

it would help you understand what Mohammed Y Kamil is saying.

Hi Nisreen,

I am attaching a link to a paper I wrote on the topic. I think it will answer your question thoroughly. Let me just point out that you need non-diseased patients as well as diseased patients in your analysis. The non-diseased patients will give you the specificity and negative predictive values. They will also contribute to the ROC curve value.

I hope this helps

Ariel

Article Measuring diagnostic and predictive accuracy in disease mana...

Dear Colleagues,

Good Day,

Here are some websites/ articles/ papers/ youtube related to your question, even though it is general one ( is it the accuracy of measurement / instrument etc. etc. 

https://www.sophia.org/tutorials/accuracy-and-precision--3

http://digital.ni.com/public.nsf/allkb/8BA2242D4BCC41B286256D1D00815B90

http://www.mathsisfun.com/accuracy-precision.html

http://www.ehow.com/how_6199228_calculate-percent-accuracy.html

http://www.wikihow.com/Calculate-Precision

https://www.youtube.com/watch?v=5APhVxCEPFs

https://www.youtube.com/watch?v=-7rzmMUEaW8

in my research i used 

The accuracy of the algorithm detecting the ROI.

Accuracy = TP/ (TP+TN)*100%

where;

The true positive (TP) and true negatives (TN) are correct classification. A false positive (FP) is when the outcome of the algorithm is incorrectly predicted, when the in reality it is actually present in the image.

But if your using classification methods like BOF or Neutral Network the classification rate produce by these techniques can be consider as accuracy as well. Because in reality classification categorizes the objects we detect using image processing. 

hi,

Accuracy = True Positive / (True Positive+True Negative)*100.

If your aneurysms detection and segmentation are two different steps, It's better to evaluate your detection.and segmentation methods seperately.

you can use precision and TP rate (TPR or recall rate) to demonstrate your result, e.g., precision=17/(17+4), and TPR= 17/17.

For evaluation you segmentation method, you can use TPR, FPR and Similarity index to evaluate the area error between your segmentation result and the ground truth(can be manually drawed by radiologist).

Thank you all for answering my question.

@ Mr Min Xian,

That means,

In this case: TPR=17, FPR=4.

precision=17/(17+4)= 80.95%

BUT, Why have you written TPR=17/17=100%  ??

*TPR=TP/(TP+FN)  I don't have FN value!

false negative is  when particular algorithm is unable to detect the object but the object in present within the test area. 

For example ; when a algorithm is to detect tumor but returns a negative result but the patient actually does have tumor.

So how many times your algorithm fails to detect the real thing?

@ Nisreen Sulayman ,

In your case TP = 17, FP = 4, and FN = 0, so the TP (recall) ratio = 100% and the Precision = 80.95%.

Your results demonstrate that your method has high TPR (sensitivity), but the precision is not so well. 

TPR and precision should be used together to illustrate the overall performance of the method. 

Thank you all.

@ Min Xian, Thank you.

Try to evaluate F1 score https://en.wikipedia.org/wiki/F1_score

Min Xian may i ask why TP=17 ?

To calculate the overall accuracy you add the number of correctly classified sites and divide it by the total number of reference site. We could also express this as an error percentage, which would be the complement of accuracy: error + accuracy = 100%.

Classification Accuracy = TP+TN/TP+TN+FN x 100%





"Correct classified = Total Classified - Miss-classified" is a simple formula

Confusion matrix provides a performance evaluation measurement for a recognition scenario. It is a grid of actual classification vs. predicted classification. Suppose in a box there some rectangular objects (R) and some triangular (T) objects. An object is taken out at random and the computer is asked to predict its type (class). Four situations can arise for each class :

– Actually R and predicted as R (correct classification : true positive)

– Actually not R and predicted as not R (correct classification : true negative)

– Actually not R but predicted as R (incorrect classification : false positive)

– Actually R but predicted as not R (incorrect classification : false negative)

Following measures can be calculated :

(1) Accuracy = Correct predictions / Total predictions

= (TP + TN) / (TP + TN + FP + FN)

(2) Precision = True positive / Predicted positive = TP / (TP + FP)

(3) Recall = True positive / Actual positive = TP / (TP + FN)

Different systems are compared with respect to these values to find out the better system

F1 Score (Weighted average of Precision and Recall) = 2* ((precision * recall)/(precision + recall))

In your case, acc of 80.95% is correct. However, you should Senstivity and Specificity in case of medical imaging if the data is imbalance. For balanced data, precision, recall and F-measure are also used. Imp, use AUC and ROC which are also good and advanced level analysis. best wishes

how to find the accuracy of experimental value to simulation value.

I agree with Dr. Mohammed Y. Kamil

For more detail we have:

https://www.mathworks.com/help/images/ref/bfscore.html

https://stackoverflow.com/questions/13974167/how-to-test-accuracy-of-segmentation-algorithm

Dear Nisreen Sulayman

We have many methods to calculate accuracy. This article will clarify this to you

Article Novel Feature Selection and Voting Classifier Algorithms for...

Min Xian I have a question, if I have 20 pics for testing, 13 giving correct results and 7 gives wrong results so I have the accuracy = 20/(20 + 7) is it correct or not

http://www.50northspatial.org/classification-accuracy-assessment-confusion-matrix-method/

https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&cad=rja&uact=8&ved=2ahUKEwiAlZn0utfyAhVMVH0KHe1QD_0QFnoECB4QAQ&url=http%3A%2F%2Fweb.pdx.edu%2F~nauna%2Fresources%2F9-accuracyassessment.pdf&usg=AOvVaw1q6MiParvvafj1I8ml5ovQ

For analyzing the segmentation performance some more popular measures include Relative volume difference(RVD), Dice score, Jaccard Index, Structural similarity index. In addition, PSNR can also give you the difference between the original image and the segmented image. In segmentation, we always have to analyze the amount of correct segmentation.