Course talk:CPSC522/Automatic Classification of Morphological Heart Arrhythmia

Hi Mehrdad,
Excellent work! I learned a lot from reading your page as I wasn’t really exposed to this topic before.
Comments/suggestions:

• Do you think using higher k values instead of 1 KNN may result in noise reduction and improve your results?
• Could you briefly explain how you fit a Hermite function to the heart beats and how the function classifies the heart beats?
• I am also eagerly waiting to see how the neural networks portion of your classification of feature vectors works out and performs relative to the k-Nearest Neighbors approach.
  

I really enjoyed reading your page. Keep up the good work.

Best Regards,
Adnan

Hi Adnan,

The results with higher k values are lower for some reason, the fitting is basically Gaussian for QRS complex and polynomial for the fluctuations on it. The results for Neural networks are very promising! and I will report them tonight

I really appreciate your positive comments and points, Thanks!

Hi Mehrdad Ghomi,

Very interesting topic, very informative.

I have some questions for you:

1. In Modeling part, you changed 203 intervals into 25 to reduce the complexity. Will this actually affect the precision and the sensitivity of your model?

2.About your Hermitian reduction function, I can understand that by that you make the system less complex, but can you also add more evidence to show us how it improved the result?

3.For most the methods mentioned in this page are all theories, but did not touch how did you apply those methods in your training. Can you talk more about like how the pattern is formed?

4. About your K-NN, you mentioned that the use of 1-NN is to reduce the complexity, and it works better than others like 3-NN, I really want to know why it is better, and what are the differences between them?

Hi,

The reduction in the feature vector size didn't really affect the results, as the new re-modelled signal is pretty good in terms of results.
Yes, I can perhaps put the results without the modelling there too for the second draft!
Actually the reason that why the 1-NN is better is still a question mark for me, but the results are certain that it is better.

I really appreciate your positive comments and points, Thanks!

Hey Mehrdad,

• The definition for H doesn't seem to need two base cases. H_1 is defined by the recursion and H_0.
• "Hermite polynomials are not orthogonal in general" - but wikipedia says they are... (Though I've no idea what orthogonal polynomials even mean.) Is there a difference in your definition that I'm missing?
• How did you fit a Hermite function to the heart beats? I assume this is how you're constructing the input features for your model.
• Essentially, how is the Hermite basis function used in processing the dataset and how does it characterize the heart beats?
• Is "1-NN" referring to k-NN with k=1? Seems like the results are showing that ECG is an amazing indicator of this heart condition... Is k-NN supposed to be better than a human expert? Is it useful because human doctors have to spend a lot of time testing if a patient has this heart condition?

Hi,

- The "H" in all of the sources that I looked has been initiated in two steps. - For the fitting point you mentioned, the QRS complex is remodelled via a Gaussian function, and then there is polynomial over it to cover the fluctuations - 1-NN is k-NN with k=1, and the results are high and certainly in a level that is possible to compete (and even beat?) human-level. - I would say they probably wont spend "a lot of time", but it is certainly an unnecessary task for them if the results are getting to the same level.

I really appreciate your positive comments and points, Thanks!

Hi Mehrdad,
This is going to be a rather long message. I found your project really exciting and I have lots of questions.
Some initial cosmetic suggestions that I would like to get across are as follows:
1. In a line or two, right at the beginning you might want to mention that heart arrhythmia is an irregular heart beat. Also which type of arrhythmia are you trying to classify or model? Bradyarrythmias or tachyarrhythmias?
2. You might want to add the definition of QRS complex (maybe in brackets where you first mention it) even though i noticed you did include it in your figure but it might not be very obvious.
3. Right above your gaussian section, there is no link to click for further information on Hermite basis functions. Same for the Gaussian function ,the KNN algorithm and neural networks.

Questions that came up in my mind as I was going through your page are outlined below:
1. How did you chop the continuous signals into units containing only one peak? Also, does this mean you extract more than one peak from each patient’s 30 minute ECG signal?
2. Is it possible for you to show how you remodelled the 203 intervals to 25 intervals and what was your resulting weighted vector for the combination of Hermite basis functions?
3. Using k value as 1 makes me a little uncomfortable. Isn’t it true that larger values of k reduce the effect of noise on classification? Do you think, using k=3 or 5 would improve your results?
4. The N training vectors that you are using are continuous variables? Or are they discrete y values at each of the 25 intervals? Would i be right in assuming the dimension of each training vector to be [1x25].
I really found your project very interesting and I think you’ve done a very good job explaining your problem description and the hypothesis is also very clear. You could probably add some more input to how you model your problem (for instance you could write your x,k_i and y vectors explicitly. I am really looking forward to you Neural network implementation and how it wold compare to this 1-KNN model as it seems almost everywhere neural nets are taking over.
Regards,
Ritika

Hello Ritika,

Thanks for your very useful comments. I have already fixed the links. I would certainly add the demanded information regarding the heart arrhythmia and my modelling for the next draft.

Thanks,

Mehrdad Ghomi