1
$\begingroup$

So I am doing some research in the field of Medical Physics, in particular radiation treatment with photons and protons.

Now, I have two cases I am unsure about.

1) Here I have several patients. Each patient is treated (only simulated) with both photons and protons. From those treatment plans one can get several metrics out describing how good/bad each modality is. So for example, the mean dose to the target/tumour should be pretty much the same for both modalities, i.e. it shouldn't change much. However, how much radiation the brainstem gets (since this is radiation to the brain) is different depending on modality - and patient. So what tests could I do here? For me it's a bit unclear. I could compare between modalities, and see what is best, but that is only two data points for each patient. And since the brainstem toxicity can be VERY different for each patient, is there really any good way to use statistics on all patients as a group?

So the hypothesis would be something like: Is one treatment modality better than the other one, or are they the same.

2) Same patient group. And also photon vs. proton related. However, here we look at the movement within the brain instead. So we take weekly images, and see if anything has moved (target etc.), and then use the same dose calculation on each of these images to see if by using the same treatment plan as originally intended, how much difference would we get in brainstem toxicity and target coverage. So here I have for example brainstem toxicity for each scan/week (7 in total) for each patient. What tests would I use here? First a test to see if the weekly scans/movement has anything important to say, and then after than compare photons and protons? (See image for data example)

I guess the hypothesis is pretty much the same as before. Is one better than the other when looking at how they perform during weekly movement in patient.

enter image description here

I'm quite new to this field, and my statistic skills are kind of rusty at the moment. So I'm looking for help until my statistics course will begin later this year :)

$\endgroup$
1
  • 1
    $\begingroup$ It might be better if you split this into two separate questions. And you could improve them by clarifying what your specific hypothesis is in the first question ('better' is hard for us to judge), and your experimental design in the second one ('...by using the same treatment plan as originally intended' is also unclear). $\endgroup$
    – mkt
    Jul 25, 2017 at 4:49

1 Answer 1

1
$\begingroup$

Based on you statement, I think

For the first one, Anova or 2 Sample t test can be applied(if the sample size is < 30) and choose the hypothesis based on the significance of the p-value.

For the second one, Paired t test can be done since the data will be based on pre and post treatment or each week progress.

$\endgroup$
5
  • $\begingroup$ So for the second one, would I use the paired t-test on all pre- and post treatment data for one patient at a time ? For example, for each patient I have 10 weeks/images where this treatment is performed, i.e. I get 10 different data sets in total per patient - for each treatment modality (photons and protons). So would I just take Patient 1 and use the test on one set of data from the photon case against the proton case, or...? I'm thinking, since the data can vary a lot week by week, and patient by patient, I shouldn't just add all patients up for the same type of data, right ? $\endgroup$ Jul 27, 2017 at 3:03
  • $\begingroup$ Hmmm, should you have written something that I can't see, or should I be imagining it ? :) $\endgroup$ Jul 27, 2017 at 4:31
  • $\begingroup$ Based on your comment I assume that the data looks like this, Patient 1 - (Week 1,2,3,4,5,6,7,8,9,10), Patient 2 - (Week 1,2,3,4,5,6,7,8,9,10) and so on. If this is the case then take 2 data sets for each test like, comparing before treatment and after week 1 or after week 1 and after week 2 or after week 5 and after week 10 and so on. I don't understand this statement form your comment " would I just take Patient 1 and use the test on one set of data from the photon case against the proton case, or...? ". Could provide a sample data set to let me imagine about what you are asking. $\endgroup$
    – Sam Gladio
    Jul 27, 2017 at 4:35
  • $\begingroup$ I have attached a data example to my main post. That is an example of one metric for photons, for one patient, for all weeks. I know have this for also protons, and several other patients - including other metrics/data points as well... $\endgroup$ Jul 27, 2017 at 4:47
  • $\begingroup$ According to your obejective - So here I have for example brainstem toxicity for each scan/week (7 in total) for each patient. What tests would I use here? First a test to see if the weekly scans/movement has anything important to say, and then after than compare photons and protons? - 1) You have to compare weekly data to each other for all patients i.e. Week 1 as one column with Week 2 as another column of all patients - paired t test. 2) Comparing protons(as one column) and photons(as another) involve 2 sample t test where you can do for each week(all patients) or each patient(all weeks). $\endgroup$
    – Sam Gladio
    Jul 27, 2017 at 5:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.