Writing a research paper is difficult, very difficult. I found it out the hard way when I had to write my first research paper. Reading a paper is very easy but writing something that you have done (without copying and pasting from similar papers (BTW also called plagiarism)) is very difficult and you have to be very innovative but at the same time sound technical, clear and interesting. After I actually wrote the paper (which took me and my colleague around 6 months), the way I look at the papers completely changed. Now, when I see certain portions of the published peer reviewed papers, I know that this section was completely due to the pressure of the reviewers, this section was because the authors wanted to highlight certain aspects (and hide some others). So, now you would be thinking, why have I started this rant of mine? Good question! It is because there are certain things that have really annoyed me about the papers which might be inherent problems of the field. But still annoying. Specifically, I am upset about some of the practices people do and the way they write their methods section. 1. Talaraich Co-ordinates. I do respect the idea and effort that was put in making the Talaraich atlas. But, now I believe if you really want to confuse people and don’t want your study to be replicated, you use these co-ordinates. I know that there would be people who would disagree with me, but I guess this is my opinion. First, there are well known issues with the Talaraich atlas (from a single subject etc.). A lot has been written about it, so I won’t elaborate here. Second of all, there is no single accepted way of transforming it into MNI. There are ways - Matthew Brett’s script and lot of others (see http://www.sdmproject.com/utilities/?show=Coordinates). But mostly they are empirical and approximations. So, now I just get annoyed when I see Talaraich coordinates. 2. Cluster level correction: The field is well aware of the issues of multiple comparison and a lot has been said – some in a funny way and some in more serious ways. However, the issues has now taken a different direction and incorrect or less understood ways of correction have come up. One of them being the Alphasim/ monte carlo simulation method which is often poorly described/incorrectly used in the studies (see Bennett, Wolford, & Miller, 2009). I am not saying that the method is wrong. But the way it is written is often unclear (probably even to the authors). This leads to misleading statements like ‘The results are corrected for multiple comparisons’ in the papers when they are corrected at cluster level and not on the voxel level. Also, FDR voxel level has been proven not to be appropriate for inference on peaks or regions of activation than the cluster level FDR (Chumbley & Friston, 2009). SPM software has already stopped showing results at FDR voxel level. However, FDR voxel level are still prevalent in the papers. 3. Results on Brain slices. I know that a lot of people will disagree with me and say that the way of presenting your fmri results is upto you. But I would argue that if you really like your results on brain slices, put them in supplementary or next to a whole brain image. What brain slices do, is that they let u see only what the authors want you to see which I believe is against the idea of a results section. The results section should have all the results as it is without the interference from the authors – their way of thinking or their beliefs. I know that this might be difficult to take out from the paper, but we can try with results that just don’t show one slice of brain! So now, whenever I see fmri results on a brain slice, I start believing that there were probably some activations, that the authors didn’t want me to see and more often than not, when I see the results table, there are other regions which are activated. Lot of times hiding other activations might not be the aim of the authors but to showcase specifically the activations, however, often what this leads to is a bias in the mind of readers (quite frankly how many of us go through the tables for list of areas activated in each task unless we are looking for co-ordinates or specific regions). The reason I stress on this is because, what this eventually leads to is strong ideas in the field about the task and if the paper is influential enough, to studies that tend to approve the regions in the studies forming a vicious circle questioning studies that show other regions activated in similar tasks. 4. Short methods sections. This is very true of some journals that simply think that methods are not so important, it’s the results that are critical. I believe that methods are almost as important as the methods. For example, there are multiple ways to make a something that looks like a cake, but all of them might not be edible. 5. Previously used methods: The sure shot way to ensure that your results are never replicated ever – “The protocol previously used in the papers (1 2 also see 3 ) was used for the data”. Most often the papers 1,2 as well as 3 refer to different protocols which are somewhat related to the current paper but possibly use completely different parameters (which is not very obvious). A reader only interested in the fascinating results might not really bother about them but people who would like to replicate the experiment would be sent on a treasure hunt to find what parameters were actually used. Strangely, the journals do not mind such statements. 6. Methods in books. This is another way to ensure experiments are not replicated. “The values were calculated using the criteria X (reference to a book)” Often this book is a textbook, not freely available and even if u manage to get your hands on it, you are on another journey to hunt the location of the criteria in the book. The reporting of methods is improving however, with efforts like pre-registration of studies and other practices of reporting results. Hopefully, this will improve further in future.
0 Comments
It may not be a great achievement, not even an achievement at all. But it is a sort of moral booster. I guess there have to be events of personal joy where you take a rest from the depression of the research work and enjoy a bit of success (how small it may be). The extension is a work of my getting fed up of the laborious work of using aal (Automated Anatomical Labeling atlas) every time I had to find out what regions are activated in my activation map (although I have become better than before in distinguishing Prefrontal from Parietal!! now). The extension although was written in 2012 but I was too lazy to make documentation or make it organized (nor that it is organized right now). However, it's better now and I feel that atleast someone might find it useful to automatically batch aal results and print a copy for later retrieval.
Well, I almost forgot to add the link : https://github.com/ateshkoul/NeuroImaging/tree/master/spm/batch_aal or you can download it as a zip here: https://github.com/ateshkoul/NeuroImaging/blob/master/spm/batch_aal.zip The script is on github. I believe all our programs and scripts should be there for everyone to use and modify. The extension is easy to use and install: 1. If you have the zip, extract the contents to a folder batch_aal 2. Copy it to your '....spm/toolbox/' folder. (if on windows something like 'c:\Programs\spm\toolbox\batch_aal') 3. Use it from SPM -> Tools -> Batch AAL. Thats it!!! |
About meI am a neuroscientist working on social cognition. (I was told not to be fancy.) Archives
June 2016
Categories |