I agree with the implicit assumption in your question. When trying to evaluate whether there is a serious problem, in any area, we should rely on data. These stories, heartbreaking as they are, are just anecdotes. Thus, lacking proper data, I can't tell whether there is a systemic problem with police brutality in the US: I can't answer your question. I'm sure such statistics do exist, but since I'm not from the US, it's not exactly something I tend to pick up on.
However, I am somewhat familiar with the dynamics involved with issues like this. There are usually some few exceptions that stand out enough to be picked up in the media, but most don't. I'd guess that there are an order of magnitude more cases that are either not severe enough to be reported on, or too common to be interesting, but are still beyond what a society would consider ideal, were they to come to light.
So, let me rephrase: do you have trustable statistics showing that police brutality is not systemic? How about all those 'exceptional' cases where the system very much seems to work against correcting itself by protecting the offending police officers?
See, I'm not surprised that in a ~300M nation there are lots of cases of authority acting violently and that we only hear about those cases where people are treated badly - but what's worrying to me is how bad the 'exception handling' usually appears to work. If your program throws an exception every 10'000 users, but the exception leads to the user being electro-shocked, jailed and possibly loosing his job, wouldn't you throw in some additional exception handling?