It is not possible to run your code.
The filter may be working, however note that it is only order 2. That may be appropriate for some designs (for example an elliptic filter), however it may not be enough for a Butterworth design. Each filter design has a function to estimate the appropriate order and return other values, for Butterworth filters it is buttord. Using it might help with your filter design, and returning the zero-pole-gain output from butter and then using zp2sos and then filtfilt might also produce better results. So long as you are designing the filters in a loop, an equivalent approach could be to initially filter the signal with the first filter and then the output of that filter with the subsequent filters. It might be easier to use the bandstop function, actually (include 'ImpulseResponse','iir'), unless you have to design your own filters.
Onee problem however is how you are analysing the results. The spectrogram arguments are first the signal, second the window length, and third the overlap. In your call to it, the overlap is equivalent to the sampling frequency. (The sampling frequency is actually the fifth argument to spectrogram.) With a smaller overlap (for example using the default value), and using square brackets [] for the arguments you do not want to specify (in that event they become the function default values), the filtered results might be mor to your liking.
