It depends on where you define ‘start’ and ‘end’.
If you want to consider the middle of the signal rather tha either peak, first use the bandpass function to elimiinate the baseline drift and high-frequency noise. (Use the fft function or the Signal Analyzer App to determine the frequency content in order to design the filter correctly.) Then, to determine the approximate zero-crossing indices, use:
zxidx = find(diff(sign(filtered_signal)));
You can then use a for loop and interp1 to determine the exact aero-crossing times if that is important.
Iff you define them by the peaks or valleys, use findpeaks,, or islocalmax or islocalmin to determine those indices.