the typical cell of this eq is good but the problem is, like seem from this image, that all stages (cells and local summing ampplifiers ) are putted in series, so the audio signal go into too stages.

The eq can be improved for having all cells in parallel config.

This is very debatable.

Whatever the configuration (series or parallel), there will be he same number of biquads for a given number of bands. So let's take them out of the comparison.

In the series topology, there will be as many gain/boost opamps as there are bands. The total degradation of signal produced by these N stages will be: quadratic noise addition (4 bands => +6dB) and THD (anything between +6 and +12dB).

In the parallel topology, there is only one opamp, but its noise gain increases LINEARLY with the number of bands. So the degradation of signal increases by 12dB for a 4 band design.

In addition, in parallel topology, two adjacent bands interact in an "unpredictable" way. See attached curves. IMO, the series topology does what it should, i.e. the resulting curve is exactly the sum of the individual bands - very instinctive.