We previously showed that reentry in atrial tachycardia (AT) consistently occurs in pairs - the paired loop paradigm (PLP). These pairs occur around two critical boundaries (CBs), which when connected by an ablation line consistently stops the AT. CBs can be identified by calculating the phase index (PI), which in our previous study was done using local activation times (LATs). However, LATs can be challenging to identify, especially if we want to investigate the PLP in more complex arrhythmias like atrial fibrillation. Therefore, here we extend our approach by showing that PI can also be computed with a novel signal-based method using cross-correlation. We compared both methods (LAT versus signal) by using a set of 380 spherical simulations of anatomical reentry, with varying boundary sizes and positions (201 3-boundary cases, 179 4-boundary cases). 1610 evenly placed unipolar electrograms (EGMs) were generated around each sphere. After pre-processing, we derived (1) LATs, and (2) delays between neighbors through cross-correlation. PI was calculated as the sum of (1) LAT differences or (2) correlation-delays, divided by the cycle length. Accuracy was calculated by marking boundaries as correctly annotated when the difference between PI and ground truth (1 for CB, 0 otherwise) was < 0.5. To test robustness, both methods were repeated after adding white Gaussian noise to the original EGMs, with decreasing signal-to-noise ratios (SNR = 10 dB, 5 dB, 0 dB). With no noise, both methods had 100% accuracy. For the respective SNR values, accuracy for the LAT method was 96.1%, 94.1%, and 77.4%, while for the cross-correlation method it was 100%, 98.6%, and 83.5%. In conclusion, the cross-correlation method for computing PI was validated, replacing the need to compute LATs. As noise increased, it consistently showed higher accuracy, which degraded at a lower rate, demonstrating greater robustness.