In post-ischemic ventricular tachycardia treatment, intracardiac electrogram (EGM) fragmentation is widely used to identify abnormal ventricular potentials (AVPs). However, the conventional time-domain approach may not fully reflect pathological dynamics of arrhythmogenic substrate. In this work, we explore fragmentation analysis in the time-frequency domain, through Continuous Wavelet Transform (CWT). We used the ARGO dataset (1716 EGMs @1000 Hz, 2.5 s each), annotated by a consensus of experts. A CWT decomposition was applied on each EGM up to 40th scale, and fragmentation was evaluated both at single-scale level (fr_sCWT) and as a cumulative index (fr_CWT) summing single-scale fragmentations. The optimal configuration was selected based on AUC analysis and compared against time-domain fragmentation (fr_time). The optimal fr_CWT, summing the first 19 fr_sCWT, achieved an AUC of 0.98 with Hellinger distance (H) of 0.87, against an AUC of 0.91 and H of 0.62 for fr_time. These results suggest that evaluating fragmentation across CWT scales captures discriminative patterns that a time-domain count alone cannot resolve. Such a result could be valuable in the context of automated AVP detection pipelines for ablation target identification in clinical practice.