Introduction: biventricular digital twins (BDTs) offer promise for preci-sion cardiology, yet most models lack key pathological and anatomical de-tails such as transmural scar heterogeneity and the Purkinje network. Accurate transmurally heterogeneous distribution of scar and Purkinje–myocardial junctions (PMJs) are essential for simulating arrhythmia dynamics. Methods: BDTs were constructed from CMR, EGM, and ECG data, in-corporating fibre orientation, transmurally heterogeneous scar distribution (dense fibrosis and border zone defined by pixel intensity), and electrophysio-logical parameters. PMJs were generated using a novel algorithm based on Purkinje morphology and transmural extent. PMJ conductivity was calibrated by tuning it iteratively so that the simulated local activation time (LAT) at the scar-free calibration point in the basal LV endocardium matched the clini-cal LAT, using the pacing site identified from the clinical LAT map. Accura-cy was validated at additional non-calibrated locations. ventricular tachycardia (VT) induction and re-entry analysis were performed in BDTs. Results: BDTs were constructed for 3 VT patients (2 endocardial and 1 ep-icardial ablation cases). Simulated LATs at 10 structurally representative en-docardial sites showed strong agreement with clinical LATs (within 7.46 ± 3.27 ms). VT was successfully induced in all BDTs. Simulations re-produced reentry circuits that matched clinical ablation locations: reentry was inducible only from pacing sites consistent with the clinical access route (ep-icardial or endocardial). Conclusion: This pipeline enables construction of physiologically de-tailed BDTs that accurately replicate patient-specific activation and reentry patterns, supporting non-invasive VT risk assessment and guiding personal-ised ablation strategies.