Introduction. Larger one-time values of the spatial QRS-T angle (SA) are associated with risk. However, experience how serial changes in SA (ΔSA) should be interpreted is lacking. Even within normal limits, any change in SA likely signifies electrical remodelling. Possible pseudo-normalization (masking of one pathology by another one with opposite effect) explains that in patients with acquired larger SA values, new emerging pathology can decrease SA again, while the clinical status deteriorates. Aim of the current study was to assess the impact of choosing either ΔSA or |ΔSA| as one of a set of serial ECG difference features that constitute the input for our deep learning serial-ECG classifier (DLSEC). Methods. We trained and tested our DLSEC to detect emerging pathology in serial standard ECGs of post-myocardial infarction patients who developed heart failure, and of stable angina patients in whom acute ischemia was created during elective percutaneous transluminal coronary angioplasty. Serial ECGs of control patients (who remained stable) were added to facilitate learning and testing, thus creating a “heart failure database” (48/81 cases/controls) and an “ïschemia database” (82/398 cases/controls). Either ΔSA or |ΔSA| were among 13 features that represented the serial-ECG differences. Neural network structures/coefficients were dynamically generated during learning, and testing receiver operating characteristics (ROCs) and corresponding areas under the curve (AUCs) were computed. Results. Testing AUCs were 72% (P<10-3) and 84% (P<10-6) for the heart failure database and 77% (P<10-11) and 83% (P<10-11) for the ischemia database, for ΔSA or |ΔSA| among the features, respectively. Discussion and conclusions. The DLSECs performed well in both emerging heart failure and acute ischemia; |ΔSA| among the features was superior to ΔSA in discriminating cases and controls. Our study reinforces the concept that any change in SA, irrespective the sign, indicates a worsening situation. Further corroboration requires studies in other clinical situations.