Abstract Background Conventional methods for the estimation of infarct size by late-enhanced cardiac magnetic resonance imaging use analyzes from each slice which in turn is added to generate a volume. We present a novel method based on machine learning, called texture based probability mapping (TPM). TPM is based upon expressing the probability of scarring in a pixel based upon analysing textural information in patches around each pixel. Purpose To explore our method and assess its utility as a tool to scar size and compare it with four established methods. Methods In 54 patients with ischemic scars, our method was compared with four 2D methods. 2 patients were excluded due to low image quality and artefacts. The most basal and apical short axis image slices were exclude due to partial volume artefacts. Myocardial infarction (MI) size was estimated manually, and TPM performed. Cardiac segments were defined as myocardial regions with pixel probabilities within a specified range. By varying the lower probability threshold while keeping the upper threshold fixed at 1, different cardiac segments were defined. Using the Sørensen-Dice coefficient the optimal probability range with the highest correlation to manual estimation of infarct size was identified. Results TPM-range 0.341–1.0 is best correlated to manual demarcation (median Dice 0.70). The method demonstrated stronger correlations between scar size and left ventricular remodeling parameters (LV ejection fraction: r=−0.731, p<0.0005; LV end-diastolic volume: r=0.641, p<0.0005; LV end-systolic volume: r=0.672, p<0.0005) compared with manual method. Conclusion Using TPM, infarct size can be measured automatically without using signal intensity as a reference value. It is without the need for manual demarcation of the scar and is less time consuming and less vulnerable to interobserver variability. It has a strong correlation with left ventricular remodeling parameters crucial in guiding further medical therapy. The use of this method is however not primarily dedicated for scar size estimation, but a tool to evaluate different properties of the tissue that cannot be visualized by the naked eye. It's an important step towards better understanding of myocardial damage pattern causing ventricular arrhythmia and SCD. Funding Acknowledgement Type of funding source: Public grant(s) – National budget only. Main funding source(s): Helse Vest