Cystoscopy has limited ability to detect ureteral injury caused by electrosurgical device lateral thermal spread during gynecologic procedures. We assessed the feasibility of endoscopic optical coherence tomography (OCT) to detect electrothermal ureteral damage. A convolutional neural network (CNN) was developed to automate detection and to provide higher-level insights into characteristic features of electrothermal injury on OCT images. Bipolar electrothermal energy was externally applied to nine freshly excised porcine ureters. Three segments of each ureter were treated for 5 seconds at low (16 watts), medium (26 watts), and high (36 watts) powers (total n = 27). Volumetric OCT images of each lesion were acquired using a swept source OCT laser endomicroscopy system (Ninepoint NVisionVLE). Lesions were compared to untreated controls on histology. OCT datasets were visually inspected for characterization of normal and electrothermally injured tissue architecture. Based on ground-truth interpretation, labelled images were used to train and validate the machine learning algorithm. The effect of power on lesion length as measured with OCT was compared using a one-way analysis of variance. Transmural electrothermal injury was detected in all histology-matched lesions (23/23, 100%) on OCT images. The mean lesion size on OCT was 0.36 ± 0.2 cm, 0.43 ± 0.1 cm, and 0.70 ± 0.3 cm for low, medium, and high powers, respectively (p=0.017). The CNN successfully identified all lesions but with several false positives, including artifacts from tissue dissections, needles, and air pockets. Endoscopic OCT could fulfill an unmet clinical need for the timely detection of electrothermal ureteral damage.