I've linked to numerous papers on this subject in the past. This recent review may be the best I've seen. It includes a case report of a patient with a strong clinical prior probability of PE who, due to clinical and external factors, could not be imaged. She had several electrocardiographic features suggestive of PE. Based on clinical suspicion and the ECG features the authors treated her, with a satisfactory outcome. Later on clinical circumstances enabled her to be imaged and the diagnosis was confirmed.
There are two potential uses of electrocardiography in pulmonary embolism. One is in severity assessment. The electrocardiogram is helpful in predicting complications of PE because it detects right ventricular strain, and an emerging literature is helping define its most effective use in this area. The other use, the diagnosis of PE itself, is more problematic. No single abnormality has high specificity. Several studies, as summarized in the paper, suggest that combinations of abnormalities have high specificity and positive predictive value.