In this paper, we explore the use of video analysis for detecting asthma in horses. We develop a single-camera model to identify asthma symptoms based on the movements of the nostrils and abdomen, body parts selected based on recommendations from veterinary experts. This model provides ambulatory vets with diagnostic and follow-up support in situations where invasive clinical procedures, such as respiratory endoscopy, bronchoalveolar lavage, and thoracic radiography, typically conducted in a hospital setting, cannot be performed. We gather a dataset with video recordings of asthmatic and healthy horses, and we propose two different methods for detecting asthma: one based on features extracted from segmented images; the other incorporating image subtraction with transfer learning from pre-trained image classifiers. Various classifiers and image subtraction techniques are evaluated in those two methods. Our best-performing model achieves an average accuracy of 89% by using nostril data alone.