The number of people suffering from lung-related diseases accounts for a high percentage. Patients with lung diseases are of all ages and genders, especially children and the elderly with low resistance. To accurately diagnose lung diseases, doctors often diagnosis based on chest X-ray images. Chest X-ray is one of the most widely used techniques in medicine, helping doctors to assess and detect lung lesions. However, it is difficult to accurately and quickly diagnose lung lesions with a large number of patients taking the time of doctors in reading test results. Therefore, in this paper, we propose a novel method to accurately detect lung lesions by constructing deep learning networks in Spark parallel and distributed computing environments. Experimental results show that the proposed method achieves an accuracy of 96% with the training time reduced by 57% compared to training in a stand-alone computing environment. This supports doctors in quickly making a preliminary diagnosis of lung lesions for timely treatment.