Improving Operational Machine Learning in the Pulmonary Radiology Domain Towards Clinical Adoption