This doctoral project will develop the latest computer vision deep learning techniques to better predict outcomes of complex respiratory disease. Through image analysis of chest X-rays and CT scans, we could provide clinicians with more reliable and more objective evidence of specific disease types, and provide greater justification for what are often challenging and high-risk treatments. We currently lack adequate tools for diagnosing complex respiratory infection. Using existing methods, it can be difficult to know whether newly-identified bacteria have caused the severe lung disease, or whether they simply reflect the severe disease, acting as so-called'by-standers' having thrived in this environment. This is a problem, as (i) some infections such as those caused by Pseudomonas and Mycobacteria may require intravenous and potentially long and toxic therapy; and (ii) antibiotic usage when it is not required for treatment drives increasing drug-resistance and this remains a global emergency.