10083748End-to end AI-assisted workflow for PSMA PET/CT reportingClosedCollaborative R&DInnovate UKProstate specific membrane antigen (PSMA)-targeted imaging agents are a relatively new family of radiotracers used for imaging prostate cancer patients with Positron Emission Tomography (PET) for diagnostics, therapy selection and response assessment and a major growth area in clinical practice. Whilst PSMA PET/CT offers high sensitivity in detecting metastatic disease, it is also subject to potential interpretative pitfalls and equivocal findings, with uptake in regions of the image which does not reflect prostate cancer, including normal biodistribution in several organs, excretion via the urinary tract, expression in a variety of benign and other malignant conditions, and potential for non-specific bone uptake. A number of standardised reporting systems have been proposed in radiology that aim to standardize the interpretation and reporting of findings from a specific imaging modality, including for oncologic PET/CT. The focus of this project is on developing and evaluating an end-to-end AI-assisted PSMA PET/CT reporting tool, in order to assist clinicians in better stratifying patients to the appropriate treatment, together with creating the tools for standardisation and automatic reporting to assess efficiency gains and consistency of reporting quality. The solution will be built on the functionality of Mirada XD diagnostic software which is used in thousands of hospitals worldwide in clinical routine. The project will involve a collaboration between Mirada Medical, and clinicians at Leeds Teaching Hospitals, University Hospitals Bristol and Newcastle upon Tyne Hospitals NHS Foundation Trust. It will build on our recent work on AI-assisted FDG-PET/CT reporting applied to lesion detection and segmentation in lymphoma patients (part of the NCIMI research program, InnovateUK104688). The project focusses on oncology and novel digital and data-enabled tools to more accurately diagnose conditions and stratify patients to the most appropriate treatment. It will include prototyping, evaluation of the solution within a clinical environment, and demonstration of clinical utility and effectiveness. There has been much public discussion about the risks and benefits of using AI to support radiology processes. We will pay careful attention to the role that AI plays within our solution. We will use AI most where it will be supporting the radiologist by collating and organising information, rather than mimicking the core role of the radiologist. For these supportive tasks it will be easier for the radiologist to check that the AI has worked correctly, which will also make it simpler to prove the tool is safe to use on patients. We will use AI alongside more conventional software programs, deciding which is most appropriate for each part of the process. Technical and clinical results will be published and presented at international conferences.