Background: In the absence of a gold standard, a panel of experts can be invited to assign a reference diagnosis for use in research. Available literature offers limited guidance on assembling and working with an expert panel for this purpose. We aimed to develop a protocol for an expert panel consensus diagnosis and evaluated its applicability in a pilot project. Methods: An adjusted Delphi method was used, which started with the assessment of clinical vignettes by 3 experts individually, followed by a consensus discussion meeting to solve diagnostic discrepancies. A panel facilitator ensured that all experts were able to express their views, and encouraged the use of argumentation to arrive at a specific diagnosis, until consensus was reached by all experts. Eleven vignettes of patients suspected of having a primary neurodegenerative disease were presented to the experts. Clinical information was provided stepwise and included medical history, neurological, physical and cognitive function, brain MRI scan, and follow-up assessments over 2 years. After the consensus discussion meeting, the procedure was evaluated by the experts. Results: The average degree of consensus for the reference diagnosis increased from 52% after individual assessment of the vignettes to 94% after the consensus discussion meeting. Average confidence in the diagnosis after individual assessment was 85%. This did not increase after the consensus discussion meeting. The process evaluation led to several recommendations for improvement of the protocol. Conclusion: A protocol for attaining a reference diagnosis based on expert panel consensus was shown feasible in research practice.