Scientific Advisory Board
The Scientific Advisory Board (SAB) was formed to review the study design and outcomes of the PI-CAI challenge. In turn, they will recommend future guidelines for prostate-AI development and validation at the end of the challenge, through a Delphi consensus study (including several other experts in the domain).
For validation of AI and human-reader performance, a strong reference standard for the presence or absence of clinically significant cancer, representative testing cohort distributions, and frameworks for bias-free performance estimation, are crucial. To substantiate any outcomes or conclusions derived from the PI-CAI challenge, these factors must be discussed and agreed upon with multi-disciplinary experts in this domain. Lessons learned through the PI-CAI challenge can help unify present-day guidelines, and ensure meaningful validation of prostate-AI towards clinical translation (Reinke et al., 2022).
Disclaimer(s):
- SAB members were not allowed to review any individual patient exam (images, clinical metadata, annotations) in the hidden validation and testing cohorts. Furthermore, information regarding the exact csPCa prevalence and the overall distribution of PI-RADS and ISUP lesions in these cohorts were redacted during the review process.
- Prostate-AI teams and radiologists from the same institutes as SAB members (including SAB members themselves), are allowed to participate in the PI-CAI challenge, as they carry no additional knowledge or advantage over regular participants. All information shared with the SAB has been made publicly available for all participants through the PI-CAI challenge website and our preregistered end-to-end study design:
zenodo.org/record/6667655, clinicaltrials.gov/NCT05489341.
zenodo.org/record/6667655, clinicaltrials.gov/NCT05489341.
Artificial Intelligence
Radiology
