Artificial Intelligence and Radiologists at Prostate Cancer Detection in MRI: The PI-CAI Challenge

The PI-CAI challenge aims to validate the diagnostic performance of artificial intelligence (AI) and radiologists at clinically significant prostate cancer (csPCa) detection/diagnosis in MRI, with respect to histopathology and follow-up (≥ 3 years) as reference. The study hypothesizes that state-of-the-art AI algorithms, trained using thousands of patient exams, are non-inferior to radiologists reading bpMRI. As secondary end-points, it investigates the optimal AI model for csPCa detection/diagnosis, and the effects of dynamic contrast-enhanced imaging and reader experience on diagnostic accuracy and inter-reader variability.

Study Overview

• Status: Completed
• Conditions: Prostate Cancer
• Intervention / Treatment: Diagnostic test: Histopathology and Magnetic Resonance Imaging; Diagnostic test: Histopathology and Magnetic Resonance Imaging with Follow-Up

Detailed Description

Prostate cancer (PCa) is one of the most prevalent cancers in men worldwide. One million men receive a diagnosis and 300,000 die from clinically significant PCa (csPCa) (defined as ISUP≥2), each year, worldwide. Multiparametric magnetic resonance imaging (mpMRI) is playing an increasingly important role in the early diagnosis of prostate cancer, and has been recommended by the European Association of Urology (EAU), prior to biopsies. However, current guidelines for reading prostate mpMRI (i.e. PI-RADS v2.1) follow a semi-quantitative assessment that mandates substantial expertise for proper usage. This can lead to low inter-reader agreement (<50%), sub-optimal interpretation and overdiagnosis.

Modern artificial intelligence (AI) algorithms have paved the way for powerful computer-aided detection and diagnosis (CAD) systems that rival human performance in medical image analysis. Clinical trials are the gold standard for assessing new medications and interventions in a controlled and comparative manner, and the equivalent for developing AI algorithms are international competitions or "grand challenges", where increasingly large datasets are released to public to solve clinically relevant tasks with AI. Grand challenges can address the lack of trust, scientific evidence and adequate validation among AI solutions, by providing the means to compare algorithms against each other using common datasets and a unified experimental setup.

PI-CAI (Prostate Imaging: Cancer AI) is an all-new grand challenge, with over 10,000 carefully-curated prostate MRI exams to validate modern AI algorithms and estimate radiologists' performance at csPCa detection and diagnosis. Key aspects of the study design have been established in conjunction with an international, multi-disciplinary scientific advisory board (16 experts in prostate AI, radiology and urology) -to unify and standardize present-day guidelines, and to ensure meaningful validation of prostate-AI towards clinical translation.

The 2022 edition of PI-CAI will focus on validating AI at automated 3D detection and diagnosis of csPCa in bpMRI. PI-CAI primarily consists of two sub-studies:

• AI Study (Grand Challenge): An annotated multi-center, multi-vendor dataset of 1500 bpMRI exams (including their clinical and acquisition variables) is made publicly available for all participating teams and the research community at large. Teams can use this dataset to develop AI models, and submit their trained algorithms (in Docker containers) for evaluation. At the end of this open development phase, all algorithms are ranked, based on their performance on a hidden testing cohort of 1000 unseen scans. In the closed testing phase, organizers retrain the top-ranking 5 AI algorithms using a larger dataset of 7500-9500 bpMRI scans (including additional training scans from a private dataset). Finally, their performance is re-evaluated on the hidden testing cohort (with rigorous statistical analyses), to determine the top 3 AI algorithms for automated 3D detection and diagnosis of csPCa in bpMRI (i.e. the winners of the grand challenge).
• Reader Study: 50+ international prostate radiologists perform a reader study using a subset of 400 scans from the hidden testing cohort. For each case, radiologists complete their assessments in two rounds. At first, using clinical and acquisition variables plus bpMRI sequences only, enabling head-to-head comparisons against AI trained on the same. And then, using clinical and acquisition variables plus full mpMRI sequences, enabling comparisons between AI and current clinical practice (PI-RADS v2.1). Overall, the goal of this study is to estimate the performance of the average radiologist at detection and diagnosis of csPCa in MRI.

In the end, PI-CAI aims to benchmark state-of-the-art AI algorithms developed in the grand challenge, against prostate radiologists participating in the reader study -to evaluate the clinical viability of modern prostate-AI solutions at csPCa detection and diagnosis in MRI.

• Study Type: Observational
• Enrollment (Actual): 10207

Contacts and Locations

Study Locations

• Netherlands
  • Gelderland
    • Nijmegen, Gelderland, Netherlands, 6525 GA
      • Radboudumc

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: 14 years and older (Adult, Older Adult)
• Accepts Healthy Volunteers: Yes

• Sampling Method: Non-Probability Sample

Study Population

All patient exams are of men suspected of harboring csPCa, with elevated levels of prostate-specific antigen (≥ 3 ng/mL) and/or abnormal findings on digital rectal exam, and without a history of treatment or any prior positive histopathology (ISUP ≥ 2) findings. Patients underwent prostate MRI, and were primarily examined at one of three Dutch centers (Radboud University Medical Center, Ziekenhuisgroep Twente, University Medical Center Groningen) or one Norwegian center (Norwegian University of Science and Technology) during regular clinical routine, between 2012-2021.

Description

Inclusion Criteria:

• Men suspected of harboring csPCa, with elevated levels of prostate-specific antigen (≥ 3 ng/mL) and/or abnormal findings on digital rectal exam, who subsequently underwent prostate MRI.

Exclusion Criteria:

• Patients who opted-out or did not give permission to reuse clinical data.
• Patients with a history of prior prostate treatment.
• Patients with a history of prior positive csPCa findings in histopathology (ISUP ≥ 2).
• Patients whose prostate MRI exhibit severe artifacts (e.g. heavy warping due to rectal air, metal artifacts from hip prostheses, heavy motion blur), thereby impeding their usage.
• Patients, whose positive histopathology findings (ISUP ≥ 2) cannot be reliably localized on MRI (e.g. MRI-invisible lesions, systematic biopsy diagnostic reports with ambiguous, "random" or missing location information).

Study Plan

How is the study designed?

Design Details

• Observational Models: Cohort
• Time Perspectives: Retrospective

Number of groups / cohorts

4

Cohorts and Interventions

Group / Cohort	Intervention / Treatment
Public Training and Development Set (1500 cases); Available for all participants and researchers, to train and develop AI models. Includes multi-vendor (Siemens Healthineers, Philips Medical Systems) prostate bpMRI cases from three Dutch centers (Radboud University Medical Center, Ziekenhuisgroep Twente, University Medical Center Groningen), acquired between 2012-2021. All data is fully anonymized and made available under a non-commercial CC BY-NC 4.0 license. Includes 328 cases from the PROSTATEx challenge (prostatex.grand-challenge.org). Imaging data has been released via: zenodo.org/record/6624726 (DOI: 10.5281/zenodo.6624726). Lesion annotations of csPCa have been released and are maintained via: github.com/DIAGNijmegen/picai_labels.	Diagnostic test: Histopathology and Magnetic Resonance Imaging; Reference standard establishes histologically-confirmed (ISUP ≥ 2) cases of csPCa as positives, and histopathology- (ISUP ≤ 1) or MRI- (PI-RADS ≤ 2) confirmed cases of indolent PCa or benign tissue as negatives.
Private Training Set (7500-9500 cases); Used exclusively by the organizers to retrain the top-ranking 5 AI algorithms, with large-scale data. Includes multi-vendor (Siemens Healthineers, Philips Medical Systems) prostate bpMRI cases from three Dutch centers (Radboud University Medical Center, Ziekenhuisgroep Twente, University Medical Center Groningen), acquired between 2012-2021.	Diagnostic test: Histopathology and Magnetic Resonance Imaging; Reference standard establishes histologically-confirmed (ISUP ≥ 2) cases of csPCa as positives, and histopathology- (ISUP ≤ 1) or MRI- (PI-RADS ≤ 2) confirmed cases of indolent PCa or benign tissue as negatives.
Hidden Validation and Tuning Cohort (100 cases); Used for a live, public leaderboard that enables AI model selection and tuning throughout the open development phase of the challenge. Includes multi-vendor (Siemens Healthineers, Philips Medical Systems) prostate bpMRI cases from three Dutch centers (Radboud University Medical Center, Ziekenhuisgroep Twente, University Medical Center Groningen), acquired between 2012-2021, that remain fully hidden throughout the course of the challenge.	Diagnostic test: Histopathology and Magnetic Resonance Imaging with Follow-Up; Reference standard establishes histologically-confirmed (ISUP ≥ 2) cases of csPCa as positives, and histopathology- (ISUP ≤ 1) or MRI- (PI-RADS ≤ 2) with follow-up (≥ 3 years) confirmed cases of indolent PCa or benign tissue as negatives.
Hidden Testing Cohort (1000 cases); Used to benchmark AI, radiologists, and test all hypotheses at the end of the PI-CAI challenge. A subset of 400 cases from this cohort is used to facilitate the PI-CAI: Reader Study. Includes multi-vendor (Siemens Healthineers, Philips Medical Systems) internal testing data (unseen prostate bpMRI cases from three seen Dutch centers {Radboud University Medical Center, Ziekenhuisgroep Twente, University Medical Center Groningen}) and external testing data (unseen prostate bpMRI cases from one unseen Norwegian center {Norwegian University of Science and Technology}), acquired between 2012-2021.	Diagnostic test: Histopathology and Magnetic Resonance Imaging with Follow-Up; Reference standard establishes histologically-confirmed (ISUP ≥ 2) cases of csPCa as positives, and histopathology- (ISUP ≤ 1) or MRI- (PI-RADS ≤ 2) with follow-up (≥ 3 years) confirmed cases of indolent PCa or benign tissue as negatives.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
AI vs Radiologists from Reader Study	Diagnostic performance of the top 5 AI models from the grand challenge and 50+ radiologists from the reader study, at csPCa detection/diagnosis in prostate bpMRI, with respect to histopathology and MRI with follow-up (≥ 3 years) as reference, to assess the clinical viability of present-day AI solutions.	6 months
AI vs Radiologists from Clinical Routine	Diagnostic performance of the top 5 AI models from the grand challenge and the historical reads of radiologists from clinical routine, at csPCa detection/diagnosis in prostate bpMRI, with respect to histopathology and MRI with follow-up (≥ 3 years) as reference, to assess the clinical viability of present-day AI solutions.	6 months

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
AI vs AI	Diagnostic performance of the top 5 AI models from the grand challenge, at csPCa detection/diagnosis in prostate bpMRI, with respect to histopathology and MRI with follow-up (≥ 3 years) as reference, to deduce the optimal AI model architecture for this given task.	6 months
Radiologists vs Radiologists from Reader Study	Diagnostic performance and inter-reader variability of 50+ radiologists from the reader study, at csPCa detection/diagnosis in prostate mpMRI, with respect to histopathology and MRI with follow-up (≥ 3 years) as reference, to deduce the effects of dynamic contrast-enhanced imaging and reader experience.	6 months

Collaborators and Investigators

• Sponsor: Radboud University Medical Center
• Collaborators: Norwegian University of Science and Technology; University Medical Center Groningen; Ziekenhuisgroep Twente
• Investigators: Principal Investigator: Henkjan Huisman, PhD, Radboud University Medical Center

Publications and helpful links

Helpful Links

• Official website.

Study record dates

Study Major Dates

• Study Start (Actual): February 1, 2022
• Primary Completion (Actual): June 1, 2023
• Study Completion (Actual): November 1, 2023

Study Registration Dates

• First Submitted: August 3, 2022
• First Submitted That Met QC Criteria: August 3, 2022
• First Posted (Actual): August 5, 2022

Study Record Updates

• Last Update Posted (Estimated): November 17, 2023
• Last Update Submitted That Met QC Criteria: November 16, 2023
• Last Verified: July 1, 2022

More Information

Terms related to this study

• Keywords: Prostate Cancer; Magnetic Resonance Imaging; Artificial Intelligence; Computer-Aided Detection and Diagnosis
• Additional Relevant MeSH Terms: Neoplasms; Urogenital Neoplasms; Neoplasms by Site; Genital Neoplasms, Male; Prostatic Diseases; Urogenital Diseases; Male Urogenital Diseases; Genital Diseases, Male; Genital Diseases; Prostatic Neoplasms
• Other Study ID Numbers: CMO2016-3045-Project-20011

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: YES
• IPD Plan Description: To facilitate open and transparent science, our end-to-end study protocol and our source code for preprocessing prostate MRI data archives, training baseline diagnostic AI models, evaluating lesion detection/diagnosis performance, and implementing statistical tests for AI/radiologists vs AI/radiologists comparisons, have been publicly released. Furthermore, a fully-anonymized dataset of 1500 prostate bpMRI scans from the PI-CAI challenge, and their outcomes, have been released to promote further research.
• IPD Sharing Time Frame: Individual Participant Data Set (PI-CAI: Public Training and Development Set), Study Protocol, Statistical Analysis Plan (SAP) and Analytic Code has been shared with all participants of the PI-CAI challenge and the research community at large, towards the start of the challenge (June 2022). Clinical Study Report (CSR) will be released in the form of multiple publications after the completion of the challenge (tentatively May 2023). All of the aforementioned IPD will remain publicly accessible perpetually.
• IPD Sharing Access Criteria: Please refer to the "References" section of this protocol.
• IPD Sharing Supporting Information Type: STUDY_PROTOCOL; SAP; ANALYTIC_CODE; CSR

Study Data/Documents

1. Individual Participant Data Set
   Information identifier: 10.5281/zenodo.6624726
   Information comments: Imaging for the PI-CAI: Public Training and Development Dataset, containing 1500 fully-anonymized prostate bpMRI scans from 1476 patients, acquired between 2012-2021, at three Dutch centers (Radboud University Medical Center, University Medical Center Groningen, Ziekenhuisgroep Twente).
2. Study Protocol
   Information identifier: 10.5281/zenodo.6667655
   Information comments: Preregistration of the PI-CAI challenge study design, in compliance with BIAS reporting guidelines (https://www.equator-network.org/reporting-guidelines/bias-transparent-reporting-of-biomedical-image-analysis-challenges/).
3. Analytic Code
   Information comments: Source code for preprocessing prostate MRI data archives.
4. Analytic Code
   Information comments: Source code for training baseline diagnostic AI models.
5. Analytic Code
   Information comments: Source code for evaluating csPCa detection and diagnosis performance, and performing all statistical tests with respect to the same.
6. Individual Participant Data Set
   Information comments: Annotations for the PI-CAI: Public Training and Development Dataset, containing basic clinical and acquisition variables, csPCa annotations and outcomes for 1500 fully-anonymized prostate bpMRI exams from 1476 patients, acquired between 2012-2021, at three Dutch centers (Radboud University Medical Center, University Medical Center Groningen, Ziekenhuisgroep Twente).

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No