AI vs. Tumor Board Recommendations for Pancreatic Cancer Postoperative Care

A pilot study from October 2020 to December 2023 developed an AI model using clinical data from 67 pancreatic cancer patients discussed in multidisciplinary tumor boards. The model's recommendations were compared against expert decisions for 15 independent patients from 2024, showing an 80% concordance rate. Author Samet Yigman published the findings under a CC-BY-4.0 license.

Use Cases

• Benchmark AI clinical decision models based on concordance rates with expert panels.
• Study discrepancies between AI and human expert recommendations based on case analysis.
• Develop supportive tools for postoperative management based on clinical guidelines and patient data.

Strengths

• Direct comparison of AI recommendations with expert tumor board decisions for 15 independent cases.
• Reported Cohen's kappa coefficient of 0.625 provides a statistical measure of agreement.
• Clinical data spans from October 2020 to December 2023, providing a multi-year basis for model development.

Limitations

• Dataset is very small at 28.0 KB, indicating limited scope and sample size.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

figshare

Collection Method

Clinical data from patients discussed in multidisciplinary tumor boards, combined with treatment guidelines.

Time Range

October 2020 to December 2023 for model development; independent cohort from 2024.

Freshness

Last updated 2026-06-01 04:15:13; freshness should be verified.