FTS 15 - Day 3 - AI of Pathology Specimens to Predict Outcome in Prostate Cancer

Описание: FTS 15 - Day 3 - AI of Pathology Specimens to Predict Outcome in Prostate Cancer

Dr. Peter Carroll discusses the limitations of current risk assessment models for prostate cancer, particularly the NCCN guidelines, which he argues are outdated and don't accurately reflect modern biopsy practices. He emphasizes the importance of:
• Histologic Subtype: Recognizing the significance of histologic subtypes, particularly unfavorable features like large or infiltrative Gleason Pattern 4, in predicting outcomes.
• AI in Pathology:
o AI algorithms can improve the accuracy and consistency of Gleason grading, addressing inter-pathologist variability.
o AI can potentially predict outcomes, potentially reducing the need for other prognostic tests like genomic profiling.
• Impact of AI on Pathology: Dr. Carroll acknowledges that AI may eventually lead to a reduction in the need for human pathologists, but he believes that AI will ultimately augment the role of pathologists by improving accuracy and efficiency.
• AI in Treatment Decision-Making: He discusses the potential of AI to predict treatment response, such as the need for hormonal therapy in patients receiving radiation.
• UCSF Research: Dr. Carroll highlights his own research at UCSF, including a large dataset of prostate cancer cases with extensive imaging and genomic data, which will be used to develop and validate AI models for predicting outcomes.
Key Takeaways:
• Current risk assessment models for prostate cancer have limitations and may not accurately reflect modern clinical practice.
• Histologic subtype is a crucial factor in predicting outcomes and should be routinely reported by pathologists.
• AI has the potential to revolutionize prostate cancer pathology by improving diagnostic accuracy, predicting outcomes, and potentially reducing the need for other prognostic tests.

Dr. Carroll emphasizes the importance of large datasets and multidisciplinary collaboration to develop and validate AI models for prostate cancer.