✅️Potential Impact:
• faster TAT
• avoid a portion of rapid tests
• inc tissue availability fot NGS
• dec need for re biopsies
Ultimately leads to ➡️ improve cancer care
#ESMOAI25
How can we use AI to develop precision biomarkers by Gabriele Campanella
• Computational biomarker : less exhaustive (tissue) and shorter turn around time ⏩️ an ideal test indeed
#ESMOAI25
• role of AI in drug development is huge ➡️ high potential esp in different phases of clinical trial implementation (but might still be far to date)
"The future is in science of science"
@marinagarassino#ESMOAI25
•Clinical trials are moving so fast, but knowledge cannot keep up
• another challenge : less no. Of patients eligible for targeted therapy trials
#ESMOAI25
AI and Medical Education Summary
✅️ AI tutors offer opportunities and personalize learning
✅️ #GenAI can inc efficiency in preparing teaching materials
✅️ newer models are more trustworthy but critical review is stil a MUST
#ESMOAI25
Use of generative AI for teaching: How are we educationg doctors of the future? By Ramona Woitek
Utility of AI in medical education:
• generation of clinical vignettes ➡️ but it is not perfect, still need to check for errors (spelling, etc)
#ESMOAI25
✔️In summary
AI literacy roadmap for Oncologists
Start with awarenes and foundational knowledge. Thankful to @myESMO for this inaugural conference 🧬🤖
#ESMOAI25
AI literacy for Oncologists by Anita Grigoriadis
Initial concerns of clinicians re: AI
• trust issues that include privacy concerns and AI being a "black box"
#ESMOAI25
What medical doctors should know about AI by Joost Huiskens
Challenge: more than 95% of health data is unstructured multimodal data
Microsoft is now working of tapping multiple agents in AI
Ex #MedImageParse#ESMOAI25
Conclusion:
• absence of DeepGrade high clusters constitutes an independent prognostic marker ➡️ identifying low-risk patients
• study highlights that spatial distribution of AI-markers requires stronger attention (prospective studies underway)
#ESMOAI25
Identification of a very low risk subgroup in breast cancer using a combination of spatial representation and AI-derived prognostic markers by Constance Boissin
#ESMOAI25
• tumor heterogeneity is known to be involved in increased proliferation of the tumors
• intermediate risk pts: ER+/HER2- (training set)- 12 year survival:
❗️ high risk patients: 93%
✅️ low risk patients: 97.2% (p <0.001)
#ESMOAI25
In conclusion
- little experience so far in oncology beyond image-based AI
- AI has the potential to facilitate and amplify human interactions
- help & guide patients and communicate on recommendations
- follow the recently published ELCAP guidelines
#ESMOAI25
❗️The downside of AI in HCP-Patient Interactions
- shift attention away from the patient
- undermines clinical intuition and trust
- create a 'data power imbalance'
- could accelerate burnout instead of relieving it
- erodes moral dimension of medicine
#ESMOAI25