Scribd
Upload
13 views7 pages

AI X Oncology: Ai Research That Is Shaping Cancer Care in 2025

The document outlines the impact of AI on oncology, highlighting 10 AI hallmarks across prevention, treatment optimization, and therapy innovation. It discusses advancements in AI models for cancer detection, treatment planning, and trial matching, showcasing their effectiveness compared to traditional methods. The future of AI in oncology includes the development of autonomous AI agents and the integration of multi-omics data for improved patient outcomes.

Uploaded by

Amr Abbas
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
13 views7 pages

AI X Oncology: Ai Research That Is Shaping Cancer Care in 2025

The document outlines the impact of AI on oncology, highlighting 10 AI hallmarks across prevention, treatment optimization, and therapy innovation. It discusses advancements in AI models for cancer detection, treatment planning, and trial matching, showcasing their effectiveness compared to traditional methods. The future of AI in oncology includes the development of autonomous AI agents and the integration of multi-omics data for improved patient outcomes.

Uploaded by

Amr Abbas
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

AI x Oncology

AI RESEARCH THAT IS SHAPING


CANCER CARE IN 2025

Vidith Phillips
Clinical AI Researcher
Johns Hopkins Medicine
Framework outlines 10 distinct “AI hallmarks”, across 3
domains:

Prevention & Diagnosis: screening, detection, profiling

Treatment Optimization: outcome prediction, adaptive


planning, biomarker discovery

Therapy Innovation: combo prediction, vulnerability


mining, AI-guided drug design

Authors analyzed 31 interventional trials using AI most


common in trial matching and outcome prediction

📌 Chang, TG., Park, S., Schäffer, A.A. et al. Hallmarks of artificial intelligence
contributions to precision oncology. Nat Cancer 6, 417–431 (2025).
A CNN trained on 8,343 colorectal cancer cases reached
an AUROC of 0.96 for MSI status. CAMIL, a context-aware
attention model, improved accuracy by understanding
spatial relationships in tumor regions.

AI-assisted PD-L1 scoring from whole slide images matched


or outperformed pathologist assessments in identifying
responders to immunotherapy.

In one trial of HER2 classification, AI disagreed with manual


scoring in 12/71 cases and manual scores were revised in 8
after review, showing AI’s added value.

📌 Fountzilas, E., Pearce, T., Baysal, M.A. et al. Convergence of evolving


artificial intelligence and machine learning techniques in precision oncology.
npj Digit. Med. 8, 75 (2025).
Current deep learning models solve one task; agents go
further by planning, executing, and refining entire research
pipelines.

AI agents can now:


Select which AI tools to use
Sequence and execute tasks autonomously
Self-reflect and learn from trial-and-error

Multi-agent systems proposed: e.g., one agent proposes


hypotheses, another critiques them.

Envisioned use cases: clinical trial design, treatment planning,


hypothesis generation​

📌 Lee, Y., Ferber, D., Rood, J.E. et al. How AI agents will change cancer
research and oncology. Nat Cancer 5, 1765–1767
Developed a multimodal deep learning model (GMLF) that
integrates:
H&E histology slides (via SlideGraph+ GNNs for spatial
+ morphological features)
Gene expression data (via MLP)

Achieved AUC of 0.74 (5-fold CV) and 0.72 (holdout) in


predicting complete NAC response.

SHAP-based interpretation identified TP63, CCL5, DCN as


top genes linked to response.

Cell-type enriched patches helped distinguish responders


vs non-responders using tumor-stromal ratios

📌 Bai, Z., Osman, M., Brendel, M. et al. Predicting response to neoadjuvant


chemotherapy in muscle-invasive bladder cancer via interpretable multimodal
deep learning. npj Digit. Med. 8, 174 (2025)
Digital twins, AI-guided trial recruitment, and multi-omics
fusion expected to see rapid deployment.

Emphasis on integrating spatial omics and longitudinal


data into AI models for better dynamic predictions.

Push toward AI systems that can simulate patient response


to guide treatment pathways before clinical application.

Recognized the growing need for standardized evaluation


and equity-centered design in cancer AI deployment.
AI x Oncology
AI is no longer a research experiment. It’s a
clinical force.

📌 Save this roundup


🔁 Share it with someone shaping the future
of cancer care

Vidith Phillips
Clinical AI Researcher
Johns Hopkins Medicine

You might also like