A
Here’s a concrete example of how an LLM could be used in oncology clinical practice.
🧬 Example: LLM Assisting in a Breast Cancer Case
🧾 Patient Data (Input to LLM)
52-year-old female
Invasive ductal carcinoma
Stage II (T2N1)
ER-positive (90%)
PR-positive (70%)
HER2-negative
Ki-67: 25%
No distant metastases
🤖 Step 1: Structured Summary
The LLM reads the pathology + staging report and outputs:
“This is a hormone receptor–positive, HER2-negative, node-positive early-stage breast cancer in a postmenopausal woman.”
📚 Step 2: Guideline-Based Treatment Suggestions
The LLM cross-references guidelines (e.g., from the National Comprehensive Cancer Network and American Society of Clinical Oncology):
It generates:
Consider adjuvant chemotherapy due to node positivity
Followed by endocrine therapy (aromatase inhibitor preferred in postmenopausal patients)
Consider genomic assay (e.g., Oncotype DX) to assess recurrence risk
Radiation therapy after surgery
⚠️ Final decision remains with the oncologist.
🧪 Step 3: Clinical Trial Matching
The LLM searches trial databases and identifies:
“Eligible for ongoing Phase III trial investigating CDK4/6 inhibitor in high-risk HR+ early breast cancer.”
It summarizes eligibility criteria in plain language.
📝 Step 4: Patient-Friendly Explanation
The LLM generates:
“Your cancer grows in response to hormones like estrogen. This is good news because we have effective hormone-blocking treatments. Because one lymph node is involved, your doctor may recommend chemotherapy to reduce recurrence risk.”
🏥 Example 2: LLM in Lung Cancer Molecular Tumor Board
Input:
Stage IV non-small cell lung cancer
EGFR exon 19 deletion
LLM output:
“First-line therapy: EGFR tyrosine kinase inhibitor such as osimertinib.”
Flags resistance mutation monitoring.
Summarizes key trial evidence (e.g., FLAURA study).
💡 Where This Actually Helps
Tumor board preparation
Literature summarization
EHR note drafting
Trial eligibility screening
Patient education material generation
If you'd like, I can also give:
🔬 A technical architecture example (how to build this system)
🏗 A startup product example
📊 A research use-case example
🧠 An example showing risks (hallucination scenario)
Which direction do you want?
No comments:
Post a Comment