Neurology Intelligence Agent -- Demo Guide¶

Version: 1.0.0 Date: 2026-03-22 Author: Adam Jones

Pre-Demo Setup¶

Verify Services¶

# Confirm all services running
docker compose ps

# Verify health
curl -s http://localhost:8528/health | python -m json.tool

# Expected: "status": "healthy", "collections": 14, "workflows": 9, "scales": 10

# Open Streamlit UI
open http://localhost:8529

# Open API docs
open http://localhost:8528/docs

Demo Environment Checklist¶

All 14 collections loaded (check /collections)
Streamlit UI responsive at port 8529
API health returns "healthy"
Anthropic API key configured and LLM responding
Browser tabs ready for Streamlit UI and API docs
Fallback slides prepared

Demo Scenario 1: Acute Stroke Triage¶

Clinical Setup: Emergency department. 72-year-old woman presents with sudden-onset left-sided weakness and speech difficulty. Last known well 2 hours ago. CT head negative for hemorrhage. CTA shows right M1 MCA occlusion.

Step 1: Calculate NIHSS¶

Via API:

curl -X POST http://localhost:8528/v1/neuro/scale/calculate \
  -H "Content-Type: application/json" \
  -d '{
    "scale_type": "nihss",
    "scores": {
      "1a_loc": 0,
      "1b_loc_questions": 1,
      "1c_loc_commands": 0,
      "2_gaze": 1,
      "3_visual": 2,
      "4_facial": 2,
      "5a_left_arm": 4,
      "5b_right_arm": 0,
      "6a_left_leg": 3,
      "6b_right_leg": 0,
      "7_ataxia": 0,
      "8_sensory": 1,
      "9_language": 2,
      "10_dysarthria": 2,
      "11_extinction": 1
    }
  }'

Expected output: NIHSS = 19 (Moderate-to-severe stroke). Recommendations: IV alteplase, CTA for LVO evaluation, consider mechanical thrombectomy.

Step 2: Calculate ASPECTS¶

curl -X POST http://localhost:8528/v1/neuro/scale/calculate \
  -H "Content-Type: application/json" \
  -d '{
    "scale_type": "aspects",
    "affected_regions": ["I", "M2"]
  }'

Expected output: ASPECTS = 8/10 (Favorable for intervention). Two regions with early ischemic change.

Step 3: Run Stroke Triage Workflow¶

Via Streamlit: Type the following query:

"72-year-old woman with acute left hemiparesis and aphasia, NIHSS 19, last known well 2 hours ago, CT negative for hemorrhage, CTA showing right M1 MCA occlusion. ASPECTS 8. Is she eligible for tPA and thrombectomy?"

Demo talking points: - NIHSS 19 indicates moderate-to-severe stroke -- tPA strongly recommended - Within 4.5-hour window -- tPA eligible (no contraindications mentioned) - M1 occlusion with ASPECTS >= 6 -- thrombectomy indicated - Door-to-needle target: < 60 minutes for tPA - Door-to-groin target: < 90 minutes for thrombectomy - Agent cites AHA/ASA 2019 guidelines and DAWN/DEFUSE-3 trial criteria

Demo Scenario 2: Memory Clinic -- Alzheimer's Evaluation¶

Clinical Setup: Outpatient memory clinic. 68-year-old man with 18 months of progressive memory loss. Retired professor. Wife reports word-finding difficulty and getting lost while driving.

Step 1: Calculate MoCA¶

curl -X POST http://localhost:8528/v1/neuro/scale/calculate \
  -H "Content-Type: application/json" \
  -d '{
    "scale_type": "moca",
    "domain_scores": {
      "visuospatial": 3,
      "naming": 3,
      "attention": 4,
      "language": 2,
      "abstraction": 1,
      "delayed_recall": 1,
      "orientation": 6
    },
    "education_years": 20
  }'

Expected output: MoCA = 20/30 (Mild cognitive impairment). Notably impaired: delayed_recall, abstraction. Recommendations: neuropsychological testing, structural MRI, consider amyloid PET or CSF biomarkers.

Step 2: Query Dementia Evaluation¶

Via Streamlit:

"68-year-old man with 18 months progressive memory loss, MoCA 20/30 with impaired delayed recall and abstraction. APOE e3/e4 genotype. MRI shows mild bilateral hippocampal atrophy. What is the ATN staging and is he a candidate for anti-amyloid therapy?"

Demo talking points: - MoCA 20 suggests MCI-to-mild dementia range - APOE e3/e4 increases AD risk (3x vs e3/e3) - Hippocampal atrophy suggests neurodegeneration (N+) - Recommend amyloid PET or CSF Abeta42/p-tau to determine A and T status - If A+T+, patient may be candidate for lecanemab (CLARITY AD trial criteria: MoCA 18-26, amyloid-positive) - Agent generates differential: AD most likely, but consider DLB, vascular, FTD

Demo Scenario 3: Drug-Resistant Epilepsy¶

Clinical Setup: Epilepsy clinic. 28-year-old woman with temporal lobe epilepsy since age 14. Failed levetiracetam, lamotrigine, and carbamazepine. Seizure frequency: 4-6 focal impaired awareness seizures per month.

Step 1: Query Epilepsy Classification¶

Via Streamlit:

"28-year-old woman with drug-resistant temporal lobe epilepsy. EEG shows right anterior temporal sharp waves and TIRDA. MRI shows right mesial temporal sclerosis. Failed levetiracetam, lamotrigine, and carbamazepine. 4-6 focal impaired awareness seizures per month with deja vu aura and oral automatisms. Is she a surgical candidate?"

Demo talking points: - Meets ILAE definition of drug-resistant epilepsy (failed 2+ appropriate ASMs) - TLE with hippocampal sclerosis is the most surgically remediable epilepsy syndrome - EEG-MRI concordance (right temporal spikes + right MTS) is favorable for surgery - Anterior temporal lobectomy offers 60-80% seizure freedom rate - Additional pre-surgical workup: video-EEG monitoring, neuropsychological testing, Wada test - Alternative: LITT (laser interstitial thermal therapy) or RNS - Agent identifies contraindicated ASMs for other syndromes (educational contrast)

Step 2: Scale Calculators (Show Flexibility)¶

Demonstrate other scale calculators to show breadth:

# ALSFRS-R for ALS patient
curl -X POST http://localhost:8528/v1/neuro/scale/calculate \
  -H "Content-Type: application/json" \
  -d '{
    "scale_type": "alsfrs_r",
    "scores": {
      "speech": 3, "salivation": 3, "swallowing": 3,
      "handwriting": 2, "cutting_food": 2, "dressing_hygiene": 2,
      "turning_in_bed": 3, "walking": 2, "climbing_stairs": 1,
      "dyspnea": 3, "orthopnea": 3, "respiratory_insufficiency": 3
    },
    "months_since_onset": 12
  }'

Demo Scenario 4: New Brain Mass¶

Clinical Setup: Neurosurgery referral. 55-year-old man with 3-week history of progressive headache, new-onset seizure. MRI shows 4 cm ring-enhancing left temporal lobe mass with surrounding edema.

Step 1: Query Brain Tumor Evaluation¶

Via Streamlit:

"55-year-old man with new-onset seizure, progressive headache. MRI shows 4 cm ring-enhancing mass in left temporal lobe with significant vasogenic edema and midline shift. What is the differential diagnosis and what molecular workup is needed?"

Demo talking points: - Top differential: glioblastoma (IDH-wildtype), brain metastasis, CNS lymphoma, abscess - Critical molecular markers to determine: IDH mutation status, MGMT methylation, TERT promoter, EGFR amplification - WHO 2021 classification requires integrated molecular diagnosis - If GBM (IDH-wt): Stupp protocol (maximal safe resection + RT 60 Gy + temozolomide + TTFields) - If IDH-mutant astrocytoma: different prognosis and treatment approach - MGMT methylation predicts temozolomide response - Agent cites NCCN CNS guidelines and WHO 2021 classification

Demo Scenario 5: MS Disease Monitoring¶

Clinical Setup: MS clinic follow-up. 34-year-old woman with RRMS diagnosed 3 years ago. On dimethyl fumarate (Tecfidera). New MRI shows 3 new T2 lesions and 1 Gd-enhancing lesion. EDSS 3.0.

Step 1: Calculate EDSS¶

curl -X POST http://localhost:8528/v1/neuro/scale/calculate \
  -H "Content-Type: application/json" \
  -d '{
    "scale_type": "edss",
    "fs_scores": {
      "visual": 0,
      "brainstem": 1,
      "pyramidal": 3,
      "cerebellar": 1,
      "sensory": 2,
      "bowel_bladder": 0,
      "cerebral": 0
    },
    "edss_step": 3.0
  }'

Step 2: Query MS Monitoring¶

Via Streamlit:

"34-year-old woman with RRMS on dimethyl fumarate for 2 years. New MRI shows 3 new T2 lesions and 1 gadolinium-enhancing lesion. EDSS 3.0. JCV antibody index 0.9. NfL elevated at 28 pg/mL. Should we escalate DMT? What are the options?"

Demo talking points: - NOT meeting NEDA-3: new T2 lesions = disease activity on current DMT - Elevated NfL supports active disease - DMT escalation indicated: switch from moderate to high-efficacy - Options: ocrelizumab (Ocrevus), natalizumab (Tysabri), ofatumumab (Kesimpta) - JCV index 0.9 = moderate PML risk with natalizumab (requires monitoring q6 months) - Ocrelizumab may be preferred given JCV status - Agent explains NEDA-3 criteria and DMT tier classification - Agent cites McDonald 2017 criteria and AAN DMT guidelines

Demo Tips¶

Start with stroke -- most dramatic, time-critical, clear decision points
Show scale calculators -- they respond instantly and demonstrate clinical accuracy
Use the Streamlit UI for natural language queries (more engaging for audience)
Use the API docs (Swagger) to show the full endpoint catalog
Highlight cross-collection search -- mention the 14 collections being searched in parallel
Point out citations -- the agent cites specific guidelines and trials
If LLM is slow, use the scale calculators as a bridge (they are instant)
Fallback: if Anthropic API is down, switch to search-only mode and show raw retrieval results

Quick Reference: Scale Calculator API¶

Scale	POST Path	Required Fields
NIHSS	`/v1/neuro/scale/calculate`	`scale_type: "nihss"`, `scores: {}`
GCS	`/v1/neuro/scale/calculate`	`scale_type: "gcs"`, `eye`, `verbal`, `motor`
MoCA	`/v1/neuro/scale/calculate`	`scale_type: "moca"`, `domain_scores: {}`
UPDRS	`/v1/neuro/scale/calculate`	`scale_type: "updrs_part_iii"`, `scores: {}`
EDSS	`/v1/neuro/scale/calculate`	`scale_type: "edss"`, `fs_scores: {}`, `edss_step`
mRS	`/v1/neuro/scale/calculate`	`scale_type: "mrs"`, `score`
HIT-6	`/v1/neuro/scale/calculate`	`scale_type: "hit6"`, `responses: []`
ALSFRS-R	`/v1/neuro/scale/calculate`	`scale_type: "alsfrs_r"`, `scores: {}`
ASPECTS	`/v1/neuro/scale/calculate`	`scale_type: "aspects"`, `affected_regions: []`
Hoehn-Yahr	`/v1/neuro/scale/calculate`	`scale_type: "hoehn_yahr"`, `stage`

Neurology Intelligence Agent -- Demo Guide v1.0.0 HCLS AI Factory / GTC Europe 2026