---
id: wiki-2026-0508-autism-spectrum-disorder
title: Autism Spectrum Disorder
category: 10_Wiki/Topics
status: verified
canonical_id: self
aliases: [ASD, Autism, 자폐스펙트럼장애]
duplicate_of: none
source_trust_level: A
confidence_score: 0.9
verification_status: applied
tags: [neurodevelopment, psychiatry, autism, dsm5]
raw_sources: []
last_reinforced: 2026-05-10
github_commit: pending
tech_stack:
  language: python
  framework: pandas/scikit-learn
---

# Autism Spectrum Disorder

## 매 한 줄
> **"매 social communication deficits + restricted/repetitive behaviors 의 neurodevelopmental condition — 매 dimensional spectrum."** 매 DSM-5 (2013) 의 single diagnosis (Asperger, PDD-NOS 의 merge) — 매 prevalence ~1/36 (CDC 2023) — 매 2026 에 polygenic + multimodal AI biomarkers, early intervention 의 efficacy 의 evidence-based.

## 매 핵심

### 매 DSM-5 criteria
- **A. Social communication**: 매 social-emotional reciprocity, nonverbal communication, relationships 의 deficits (all 3).
- **B. Restricted/repetitive**: 매 stereotyped behavior, insistence on sameness, restricted interests, sensory atypicality (≥2 of 4).
- **C. Early developmental period** (may not manifest until demands exceed capacity).
- **D. Functional impairment**.
- **E. Not better explained by ID/global delay**.

### 매 levels of support
- **Level 1**: requiring support.
- **Level 2**: requiring substantial support.
- **Level 3**: requiring very substantial support.

### 매 응용
1. Early screening (M-CHAT, ADOS-2).
2. Multimodal AI diagnosis (eye-tracking + voice + behavioral).
3. Personalized intervention (ABA, ESDM, JASPER).

## 💻 패턴

### M-CHAT-R/F scoring
```python
def mchat_rf_score(responses):
    # 20 items, certain answers indicate risk
    risk_answers = {
        1: "no", 2: "no", 3: "no", 4: "no", 5: "no",
        6: "no", 7: "no", 8: "no", 9: "no", 10: "no",
        11: "yes", 12: "yes", 13: "no", 14: "no", 15: "no",
        16: "no", 17: "no", 18: "yes", 19: "no", 20: "yes",
    }
    score = sum(1 for k, v in responses.items() if v == risk_answers[k])
    if score >= 8: return "high"
    if score >= 3: return "medium"  # follow-up interview
    return "low"
```

### Eye-tracking social attention
```python
import numpy as np

def social_attention_ratio(gaze_xy, face_aoi, object_aoi):
    in_face = points_in_aoi(gaze_xy, face_aoi).sum()
    in_obj = points_in_aoi(gaze_xy, object_aoi).sum()
    # Lower ratio observed in ASD vs TD
    return in_face / (in_face + in_obj + 1e-8)
```

### Repetitive behavior detection (accelerometer)
```python
from scipy.signal import find_peaks, welch

def stereotypy_score(accel_xyz, fs=50):
    mag = np.linalg.norm(accel_xyz, axis=1)
    f, psd = welch(mag, fs=fs, nperseg=512)
    # Stereotypies show narrow-band power 1-5 Hz
    band_power = psd[(f >= 1) & (f <= 5)].sum()
    total = psd.sum()
    return band_power / total
```

### Voice prosody features
```python
import librosa

def prosody_features(wav, sr):
    f0 = librosa.yin(wav, fmin=80, fmax=400, sr=sr)
    f0 = f0[f0 > 0]
    return {
        "f0_mean": np.mean(f0),
        "f0_std": np.std(f0),     # often atypical (mono- or sing-song)
        "f0_range": np.ptp(f0),
        "speaking_rate": estimate_rate(wav, sr),
    }
```

### Polygenic risk score
```python
def prs(genotype, weights):
    # weights: dict snp_id -> beta from GWAS
    score = 0.0
    for snp, dose in genotype.items():
        if snp in weights:
            score += dose * weights[snp]
    return score
```

## 매 결정 기준
| 상황 | Approach |
|---|---|
| Toddler screen | M-CHAT-R/F at 18 + 24 months |
| Diagnostic confirm | ADOS-2 + ADI-R (gold standard) |
| Early intervention (<3y) | ESDM (Early Start Denver Model) |
| School-age | ABA, social skills groups, IEP |
| Co-occurring anxiety | CBT (modified), SSRI |
| Aggression / SIB | FBA + behavioral plan; meds last |

**기본값**: 매 early screen → multidisciplinary eval → individualized plan.

## 🔗 Graph
- 부모: [[Neurodevelopmental Disorders]]
- 응용: [[ABA]]
- Adjacent: [[ADHD]] · [[Amygdala Hyperactivity]]

## 🤖 LLM 활용
**언제**: 매 caregiver psychoeducation, 매 IEP draft, 매 social-story generation.
**언제 X**: 매 diagnosis, 매 medication — 매 clinician 의.

## ❌ 안티패턴
- **Single-snapshot diagnosis**: 매 longitudinal observation 의 needed.
- **One-size-fits-all therapy**: 매 high heterogeneity — 매 individualization.
- **MMR vaccine link**: 매 debunked (Wakefield retracted 2010).
- **Cure-focused framing**: 매 neurodiversity perspective 의 respect.

## 🧪 검증 / 중복
- Verified (DSM-5-TR 2022, CDC ADDM 2023, Lord et al. Lancet 2018).
- 신뢰도 A.

## 🕓 Changelog
| 날짜 | 변경 |
|---|---|
| 2026-05-08 | Phase 1 |
| 2026-05-10 | Manual cleanup — DSM criteria + screening/biomarker patterns |