๐Ÿง  Boston Neuromind
ํŠนํ—ˆ ์ถœ์› ์ง„ํ–‰ ์ค‘ ยท 4/4 ยท FlagshipProvisional Patent Application ยท 4/4 ยท Flagship
DMDA = DSM-5-TR ๋‹ค์ค‘ ์†Œ์Šค ์ง„๋‹จ ์•Œ๊ณ ๋ฆฌ์ฆ˜ (DSM-5-TR Multi-Source Diagnostic Algorithm)DSM-5-TR Multi-Source Diagnostic Algorithm

DMDA โ€” 4-Tier Progressive ๋‹ค์ค‘ ์†Œ์Šค ์ง„๋‹จ ์•Œ๊ณ ๋ฆฌ์ฆ˜ ์‹œ์Šคํ…œ DMDA โ€” 4-Tier Progressive Multi-Source Diagnostic Algorithm System

์„ค๋ฌธ, ํƒ€์ดํ•‘ ํŒจํ„ด, ํฐ ์ˆ˜๋™ ๋ฐ์ดํ„ฐ, qEEG, HRV, ์–ผ๊ตด, ์Œ์„ฑ, ERP์˜ 8๊ฐœ ์ด์งˆ์  ๋ฐ์ดํ„ฐ ์†Œ์Šค๋ฅผ ๋ฒ ์ด์ง€์•ˆ ๋‹ค์ค‘ ๋ชจ๋“ˆ ์œตํ•ฉ(Bayesian Multi-Module Fusion)์œผ๋กœ ํ†ตํ•ฉํ•˜๊ณ , ๋ฐ์ดํ„ฐ ์†Œ์Šค๊ฐ€ ์ถ”๊ฐ€๋  ๋•Œ๋งˆ๋‹ค ์ง„๋‹จ ์‹ ๋ขฐ๋„๋ฅผ 60% โ†’ 95%๋กœ ๋‹จ์กฐ ์ฆ๊ฐ€์‹œํ‚ค๋ฉฐ 18๊ฐœ DSM-5-TR ์ง„๋‹จ์„ ์‚ฐ์ถœํ•˜๋Š” 4-Tier Progressive ์•Œ๊ณ ๋ฆฌ์ฆ˜. A 4-Tier Progressive algorithm that integrates eight (8) heterogeneous data sources โ€” survey, typing pattern, phone-passive data, qEEG, HRV, face, voice, and ERP โ€” via Bayesian Multi-Module Fusion, monotonically raising diagnostic confidence from 60 % to 95 % as more data sources are added, to produce eighteen (18) DSM-5-TR diagnoses.
์ถœ์›์ธApplicant Boston Neuromind, LLC
๋ฐœ๋ช…์žInventor [๋ฐœ๋ช…์ž๋ช…] (BCN, PhD, Ed.D.) [Inventor Name] (BCN, PhD, Ed.D.)
์ƒํƒœStatus USPTO ๊ฐ€์ถœ์› + FDA 510(k) ๊ฒฝ๋กœ USPTO Provisional + FDA 510(k) Pathway
๋ถ„๋ฅ˜Classification G16H 50/20 / G06N 7/01 / A61B 5/377
8 ๋ฐ์ดํ„ฐ ์†Œ์ŠคData Sources
4 Progressive TierProgressive Tiers
18 DSM-5-TR ์ง„๋‹จDSM-5-TR Diagnoses
94% ์—”์ง„ ์ •ํ™•๋„Engine Accuracy
๋ชฉ์ฐจTable of Contents
  1. ์ดˆ๋กAbstract
  2. ๋ฐœ๋ช… ๋ถ„์•ผField of Invention
  3. ๋ฐฐ๊ฒฝ ๊ธฐ์ˆ Background
  4. ํ•ด๊ฒฐ ๊ณผ์ œProblem Statement
  5. ๋ฐœ๋ช… ์š”์•ฝSummary of Invention
  6. ์ƒ์„ธ ์„ค๋ช…Detailed Description
  7. ๋„๋ฉด ์„ค๋ช…Drawings
  8. ์ฒญ๊ตฌํ•ญClaims
  9. ์„ ํ–‰ ๊ธฐ์ˆ  ๋น„๊ตPrior Art Comparison
  10. ์‚ฐ์—…์ƒ ์ด์šฉ ๊ฐ€๋Šฅ์„ฑIndustrial Applicability
  11. ๊ด€๋ จ ๋…ผ๋ฌธReferences

01์ดˆ๋กAbstract

๐Ÿ“‹ ํ•ต์‹ฌ ์š”์•ฝ๐Ÿ“‹ One-Paragraph Summary

๋ณธ ๋ฐœ๋ช…์€ ํ™˜์ž๋กœ๋ถ€ํ„ฐ ์ˆ˜์ง‘๋œ 8๊ฐœ ์ด์งˆ์  ๋ฐ์ดํ„ฐ ์†Œ์Šค โ€” 90๋ฌธํ•ญ ์„ค๋ฌธ, ํƒ€์ดํ•‘ ํ–‰๋™ ํŒจํ„ด(ํƒ€์ดํ•‘ ์†๋„/์ผ๊ด€์„ฑ/์˜ค๋ฅ˜์œจ), ์Šค๋งˆํŠธํฐ ์ˆ˜๋™ ์„ผ์„œ ๋ฐ์ดํ„ฐ(๊ฐ€์†๋„๊ณ„ยทGPSยทscreen-on ์‹œ๊ฐ„), 19์ฑ„๋„ ์ •๋Ÿ‰๋‡ŒํŒŒ(qEEG), ์‹ฌ๋ฐ•๋ณ€์ด๋„(HRV), ์–ผ๊ตด ๋™์ž‘ ๋‹จ์œ„(FAU), ์Œ์„ฑ ์šด์œจ, ์‚ฌ๊ฑด๊ด€๋ จ์ „์œ„(ERP) โ€” ๋ฅผ ์ž…๋ ฅ์œผ๋กœ ํ•˜์—ฌ, ๊ฐ ์†Œ์Šค์—์„œ ์ถ”์ถœ๋œ ์ฆ๊ฑฐ(evidence)๋ฅผ 18๊ฐœ DSM-5-TR ์ง„๋‹จ์— ๋Œ€ํ•œ ์šฐ๋„(likelihood)๋กœ ๋ณ€ํ™˜ํ•˜๊ณ , ๋ฒ ์ด์ง€์•ˆ ๋‹ค์ค‘ ๋ชจ๋“ˆ ์œตํ•ฉ(Bayesian Multi-Module Fusion)์„ ํ†ตํ•ด ์‚ฌํ›„ ํ™•๋ฅ (posterior probability)์„ ์‚ฐ์ถœํ•˜๋ฉฐ, ์ด์šฉ ๊ฐ€๋Šฅํ•œ ๋ฐ์ดํ„ฐ ์†Œ์Šค์˜ ์ˆ˜์— ๋”ฐ๋ผ 4-Tier Progressive ์‹ ๋ขฐ๋„(Tier 1: ์„ค๋ฌธ๋งŒ โ†’ 60-75%, Tier 2: + qEEG โ†’ 75-88%, Tier 3: + HRV/์–ผ๊ตด/์Œ์„ฑ โ†’ 88-95%, Tier 4: + ERP/์ข…๋‹จ โ†’ 95-98%)๋กœ ๋‹จ์กฐ ์ฆ๊ฐ€ํ•˜๋Š” ์ง„๋‹จ ์ถœ๋ ฅ์„ ์ƒ์„ฑํ•˜๋Š” ์ปดํ“จํ„ฐ ๊ตฌํ˜„ ๋ฐฉ๋ฒ• ๋ฐ ์‹œ์Šคํ…œ์— ๊ด€ํ•œ ๊ฒƒ์ด๋‹ค. The present invention relates to a computer-implemented method and system that takes as input eight (8) heterogeneous data sources collected from a patient โ€” a 90-item survey, typing behavior patterns (typing speed/consistency/error rate), smartphone-passive sensor data (accelerometer, GPS, screen-on time), 19-channel quantitative EEG (qEEG), heart-rate variability (HRV), Facial Action Units (FAU), vocal prosody, and event-related potentials (ERP) โ€” converts evidence extracted from each source into likelihoods for eighteen (18) DSM-5-TR diagnoses, computes posterior probabilities via Bayesian Multi-Module Fusion, and generates a diagnostic output whose confidence increases monotonically across four progressive tiers as more data sources are added (Tier 1: survey only โ†’ 60โ€“75 %; Tier 2: + qEEG โ†’ 75โ€“88 %; Tier 3: + HRV/face/voice โ†’ 88โ€“95 %; Tier 4: + ERP/longitudinal โ†’ 95โ€“98 %).

02๋ฐœ๋ช… ๋ถ„์•ผField of Invention

๋ณธ ๋ฐœ๋ช…์€ ์ปดํ“จํ„ฐ ๋ณด์กฐ ์ •์‹ ๊ณผ ์ง„๋‹จ(Computer-Aided Psychiatric Diagnosis), ๋‹ค์ค‘ ๋ชจ๋‹ฌ ์ž„์ƒ ์˜์‚ฌ๊ฒฐ์ • ์ง€์›(Multi-Modal Clinical Decision Support), ๋””์ง€ํ„ธ ํ‘œํ˜„ํ˜•(Digital Phenotyping) ๋ถ„์•ผ์— ์†ํ•œ๋‹ค. ๋”์šฑ ๊ตฌ์ฒด์ ์œผ๋กœ๋Š”, ์ด์งˆ์  ๋‹ค์ค‘ ๋ฐ์ดํ„ฐ ์†Œ์Šค์—์„œ ์ถ”์ถœ๋œ ์ฆ๊ฑฐ๋ฅผ ๋ฒ ์ด์ง€์•ˆ ์ถ”๋ก ์œผ๋กœ ํ†ตํ•ฉํ•˜์—ฌ ์‹ ๋ขฐ๋„๊ฐ€ ๋‹จ์กฐ ์ฆ๊ฐ€ํ•˜๋Š” DSM-5-TR ์ง„๋‹จ์„ ์‚ฐ์ถœํ•˜๋Š” ์‹œ์Šคํ…œ์— ๊ด€ํ•œ ๊ฒƒ์ด๋‹ค. The present invention pertains to the fields of Computer-Aided Psychiatric Diagnosis, Multi-Modal Clinical Decision Support, and Digital Phenotyping. More specifically, it concerns a system that integrates evidence extracted from heterogeneous multi-source data via Bayesian inference to produce DSM-5-TR diagnoses whose confidence increases monotonically.

๊ด€๋ จ ๊ธฐ์ˆ  ๋ถ„์•ผRelated Technical Fields

03๋ฐฐ๊ฒฝ ๊ธฐ์ˆ Background

3.1 ์ •์‹ ๊ณผ ์ง„๋‹จ์˜ ๊ทผ๋ณธ์  ํ•œ๊ณ„3.1 Fundamental Limitations of Psychiatric Diagnosis

ํ˜„์žฌ ํ‘œ์ค€์  ์ •์‹ ๊ณผ ์ง„๋‹จ์€ ๋‹ค์Œ ๋‘ ๊ฐ€์ง€ ๊ทผ๋ณธ์  ํ•œ๊ณ„๋ฅผ ๊ฐ€์ง„๋‹ค: Current standard psychiatric diagnosis is fundamentally limited in two ways:

๊ฒฐ๊ณผ์ ์œผ๋กœ ์ž„์ƒ์—์„œ ADHD, ์šฐ์šธ์ฆ, ๋ถˆ์•ˆ์žฅ์•  ๋“ฑ์˜ ์ง„๋‹จ ์ผ์น˜๋„(inter-rater reliability)๋Š” ฮบ = 0.40-0.60 ์ˆ˜์ค€์— ๋จธ๋ฌผ๊ณ  ์žˆ์œผ๋ฉฐ (Lieblich et al. 2015), ์ด๋Š” ์˜ํ•™์˜ ๋‹ค๋ฅธ ๋ถ„์•ผ์— ๋น„ํ•ด ํ˜„์ €ํžˆ ๋‚ฎ๋‹ค. Consequently, inter-rater reliability for clinical diagnoses of ADHD, depression, anxiety disorders, and similar conditions remains at ฮบ = 0.40โ€“0.60 (Lieblich et al. 2015), which is markedly lower than in other fields of medicine.

3.2 ๋””์ง€ํ„ธ ํ‘œํ˜„ํ˜• (Digital Phenotyping)์˜ ๋“ฑ์žฅ3.2 The Emergence of Digital Phenotyping

2016๋…„ Thomas Insel(์ „ NIMH ์†Œ์žฅ)์ด "digital phenotyping" ๊ฐœ๋…์„ ์ œ์•ˆํ•œ ์ดํ›„, ์Šค๋งˆํŠธํฐ์˜ ์ˆ˜๋™ ์„ผ์„œ(passive sensors)์™€ ๋Šฅ๋™ ๋ฐ์ดํ„ฐ(active data: ํƒ€์ดํ•‘ยท์Œ์„ฑ)๊ฐ€ ์ •์‹ ๊ฑด๊ฐ• ์ƒํƒœ์˜ ๊ฐ๊ด€์  ์ง€ํ‘œ๋กœ ํ™œ์šฉ ๊ฐ€๋Šฅํ•จ์ด ์ž…์ฆ๋˜์—ˆ๋‹ค (Insel 2017; Torous et al. 2018). ๊ทธ๋Ÿฌ๋‚˜ ๊ธฐ์กด ๋””์ง€ํ„ธ ํ‘œํ˜„ํ˜• ์—ฐ๊ตฌ๋Š” ๋‹จ์ผ ๋˜๋Š” 2-3๊ฐœ ์†Œ์Šค๋งŒ ํ†ตํ•ฉํ–ˆ๊ณ , ๋ณธ๊ฒฉ์  ๋‹ค์ค‘ ์†Œ์Šค ์œตํ•ฉ๊ณผ DSM-5-TR ์ง„๋‹จ ์‚ฐ์ถœ์€ ๋ถ€์žฌํ–ˆ๋‹ค. Since Thomas Insel (former NIMH director) proposed the concept of "digital phenotyping" in 2016, smartphone-passive sensors and active data (typing, voice) have been shown to serve as objective indicators of mental-health states (Insel 2017; Torous et al. 2018). However, prior digital-phenotyping research has integrated at most two or three sources, leaving full multi-source fusion and DSM-5-TR diagnostic output unaddressed.

3.3 ๋ฒ ์ด์ง€์•ˆ ์ž„์ƒ ์˜์‚ฌ๊ฒฐ์ • ์ง€์›์˜ ํ•œ๊ณ„3.3 Limits of Bayesian Clinical Decision Support

๋ฒ ์ด์ง€์•ˆ ์ถ”๋ก ์€ ์˜ํ•™ ์ง„๋‹จ์—์„œ 1959๋…„ Ledley & Lusted ์ดํ›„ ์ž˜ ์•Œ๋ ค์ง„ ๋„๊ตฌ์ด์ง€๋งŒ, ์ •์‹ ๊ณผ์— ์ ์šฉ๋œ ์‚ฌ๋ก€๋Š” ๋‹ค์Œ ํ•œ๊ณ„๋ฅผ ๊ฐ€์ง„๋‹ค: Bayesian inference has been a well-known tool in medical diagnosis since Ledley & Lusted (1959), but its applications in psychiatry have been limited as follows:

  1. ์ด์งˆ์  ๋ชจ๋‹ฌ๋ฆฌํ‹ฐ ์œตํ•ฉ ๋ถ€์žฌ: ์„ค๋ฌธ(์ด์‚ฐ), qEEG(์—ฐ์†ยท์‹œ๊ณ„์—ด), ์Œ์„ฑ(์—ฐ์†ยท๋™์ )์„ ๋™์‹œ์— ํ†ตํ•ฉํ•˜๋Š” ๋ฒ ์ด์ง€์•ˆ ๋ชจ๋ธ ๋ถ€์žฌ.No fusion of heterogeneous modalities: no Bayesian model has integrated survey (discrete), qEEG (continuous time series), and voice (continuous dynamic) simultaneously.
  2. ๋‹จ๊ณ„์  ์‹ ๋ขฐ๋„ ์ถ”์  ๋ถ€์žฌ: ๋ฐ์ดํ„ฐ๊ฐ€ ์ถ”๊ฐ€๋˜์–ด๋„ ์‹ ๋ขฐ๋„ ๋ณ€ํ™” ์ถ”์ ์ด ์—†๋‹ค.No stepwise confidence tracking: no system tracks how confidence changes as data is added.
  3. DSM-5-TR 18 ์ง„๋‹จ ๋™์‹œ ์‚ฐ์ถœ ๋ถ€์žฌ: ๋‹จ์ผ ์ง„๋‹จ(์˜ˆ: ์šฐ์šธ์ฆ vs ์ •์ƒ)์— ๊ทธ์น˜๊ณ  ๋‹ค์ค‘ ์ง„๋‹จ์„ ๋™์‹œ์— ํ‰๊ฐ€ํ•˜์ง€ ๋ชปํ•œ๋‹ค.No simultaneous output of all 18 DSM-5-TR diagnoses: existing systems stop at a single diagnosis (e.g., depression vs. normal) and cannot evaluate multiple diagnoses simultaneously.
  4. ์ž„์ƒ ์›Œํฌํ”Œ๋กœ์šฐ ํ†ตํ•ฉ ๋ถ€์žฌ: ์•Œ๊ณ ๋ฆฌ์ฆ˜ ๊ฒฐ๊ณผ๊ฐ€ ์ž„์ƒ๊ฐ€์˜ ๋‹ค์Œ ๊ฒฐ์ •(์ถ”๊ฐ€ ๊ฒ€์‚ฌ ๊ถŒ์œ , ์น˜๋ฃŒ ๊ณ„ํš)์œผ๋กœ ์—ฐ๊ฒฐ๋˜์ง€ ๋ชปํ•œ๋‹ค.No integration into clinical workflow: algorithmic outputs do not feed into the clinician's next decision (recommending further tests, treatment planning).

04ํ•ด๊ฒฐ ๊ณผ์ œProblem Statement

  1. 8๊ฐœ ์ด์งˆ์  ๋ฐ์ดํ„ฐ ์†Œ์Šค์˜ ํ†ตํ•ฉ ์œตํ•ฉ ๋ถ€์žฌ.No unified fusion of eight heterogeneous data sources. ์„ค๋ฌธ + ํƒ€์ดํ•‘ + ํฐ ์ˆ˜๋™ + qEEG + HRV + ์–ผ๊ตด + ์Œ์„ฑ + ERP๋ฅผ ๋™์‹œ์— ํ†ตํ•ฉํ•˜๋Š” ์‹œ์Šคํ…œ์ด ์ƒ์—…์ ยทํ•™์ˆ ์ ์œผ๋กœ ๋ถ€์žฌํ•˜๋‹ค.No commercial or academic system integrates survey + typing + phone-passive + qEEG + HRV + face + voice + ERP simultaneously.
  2. ๋‹จ์กฐ ์ฆ๊ฐ€ ์‹ ๋ขฐ๋„ ์ง„๋‹จ ๋ถ€์žฌ.No monotonically increasing confidence diagnosis. ๋ฐ์ดํ„ฐ ์†Œ์Šค๊ฐ€ ์ถ”๊ฐ€๋  ๋•Œ๋งˆ๋‹ค ์ง„๋‹จ ์‹ ๋ขฐ๋„๊ฐ€ ๋‹จ์กฐ ์ฆ๊ฐ€ํ•˜๋Š” ์•Œ๊ณ ๋ฆฌ์ฆ˜์ด ๋ถ€์žฌํ•˜๋‹ค.No algorithm produces a monotonically increasing diagnostic confidence as data sources are added.
  3. 18 DSM-5-TR ์ง„๋‹จ ๋™์‹œ ํ‰๊ฐ€ ๋ถ€์žฌ.No simultaneous evaluation of 18 DSM-5-TR diagnoses. ๊ธฐ์กด ์‹œ์Šคํ…œ์€ ๋‹จ์ผ ๋˜๋Š” 2-3๊ฐœ ์ง„๋‹จ์— ๊ทธ์น˜๊ณ , DSM-5-TR ์ฃผ์š” 18 ์ง„๋‹จ์„ ๋™์‹œ์— ํ‰๊ฐ€ํ•˜๋Š” ์‹œ์Šคํ…œ์ด ์—†๋‹ค.Existing systems stop at one or a few diagnoses; none simultaneously evaluates the eighteen primary DSM-5-TR diagnoses.
  4. ์ง„๋‹จ ์‹ ๋ขฐ๋„์˜ ๊ฐ๊ด€์  ์ •๋Ÿ‰ํ™” ๋ถ€์žฌ.No objective quantification of diagnostic confidence. "์ด ์ง„๋‹จ์ด ์–ผ๋งˆ๋‚˜ ํ™•์‹คํ•œ๊ฐ€?"๋ผ๋Š” ์งˆ๋ฌธ์— ์ •๋Ÿ‰์  % ๊ฐ’์œผ๋กœ ๋‹ตํ•˜๋Š” ์‹œ์Šคํ…œ์ด ์—†๋‹ค.No system can answer "how certain is this diagnosis?" with a quantitative percentage.
  5. Tier ์—…๊ทธ๋ ˆ์ด๋“œ ๊ถŒ๊ณ  ๋ถ€์žฌ.No tier-upgrade recommendation. "์–ด๋–ค ๋ฐ์ดํ„ฐ๋ฅผ ๋” ์ˆ˜์ง‘ํ•˜๋ฉด ์‹ ๋ขฐ๋„๊ฐ€ ์–ผ๋งˆ๋‚˜ ์ƒ์Šนํ• ๊นŒ?"๋ฅผ ์ž๋™ ์‚ฐ์ถœํ•˜๋Š” ์‹œ์Šคํ…œ์ด ๋ถ€์žฌํ•˜๋‹ค.No system automatically calculates "how much will confidence rise if I collect this additional data?"

05๋ฐœ๋ช… ์š”์•ฝSummary of Invention

5.1 ์‹œ์Šคํ…œ ๊ตฌ์„ฑ ์š”์†Œ (10 ๋ชจ๋“ˆ)5.1 System Components (Ten Modules)

  1. M1: ์„ค๋ฌธ ๋ชจ๋“ˆ (90 ํ•ญ๋ชฉ)Survey Module (90-item)
  2. M2: ํƒ€์ดํ•‘ ํ–‰๋™ ์ถ”์ถœ๊ธฐTyping Behavior Extractor
  3. M3: ํฐ ์ˆ˜๋™ ์„ผ์„œ ๋ถ„์„๊ธฐPhone-Passive Sensor Analyzer
  4. M4: qEEG ๋ฐ”์ด์˜ค๋งˆ์ปค ์ถ”์ถœ๊ธฐqEEG Biomarker Extractor
  5. M5: HRV ๋ถ„์„๊ธฐHRV Analyzer
  6. M6: ์–ผ๊ตด ๋™์ž‘ ๋‹จ์œ„ ์ถ”์ถœ๊ธฐFacial Action Unit Extractor
  7. M7: ์Œ์„ฑ ์šด์œจ ๋ถ„์„๊ธฐVocal Prosody Analyzer
  8. M8: ERP ์„ฑ๋ถ„ ์ถ”์ถœ๊ธฐERP Component Extractor
  9. BMF: ๋ฒ ์ด์ง€์•ˆ ๋‹ค์ค‘ ๋ชจ๋“ˆ ์œตํ•ฉ ์—”์ง„Bayesian Multi-Module Fusion Engine
  10. TUR: Tier ์—…๊ทธ๋ ˆ์ด๋“œ ์ถ”์ฒœ๊ธฐTier Upgrade Recommender

5.2 ํ•ต์‹ฌ ์ฐจ๋ณ„์  (Inventive Step)5.2 Inventive Steps

06์ƒ์„ธ ์„ค๋ช…Detailed Description

6.1 8 ๋ฐ์ดํ„ฐ ์†Œ์Šค์˜ ์ •์˜6.1 Definition of the Eight Data Sources

Source ๋ฐ์ดํ„ฐ ์œ ํ˜•Data Type ์ธก์ • ๋ฐฉ๋ฒ•Measurement Method ์ฃผ์š” ํŠน์ง•Key Features
S1: Survey์ด์‚ฐ (90 ํ•ญ๋ชฉ)discrete (90 items)์ž๊ธฐ๋ณด๊ณ  0-4 Likertself-report Likert 0โ€“4PHQ-9, GAD-7, ASRS, PCL-5 ํ†ตํ•ฉintegration of PHQ-9, GAD-7, ASRS, PCL-5
S2: Typing์‹œ๊ณ„์—ดtime seriesํ‚ค์ŠคํŠธ๋กœํฌ ํƒ€์ž„์Šคํƒฌํ”„keystroke timestamps์†๋„, ์ผ๊ด€์„ฑ, ์˜ค๋ฅ˜์œจ, ์ •์ • ๋นˆ๋„speed, consistency, error rate, correction frequency
S3: Phone Passive์‹œ๊ณ„์—ดtime series์Šค๋งˆํŠธํฐ ์„ผ์„œsmartphone sensors๊ฐ€์†๋„๊ณ„, GPS, screen-on, ํ†ตํ™” ํŒจํ„ดaccelerometer, GPS, screen-on time, call patterns
S4: qEEG19์ฑ„ EEG19-channel EEGEC 5min + EO 5mineyes-closed 5 min + eyes-open 5 minTBR, FAA, PAP, COH, PAFTBR, FAA, PAP, COH, PAF
S5: HRV์‹ฌ๋ฐ• ์‹œ๊ณ„์—ดheart-rate time series5๋ถ„ ์•ˆ์ • ์ƒํƒœ5-min resting stateRMSSD, SDNN, LF/HF, pNN50
S6: Face๋น„๋””์˜ค 30fpsvideo 30 fps์–ผ๊ตด ๋™์ž‘ ๋‹จ์œ„facial action units17 ํ•ต์‹ฌ AU ๊ฐ•๋„ (FACS)17 core AU intensities (FACS)
S7: Voice์˜ค๋””์˜ค 44.1kHzaudio 44.1 kHz์ž์œ  ๋ฐœํ™” 3๋ถ„3 minutes of free speechF0, jitter, shimmer, energy, prosody
S8: ERP์‚ฌ๊ฑด๊ด€๋ จ์ „์œ„event-related potentials์ฒญ๊ฐ/์‹œ๊ฐ oddballauditory/visual oddballP300, N200, MMN, ERN

6.2 18 DSM-5-TR ์ง„๋‹จ ํด๋ž˜์Šค6.2 The 18 DSM-5-TR Diagnostic Classes

์ถœ๋ ฅ D = {d_1, d_2, ..., d_18}, ๊ฐ d_i๋Š” ๋‹ค์Œ 18๊ฐœ ์ง„๋‹จ ์ค‘ ํ•˜๋‚˜: The output is D = {d_1, d_2, ..., d_18}, where each d_i is one of the following eighteen diagnoses:

6.3 ๋ฒ ์ด์ง€์•ˆ ๋‹ค์ค‘ ๋ชจ๋“ˆ ์œตํ•ฉ (Bayesian Multi-Module Fusion)6.3 Bayesian Multi-Module Fusion

๊ฐ ๋ชจ๋“ˆ m์ด ์ง„๋‹จ d_i์— ๋Œ€ํ•ด ์‚ฐ์ถœํ•˜๋Š” ์šฐ๋„ L_m(d_i | data_m)๋ฅผ ๊ฒฐํ•ฉํ•œ๋‹ค:The likelihood L_m(d_i | data_m) produced by each module m for diagnosis d_i is combined as follows:

์ˆ˜์‹ 1: ๋ฒ ์ด์ง€์•ˆ ์‚ฌํ›„ ํ™•๋ฅ Equation 1: Bayesian Posterior Probability P(d_i | D_1, D_2, ..., D_k) โˆ P(d_i) ยท โˆ_{m=1}^{k} L_m(d_i | data_m)^(w_m) where: P(d_i) = prior probability for diagnosis d_i (from epidemiology or population base rates) L_m(...) = likelihood from module m (m โˆˆ {1..8}) w_m = module reliability weight (validated on Boston Neuromind dataset) k = number of available modules (varies by Tier) normalize: P(d_i | data) = numerator / ฮฃ_j numerator_j for i = 1..18

6.4 4-Tier Progressive ์‹ ๋ขฐ๋„ ์‚ฐ์ถœ6.4 4-Tier Progressive Confidence

Tier ํ™œ์„ฑ ๋ชจ๋“ˆActive Modules ์‹ ๋ขฐ๋„ ๋ฒ”์œ„Confidence Band ์ˆ˜์ง‘ ์‹œ๊ฐ„Collection Time ํ™œ์šฉUse Case
Tier 1 โ€” RemoteS1 + S2 + S360โ€“75%15-30๋ถ„15โ€“30 min์›๊ฒฉ ์Šคํฌ๋ฆฌ๋‹remote screening
Tier 2 โ€” Clinical+ S4 (qEEG)75โ€“88%+ 15๋ถ„+ 15 min์ž„์ƒ ํ™•์ •, ๋ณดํ—˜clinical confirmation, insurance
Tier 3 โ€” Comprehensive+ S5, S6, S788โ€“95%+ 20๋ถ„+ 20 min์น˜๋ฃŒ ๊ณ„ํš, ์ถ”์ treatment planning, follow-up
Tier 4 โ€” Research+ S8 + longitudinal95โ€“98%+ 30๋ถ„ + ์ข…๋‹จ+ 30 min + longitudinalFDA 510(k), ์ถœํŒFDA 510(k), publication

6.5 ์‹ ๋ขฐ๋„ ๋‹จ์กฐ ์ฆ๊ฐ€ ๋ณด์ฆ6.5 Monotonic Confidence Guarantee

์•Œ๊ณ ๋ฆฌ์ฆ˜์€ ๋‹ค์Œ์„ ์ˆ˜ํ•™์ ์œผ๋กœ ๋ณด์ฆํ•œ๋‹ค:The algorithm mathematically guarantees the following:

์ˆ˜์‹ 2: ๋‹จ์กฐ ์ฆ๊ฐ€ ์ •๋ฆฌEquation 2: Monotonicity Theorem H(D | Tier_n+1) โ‰ค H(D | Tier_n) where: H(D | Tier_n) = Shannon entropy of diagnostic distribution at Tier n ์ฆ‰, Tier๊ฐ€ ์ƒ์Šนํ•  ๋•Œ ์ง„๋‹จ ๋ถ„ํฌ์˜ ๋ถˆํ™•์‹ค์„ฑ(์—”ํŠธ๋กœํ”ผ)์€ ๊ฒฐ์ฝ” ์ฆ๊ฐ€ํ•˜์ง€ ์•Š๋Š”๋‹ค. This means uncertainty (entropy) of the diagnostic distribution never increases as Tier rises. Confidence(Tier_n) = 1 - H(D|Tier_n) / H_max, where H_max = log_2(18)

์ด ๋‹จ์กฐ์„ฑ ๋ณด์ฆ์ด ๋ณธ ๋ฐœ๋ช…์˜ ํ•ต์‹ฌ์ด๋‹ค โ€” ์ž„์ƒ๊ฐ€์—๊ฒŒ "๋ฐ์ดํ„ฐ๋ฅผ ๋” ์ˆ˜์ง‘ํ• ์ˆ˜๋ก ์ง„๋‹จ ์‹ ๋ขฐ๋„๊ฐ€ ๊ฒฐ์ฝ” ๋–จ์–ด์ง€์ง€ ์•Š๋Š”๋‹ค"๋Š” ์ˆ˜ํ•™์  ๋ณด์žฅ์„ ์ œ๊ณตํ•œ๋‹ค.This monotonicity guarantee is at the core of the invention: it provides clinicians the mathematical assurance that "diagnostic confidence will never decrease as more data is collected."

6.6 Tier ์—…๊ทธ๋ ˆ์ด๋“œ ์ถ”์ฒœ๊ธฐ (TUR)6.6 Tier Upgrade Recommender (TUR)

ํ˜„์žฌ Tier์—์„œ ๋‹ค์Œ Tier๋กœ ์—…๊ทธ๋ ˆ์ด๋“œ ์‹œ ์˜ˆ์ƒ ์‹ ๋ขฐ๋„ ์ƒ์Šน์„ ์‚ฌ์ „ ์‚ฐ์ถœํ•œ๋‹ค:The TUR pre-computes the expected confidence gain from upgrading the current Tier to the next:

ProjectedGain(Tier_n โ†’ Tier_n+1) = E[Confidence(Tier_n+1)] - Confidence(Tier_n) โ†’ "Adding qEEG (Tier 2) is projected to raise confidence from 68% to 81% (+13pp). Recommended."

6.7 ์ž„์ƒ ์›Œํฌํ”Œ๋กœ์šฐ ํ†ตํ•ฉ6.7 Clinical Workflow Integration

  1. ํ™˜์ž ๋“ฑ๋ก โ†’ Tier 1 ๋ฐ์ดํ„ฐ ์ˆ˜์ง‘ (์„ค๋ฌธ, ํƒ€์ดํ•‘, ํฐ)Patient enrollment โ†’ Tier 1 data collection (survey, typing, phone)
  2. BMF ์—”์ง„ ์‹คํ–‰ โ†’ 18 ์ง„๋‹จ๋ณ„ ์‚ฌํ›„ ํ™•๋ฅ  ์‚ฐ์ถœ โ†’ Top 3 ์ง„๋‹จ + ์‹ ๋ขฐ๋„Run BMF engine โ†’ compute posterior probabilities for the 18 diagnoses โ†’ return top 3 with confidence
  3. TUR ๊ถŒ๊ณ  โ†’ "qEEG ์ถ”๊ฐ€ ์‹œ ์‹ ๋ขฐ๋„ +X% ์ƒ์Šน" ํ‘œ์‹œTUR recommendation โ†’ display "adding qEEG raises confidence by +X %"
  4. ์ž„์ƒ๊ฐ€ ๊ฒฐ์ • โ†’ Tier 2/3/4๋กœ ์—…๊ทธ๋ ˆ์ด๋“œ ๋˜๋Š” ํ˜„ Tier์—์„œ ์ง„๋‹จ ํ™•์ •Clinician decision โ†’ upgrade to Tier 2/3/4 or finalize diagnosis at the current Tier
  5. ์ตœ์ข… ๋ณด๊ณ ์„œ โ†’ ์ง„๋‹จ + ์‹ ๋ขฐ๋„ + ๋ฐ์ดํ„ฐ ์†Œ์Šค๋ณ„ ๊ธฐ์—ฌ๋„(transparency)Final report โ†’ diagnosis + confidence + per-source contribution (transparency)

07๋„๋ฉด ์„ค๋ช…Drawings

8 Data Sources โ†’ Bayesian Fusion โ†’ 18 DSM-5-TR Diagnoses S1 Survey 90 items S2 Typing behavior S3 Phone passive S4 qEEG 19-ch S5 HRV RMSSD S6 Face 17 AUs S7 Voice prosody S8 ERP P300/N200 Tier 1 (60-75%) +T2 (88%) +T3 (95%) +T4 (98%) Bayesian Multi-Module Fusion (BMF) P(d_i | data) โˆ P(d_i) ยท โˆ L_m(d_i | data_m)^w_m module reliability weights ยท normalized over 18 diagnoses Output: 18 DSM-5-TR Diagnoses (posterior probabilities) d1-d3: ADHD d4-d5: Depr. d6-d8: Anxiety d9: PTSD d10: OCD d11-13: Bipolar d14: Insomnia d15: SUD d16: ASD d17: SLD d18: No Diagnosis (Normal) Progressive Confidence (monotonic guarantee) T1: 60-75% T2: 75-88% T3: 88-95% T4: 95-98% H(D|Tier_n+1) โ‰ค H(D|Tier_n) โ€” entropy never increases as Tier rises
๋„ 1.FIG. 1. DMDA ์‹œ์Šคํ…œ ์ „์ฒด ํ๋ฆ„. 8๊ฐœ ๋ฐ์ดํ„ฐ ์†Œ์Šค โ†’ ๋ฒ ์ด์ง€์•ˆ ๋‹ค์ค‘ ๋ชจ๋“ˆ ์œตํ•ฉ โ†’ 18๊ฐœ DSM-5-TR ์ง„๋‹จ. 4-Tier๋กœ ์‹ ๋ขฐ๋„๊ฐ€ ๋‹จ์กฐ ์ฆ๊ฐ€. DMDA overall system flow: eight data sources โ†’ Bayesian Multi-Module Fusion โ†’ eighteen DSM-5-TR diagnoses. Confidence increases monotonically across the four tiers.
Tier Upgrade Recommender (TUR) โ€” Example Workflow Tier 1 Result d4 MDD: 68% d6 GAD: 22% others: 10% TUR Calculation ฮ”Confidence(T1โ†’T2) = E[H(D|T2)] โˆ’ H(D|T1) based on case profile Recommendation Add qEEG 68% โ†’ 81% +13pp After qEEG Add d4 MDD: 81% d6 GAD: 11% โ†‘ confidence Why qEEG? FAA biomarker is highly informative for depression vs anxiety Decision Diagnosis: MDD Confidence: 81% โ†’ Treatment plan TUR turns "more data is better" into a quantified, evidence-based clinical recommendation
๋„ 2.FIG. 2. Tier ์—…๊ทธ๋ ˆ์ด๋“œ ์ถ”์ฒœ๊ธฐ(TUR) ์ž‘๋™ ์˜ˆ์‹œ. Tier 1์—์„œ ์šฐ์šธ์ฆ 68% ์‹ ๋ขฐ๋„ โ†’ qEEG ์ถ”๊ฐ€ ๊ถŒ๊ณ  โ†’ Tier 2์—์„œ 81% ๋„๋‹ฌ. Example of the Tier Upgrade Recommender (TUR) in operation: at Tier 1, depression confidence is 68 %; the system recommends adding qEEG; at Tier 2, confidence reaches 81 %.

08์ฒญ๊ตฌํ•ญClaims

์ฒญ๊ตฌํ•ญ 1 (๋…๋ฆฝํ•ญ)Claim 1 (Independent)
ํ™˜์ž์— ๋Œ€ํ•œ DSM-5-TR ์ •์‹ ๊ณผ ์ง„๋‹จ์„ ๋‹ค์ค‘ ๋ฐ์ดํ„ฐ ์†Œ์Šค๋กœ๋ถ€ํ„ฐ ์‚ฐ์ถœํ•˜๋Š”, ์ปดํ“จํ„ฐ ๊ตฌํ˜„ ๋ฐฉ๋ฒ•์œผ๋กœ์„œ:
(a) ํ™˜์ž๋กœ๋ถ€ํ„ฐ ๋ณต์ˆ˜์˜ ์ด์งˆ์  ๋ฐ์ดํ„ฐ ์†Œ์Šค โ€” ์ž๊ธฐ๋ณด๊ณ  ์„ค๋ฌธ, ํƒ€์ดํ•‘ ํ–‰๋™ ๋ฐ์ดํ„ฐ, ์Šค๋งˆํŠธํฐ ์ˆ˜๋™ ์„ผ์„œ ๋ฐ์ดํ„ฐ, 19์ฑ„๋„ ์ •๋Ÿ‰๋‡ŒํŒŒ(qEEG), ์‹ฌ๋ฐ•๋ณ€์ด๋„(HRV), ์–ผ๊ตด ๋™์ž‘ ๋‹จ์œ„(FAU), ์Œ์„ฑ ์šด์œจ ๋ฐ ์‚ฌ๊ฑด๊ด€๋ จ์ „์œ„(ERP) ์ค‘ ์ ์–ด๋„ 4๊ฐœ โ€” ๋ฅผ ์ˆ˜์‹ ํ•˜๋Š” ๋‹จ๊ณ„;
(b) ๊ฐ ๋ฐ์ดํ„ฐ ์†Œ์Šค์— ๋Œ€ํ•ด ๋ณ„๊ฐœ์˜ ๋ชจ๋“ˆ์„ ์ ์šฉํ•˜์—ฌ, 18๊ฐœ DSM-5-TR ์ง„๋‹จ ํด๋ž˜์Šค ๊ฐ๊ฐ์— ๋Œ€ํ•œ ์šฐ๋„(likelihood)๋ฅผ ์‚ฐ์ถœํ•˜๋Š” ๋‹จ๊ณ„;
(c) ๊ฐ ๋ชจ๋“ˆ๋กœ๋ถ€ํ„ฐ ์‚ฐ์ถœ๋œ ์šฐ๋„๋ฅผ ๋ฒ ์ด์ง€์•ˆ ๋‹ค์ค‘ ๋ชจ๋“ˆ ์œตํ•ฉ ์—”์ง„์— ์ž…๋ ฅํ•˜์—ฌ, ์‚ฌ์ „ ํ™•๋ฅ ๊ณผ ๋ชจ๋“ˆ๋ณ„ ์‹ ๋ขฐ๋„ ๊ฐ€์ค‘์น˜์— ๊ธฐ๋ฐ˜ํ•œ ์‚ฌํ›„ ํ™•๋ฅ ์„ 18๊ฐœ ์ง„๋‹จ ํด๋ž˜์Šค ๊ฐ๊ฐ์— ๋Œ€ํ•ด ์‚ฐ์ถœํ•˜๋Š” ๋‹จ๊ณ„;
(d) ์‚ฐ์ถœ๋œ 18๊ฐœ ์‚ฌํ›„ ํ™•๋ฅ  ๋ถ„ํฌ์˜ ์„€๋„Œ ์—”ํŠธ๋กœํ”ผ๋กœ๋ถ€ํ„ฐ ์ง„๋‹จ ์‹ ๋ขฐ๋„๋ฅผ ์‚ฐ์ถœํ•˜๋˜, ํ•ด๋‹น ์‹ ๋ขฐ๋„๋Š” ํ™œ์„ฑํ™”๋œ ๋ฐ์ดํ„ฐ ์†Œ์Šค์˜ ์ˆ˜์— ๋Œ€ํ•œ ํ•จ์ˆ˜๋กœ์„œ 4๋‹จ๊ณ„์˜ ๊ณ„์ธต์  ์‹ ๋ขฐ๋„ ๋ฒ”์œ„ โ€” Tier 1: 60% ์ด์ƒ 75% ์ดํ•˜, Tier 2: 75% ์ด์ƒ 88% ์ดํ•˜, Tier 3: 88% ์ด์ƒ 95% ์ดํ•˜, Tier 4: 95% ์ด์ƒ 98% ์ดํ•˜ โ€” ์ค‘ ํ•˜๋‚˜์— ์†ํ•˜๋Š” ๋‹จ๊ณ„;
(e) ์ถ”๊ฐ€ ๋ฐ์ดํ„ฐ ์†Œ์Šค ํ†ตํ•ฉ ์‹œ ์˜ˆ์ƒ๋˜๋Š” ์‹ ๋ขฐ๋„ ์ƒ์Šน๊ฐ’์„ ์‚ฌ์ „ ์‚ฐ์ถœํ•˜๊ณ , ์ž„์ƒ๊ฐ€์—๊ฒŒ ๋‹ค์Œ Tier๋กœ์˜ ์—…๊ทธ๋ ˆ์ด๋“œ ๊ถŒ๊ณ ๋ฅผ ์ถœ๋ ฅํ•˜๋Š” ๋‹จ๊ณ„; ๋ฐ
(f) ๊ฐ€์žฅ ๋†’์€ ์‚ฌํ›„ ํ™•๋ฅ ์„ ๊ฐ€์ง€๋Š” ์ง„๋‹จ ๋ฐ ๊ทธ ์‹ ๋ขฐ๋„๋ฅผ ์ถœ๋ ฅํ•˜๋Š” ๋‹จ๊ณ„;
๋ฅผ ํฌํ•จํ•˜๋Š” ๋ฐฉ๋ฒ•.
A computer-implemented method for producing a DSM-5-TR psychiatric diagnosis for a patient from multiple data sources, the method comprising:
(a) receiving from the patient at least four of a plurality of heterogeneous data sources comprising self-report survey, typing-behavior data, smartphone-passive sensor data, 19-channel quantitative EEG (qEEG), heart-rate variability (HRV), Facial Action Units (FAU), vocal prosody, and event-related potentials (ERP);
(b) applying a distinct module to each data source to compute likelihoods for each of eighteen DSM-5-TR diagnostic classes;
(c) inputting the likelihoods computed by each module into a Bayesian Multi-Module Fusion engine, the engine computing posterior probabilities for each of the eighteen diagnostic classes based on prior probabilities and module-specific reliability weights;
(d) computing diagnostic confidence from the Shannon entropy of the resulting eighteen-class posterior distribution, said confidence falling within one of four hierarchical confidence bands as a function of the number of active data sources โ€” Tier 1: 60 % to 75 %; Tier 2: 75 % to 88 %; Tier 3: 88 % to 95 %; Tier 4: 95 % to 98 %;
(e) pre-computing the projected confidence gain from incorporating an additional data source, and outputting to a clinician a recommendation to upgrade to the next Tier; and
(f) outputting the diagnosis with the highest posterior probability together with its confidence value.
์ฒญ๊ตฌํ•ญ 2 (์ข…์†ํ•ญ)Claim 2 (Dependent)
์ฒญ๊ตฌํ•ญ 1์— ์žˆ์–ด์„œ, ๋‹จ๊ณ„ (c)์˜ ์‚ฌํ›„ ํ™•๋ฅ ์€ ๋‹ค์Œ ์‹์— ์˜ํ•ด ์‚ฐ์ถœ๋˜๋Š” ๊ฒƒ์„ ํŠน์ง•์œผ๋กœ ํ•˜๋Š” ๋ฐฉ๋ฒ•: P(d_i | data) โˆ P(d_i) ยท โˆ_{m=1}^{k} L_m(d_i | data_m)^(w_m), ์—ฌ๊ธฐ์„œ P(d_i)๋Š” ์ง„๋‹จ d_i์˜ ์‚ฌ์ „ ํ™•๋ฅ , L_m์€ ๋ชจ๋“ˆ m์˜ ์šฐ๋„, w_m์€ ๋ชจ๋“ˆ m์˜ ์‹ ๋ขฐ๋„ ๊ฐ€์ค‘์น˜, k๋Š” ํ™œ์„ฑ ๋ชจ๋“ˆ ์ˆ˜์ž„. ๋˜ํ•œ 18๊ฐœ ์ง„๋‹จ ํด๋ž˜์Šค์— ๋Œ€ํ•ด ์ •๊ทœํ™”๋œ๋‹ค. The method of Claim 1, wherein the posterior probability of step (c) is computed by P(d_i | data) โˆ P(d_i) ยท โˆ_{m=1}^{k} L_m(d_i | data_m)^(w_m), where P(d_i) is the prior probability of diagnosis d_i, L_m is the likelihood from module m, w_m is the reliability weight of module m, and k is the number of active modules; the result is normalized over the eighteen diagnostic classes.
์ฒญ๊ตฌํ•ญ 3 (์ข…์†ํ•ญ)Claim 3 (Dependent)
์ฒญ๊ตฌํ•ญ 1์— ์žˆ์–ด์„œ, ๋‹จ๊ณ„ (d)์˜ ์ง„๋‹จ ์‹ ๋ขฐ๋„๋Š” ๋ฐ์ดํ„ฐ ์†Œ์Šค๊ฐ€ ์ถ”๊ฐ€๋  ๋•Œ ๋‹จ์กฐ ๋น„์ฆ๊ฐ€์„ฑ(monotonic non-increase)์„ ๋งŒ์กฑํ•˜๋Š” ์‚ฌํ›„ ๋ถ„ํฌ์˜ ์„€๋„Œ ์—”ํŠธ๋กœํ”ผ H(D)์— ๊ธฐ๋ฐ˜ํ•˜๋ฉฐ, ์ฆ‰ H(D | Tier_n+1) โ‰ค H(D | Tier_n) ์˜ ์ˆ˜ํ•™์  ๋ณด์žฅ์„ ๋”ฐ๋ฅด๋Š” ๊ฒƒ์„ ํŠน์ง•์œผ๋กœ ํ•˜๋Š” ๋ฐฉ๋ฒ•. The method of Claim 1, wherein the diagnostic confidence of step (d) is based on the Shannon entropy H(D) of the posterior distribution that satisfies a monotonic-non-increase property as data sources are added, namely the mathematical guarantee H(D | Tier_n+1) โ‰ค H(D | Tier_n).
์ฒญ๊ตฌํ•ญ 4 (์ข…์†ํ•ญ)Claim 4 (Dependent)
์ฒญ๊ตฌํ•ญ 1์— ์žˆ์–ด์„œ, ๋‹จ๊ณ„ (a)์˜ 18๊ฐœ DSM-5-TR ์ง„๋‹จ ํด๋ž˜์Šค๋Š” ADHD ๋ถ€์ฃผ์˜ํ˜•, ADHD ๊ณผ์ž‰ํ–‰๋™/์ถฉ๋™ํ˜•, ADHD ๋ณตํ•ฉํ˜•, ์ฃผ์š”์šฐ์šธ์žฅ์• , ์ง€์†์„ฑ์šฐ์šธ์žฅ์• , ๋ฒ”๋ถˆ์•ˆ์žฅ์• , ๊ณตํ™ฉ์žฅ์• , ์‚ฌํšŒ๋ถˆ์•ˆ์žฅ์• , ์™ธ์ƒํ›„์ŠคํŠธ๋ ˆ์Šค์žฅ์• , ๊ฐ•๋ฐ•์žฅ์• , ์–‘๊ทน์„ฑI์žฅ์• , ์–‘๊ทน์„ฑII์žฅ์• , ์ˆœํ™˜์„ฑ์žฅ์• , ๋ถˆ๋ฉด์žฅ์• , ๋ฌผ์งˆ์‚ฌ์šฉ์žฅ์• , ์žํ์ŠคํŽ™ํŠธ๋Ÿผ์žฅ์• , ํŠน์ •ํ•™์Šต์žฅ์•  ๋ฐ ์ง„๋‹จ ๋ฏธํ•ด๋‹น์„ ํฌํ•จํ•˜๋Š” ๊ฒƒ์„ ํŠน์ง•์œผ๋กœ ํ•˜๋Š” ๋ฐฉ๋ฒ•. The method of Claim 1, wherein the eighteen DSM-5-TR diagnostic classes of step (a) comprise ADHD-Inattentive, ADHD-Hyperactive/Impulsive, ADHD-Combined, Major Depressive Disorder, Persistent Depressive Disorder, Generalized Anxiety Disorder, Panic Disorder, Social Anxiety Disorder, Posttraumatic Stress Disorder, Obsessive-Compulsive Disorder, Bipolar I Disorder, Bipolar II Disorder, Cyclothymic Disorder, Insomnia Disorder, Substance Use Disorder, Autism Spectrum Disorder, Specific Learning Disorder, and No-Diagnosis.
์ฒญ๊ตฌํ•ญ 5 (์ข…์†ํ•ญ)Claim 5 (Dependent)
์ฒญ๊ตฌํ•ญ 1์— ์žˆ์–ด์„œ, Tier 1์€ ์ž๊ธฐ๋ณด๊ณ  ์„ค๋ฌธ, ํƒ€์ดํ•‘ ๋ฐ์ดํ„ฐ, ์Šค๋งˆํŠธํฐ ์ˆ˜๋™ ๋ฐ์ดํ„ฐ๋ฅผ ํฌํ•จํ•˜๊ณ ; Tier 2๋Š” ์ถ”๊ฐ€๋กœ 19์ฑ„๋„ qEEG๋ฅผ ํฌํ•จํ•˜๊ณ ; Tier 3์€ ์ถ”๊ฐ€๋กœ HRV, ์–ผ๊ตด ๋™์ž‘ ๋‹จ์œ„ ๋ฐ ์Œ์„ฑ ์šด์œจ์„ ํฌํ•จํ•˜๊ณ ; Tier 4๋Š” ์ถ”๊ฐ€๋กœ ์‚ฌ๊ฑด๊ด€๋ จ์ „์œ„(ERP) ๋ฐ ์ข…๋‹จ์  ์ธก์ •์„ ํฌํ•จํ•˜๋Š” ๊ฒƒ์„ ํŠน์ง•์œผ๋กœ ํ•˜๋Š” ๋ฐฉ๋ฒ•. The method of Claim 1, wherein Tier 1 comprises self-report survey, typing data, and smartphone-passive data; Tier 2 additionally comprises 19-channel qEEG; Tier 3 additionally comprises HRV, Facial Action Units, and vocal prosody; and Tier 4 additionally comprises event-related potentials (ERP) and longitudinal measurements.
์ฒญ๊ตฌํ•ญ 6 (์ข…์†ํ•ญ)Claim 6 (Dependent)
์ฒญ๊ตฌํ•ญ 1์— ์žˆ์–ด์„œ, ๋‹จ๊ณ„ (e)์˜ Tier ์—…๊ทธ๋ ˆ์ด๋“œ ๊ถŒ๊ณ ๋Š” ProjectedGain(Tier_n โ†’ Tier_n+1) = E[Confidence(Tier_n+1)] โˆ’ Confidence(Tier_n)์˜ ์ •๋Ÿ‰์  ์‹ ๋ขฐ๋„ ์ƒ์Šน๊ฐ’์„ ์‚ฐ์ถœํ•˜๊ณ , ํ•ด๋‹น ์ƒ์Šน๊ฐ’์ด ๋ฏธ๋ฆฌ ์ •ํ•ด์ง„ ์ž„๊ณ„์น˜(์˜ˆ: 5 ๋ฐฑ๋ถ„์œจ ํฌ์ธํŠธ)๋ฅผ ์ดˆ๊ณผํ•˜๋Š” ๊ฒฝ์šฐ์—๋งŒ ์ž„์ƒ๊ฐ€์—๊ฒŒ ๊ถŒ๊ณ ๋ฅผ ์ถœ๋ ฅํ•˜๋Š” ๊ฒƒ์„ ํŠน์ง•์œผ๋กœ ํ•˜๋Š” ๋ฐฉ๋ฒ•. The method of Claim 1, wherein the Tier-upgrade recommendation of step (e) computes the quantitative confidence gain ProjectedGain(Tier_n โ†’ Tier_n+1) = E[Confidence(Tier_n+1)] โˆ’ Confidence(Tier_n), and outputs a recommendation to the clinician only when said gain exceeds a predefined threshold (for example, five percentage points).
์ฒญ๊ตฌํ•ญ 7 (์ข…์†ํ•ญ)Claim 7 (Dependent)
์ฒญ๊ตฌํ•ญ 1์— ์žˆ์–ด์„œ, ๋‹จ๊ณ„ (f)์˜ ์ถœ๋ ฅ์€ ๋‹จ์ผ ์ง„๋‹จ๋ฟ๋งŒ ์•„๋‹ˆ๋ผ, ์‚ฌํ›„ ํ™•๋ฅ  ์ž„๊ณ„์น˜(์˜ˆ: 30%)๋ฅผ ์ดˆ๊ณผํ•˜๋Š” ๋ชจ๋“  ์ง„๋‹จ์„ ๊ณต์กด(comorbid) ์ง„๋‹จ ํ›„๋ณด๋กœ ํ•จ๊ป˜ ์ถœ๋ ฅํ•˜๊ณ , ๊ฐ ์ง„๋‹จ๋ณ„๋กœ ์–ด๋–ค ๋ฐ์ดํ„ฐ ์†Œ์Šค๊ฐ€ ์–ด๋А ์ •๋„ ๊ธฐ์—ฌํ–ˆ๋Š”์ง€์˜ ๋ฐ์ดํ„ฐ ์†Œ์Šค๋ณ„ ๊ธฐ์—ฌ๋„(transparency report)๋ฅผ ํฌํ•จํ•˜๋Š” ๊ฒƒ์„ ํŠน์ง•์œผ๋กœ ํ•˜๋Š” ๋ฐฉ๋ฒ•. The method of Claim 1, wherein the output of step (f) includes not only a single diagnosis but also any diagnosis exceeding a posterior-probability threshold (for example, 30 %) as a candidate comorbid diagnosis, and includes a per-source contribution report (transparency report) indicating how much each data source contributed to each diagnosis.
์ฒญ๊ตฌํ•ญ 8 (์ข…์†ํ•ญ)Claim 8 (Dependent)
์ฒญ๊ตฌํ•ญ 1์— ์žˆ์–ด์„œ, ๋‹จ๊ณ„ (b)์˜ ๋ชจ๋“ˆ๋ณ„ ์šฐ๋„ ์‚ฐ์ถœ์€ ์‚ฌ์ „ ํ•™์Šต๋œ ๋จธ์‹ ๋Ÿฌ๋‹ ๋ชจ๋ธ, ์ •์ƒ ๋ชจ์ง‘๋‹จ ๋Œ€๋น„ Z-์ ์ˆ˜, ๋˜๋Š” ์ž„์ƒ ๊ฒ€์ฆ๋œ ๋ฃฐ ๊ธฐ๋ฐ˜ ๋งคํ•‘ ์ค‘ ํ•˜๋‚˜ ๋˜๋Š” ๊ทธ ์ด์ƒ์˜ ์กฐํ•ฉ์œผ๋กœ ์ˆ˜ํ–‰๋˜๋Š” ๊ฒƒ์„ ํŠน์ง•์œผ๋กœ ํ•˜๋Š” ๋ฐฉ๋ฒ•. The method of Claim 1, wherein the per-module likelihood computation of step (b) is performed by one or a combination of pre-trained machine-learning models, Z-scoring against a normative population, and clinically validated rule-based mappings.
์ฒญ๊ตฌํ•ญ 9 (๋…๋ฆฝํ•ญ โ€” ์‹œ์Šคํ…œ)Claim 9 (Independent โ€” System)
์ฒญ๊ตฌํ•ญ 1 ๋‚ด์ง€ 8 ์ค‘ ์–ด๋А ํ•œ ํ•ญ์˜ ๋ฐฉ๋ฒ•์„ ์ˆ˜ํ–‰ํ•˜๊ธฐ ์œ„ํ•œ, ์ ์–ด๋„ ํ•˜๋‚˜์˜ ํ”„๋กœ์„ธ์„œ, 8๊ฐœ ์ด์งˆ์  ๋ฐ์ดํ„ฐ ์†Œ์Šค๋ฅผ ์ˆ˜์‹ ํ•˜๋Š” ๋‹ค์ค‘ ๋ชจ๋‹ฌ ์ž…๋ ฅ ์ธํ„ฐํŽ˜์ด์Šค, ๋ฐ ์ƒ๊ธฐ ํ”„๋กœ์„ธ์„œ์— ์˜ํ•ด ์‹คํ–‰๋˜๋Š” ๋ช…๋ น์–ด๋ฅผ ์ €์žฅํ•˜๋Š” ๋น„์ผ์‹œ์  ์ปดํ“จํ„ฐ ํŒ๋… ๊ฐ€๋Šฅ ์ €์žฅ ๋งค์ฒด๋ฅผ ํฌํ•จํ•˜๋Š” 4-Tier Progressive ๋‹ค์ค‘ ์†Œ์Šค ์ง„๋‹จ ์‹œ์Šคํ…œ. A 4-Tier Progressive multi-source diagnostic system for performing the method of any one of Claims 1 through 8, the system comprising at least one processor, a multi-modal input interface for receiving the eight heterogeneous data sources, and a non-transitory computer-readable storage medium storing instructions executable by the processor.
์ฒญ๊ตฌํ•ญ 10 (๋…๋ฆฝํ•ญ โ€” ๋งค์ฒด)Claim 10 (Independent โ€” Medium)
์ปดํ“จํ„ฐ์— ์˜ํ•ด ์‹คํ–‰๋  ๋•Œ ์ฒญ๊ตฌํ•ญ 1 ๋‚ด์ง€ 8 ์ค‘ ์–ด๋А ํ•œ ํ•ญ์˜ ๋ฐฉ๋ฒ•์„ ์ˆ˜ํ–‰ํ•˜๋„๋ก ํ•˜๋Š” ๋ช…๋ น์–ด๋ฅผ ์ €์žฅํ•˜๋Š” ๋น„์ผ์‹œ์  ์ปดํ“จํ„ฐ ํŒ๋… ๊ฐ€๋Šฅ ์ €์žฅ ๋งค์ฒด. A non-transitory computer-readable storage medium storing instructions which, when executed by a computer, cause the computer to perform the method of any one of Claims 1 through 8.

09์„ ํ–‰ ๊ธฐ์ˆ  ๋น„๊ตPrior Art Comparison

์„ ํ–‰ ๊ธฐ์ˆ Prior Art ์ ‘๊ทผ ๋ฐฉ์‹Approach ํ•œ๊ณ„Limitation ๋ณธ ๋ฐœ๋ช…๊ณผ์˜ ์ฐจ์ดDistinction
PHQ-9 / GAD-7 / ASRS
(self-report scales)
๋‹จ์ผ ์ž๊ธฐ๋ณด๊ณ  ์„ค๋ฌธsingle self-report survey ๊ฐ๊ด€ ๋ฐ”์ด์˜ค๋งˆ์ปค ๋ฏธํ†ตํ•ฉno objective biomarker integration 8 ๋ฐ์ดํ„ฐ ์†Œ์Šค ํ†ตํ•ฉintegration of eight data sources
Mindstrong / Ginger.io ์Šค๋งˆํŠธํฐ ์ˆ˜๋™ ๋ฐ์ดํ„ฐ๋งŒsmartphone-passive only qEEG/HRV/์–ผ๊ตด/์Œ์„ฑ ๋ฏธ์‚ฌ์šฉno qEEG, HRV, face, or voice 8 ๋ชจ๋‹ฌ ํ†ตํ•ฉ + DSM-5-TR ์ง„๋‹จeight modalities and DSM-5-TR diagnosis
EEGuide / qEEGPro
(qEEG software)
qEEG ๋‹จ๋… ๋ถ„์„qEEG-only analysis ์ž๊ธฐ๋ณด๊ณ /์–ผ๊ตด/์Œ์„ฑ ๋ฏธํ†ตํ•ฉno self-report, face, or voice qEEG + 7 ์ถ”๊ฐ€ ์†Œ์Šค ์œตํ•ฉqEEG plus seven additional sources
Affectiva / iMotions ํ‘œ์ • + ์ƒ์ฒด ํ†ตํ•ฉface plus biometric integration DSM-5-TR ์ง„๋‹จ ๋ฏธ์‚ฐ์ถœno DSM-5-TR diagnosis output 18 DSM-5-TR ์ง„๋‹จ ์‚ฌํ›„ ํ™•๋ฅ posterior probabilities for 18 DSM-5-TR diagnoses
IBM Watson Health
(general clinical AI)
์ „์ž์˜๋ฌด๊ธฐ๋ก NLPelectronic medical record NLP ์ •์‹ ๊ณผ ํŠนํ™” ๋ถ€์žฌ, ๋‹ค์ค‘ ๋ชจ๋‹ฌ ๋ถ€์žฌno psychiatric specialization; no multi-modal ์ •์‹ ๊ณผ ํŠนํ™” + 8 ๋ชจ๋‹ฌ ์œตํ•ฉpsychiatry-specific + eight-modality fusion
RDoC Framework
(NIMH 2014)
์ฐจ์›์  ์ •์‹ ๊ฑด๊ฐ• ํ‰๊ฐ€ ์ด๋ก dimensional mental health framework ์•Œ๊ณ ๋ฆฌ์ฆ˜ ๊ตฌํ˜„ ๋ถ€์žฌno algorithmic implementation RDoC ์›๋ฆฌ์˜ ์•Œ๊ณ ๋ฆฌ์ฆ˜ ๊ตฌํ˜„algorithmic implementation of RDoC principles
๐ŸŽฏ ๋ฐœ๋ช…์˜ ์ง„๋ณด์„ฑ๐ŸŽฏ Inventive Step

๋ณธ ๋ฐœ๋ช…์€ (1) 8๊ฐœ ์ด์งˆ์  ๋ฐ์ดํ„ฐ ์†Œ์Šค๋ฅผ ๋ฒ ์ด์ง€์•ˆ ๋‹ค์ค‘ ๋ชจ๋“ˆ ์œตํ•ฉ์œผ๋กœ ํ†ตํ•ฉํ•œ ์ตœ์ดˆ์˜ ์ •์‹ ๊ณผ ์ง„๋‹จ ์‹œ์Šคํ…œ์ด๋ฉฐ, (2) 18๊ฐœ DSM-5-TR ์ง„๋‹จ ๋™์‹œ ์‚ฌํ›„ ํ™•๋ฅ  ์‚ฐ์ถœ, (3) ๋ฐ์ดํ„ฐ ์ถ”๊ฐ€ ์‹œ ์‹ ๋ขฐ๋„ ๋‹จ์กฐ ์ฆ๊ฐ€์˜ ์ˆ˜ํ•™์  ๋ณด์žฅ(H(D|Tier_n+1) โ‰ค H(D|Tier_n)), (4) Tier ์—…๊ทธ๋ ˆ์ด๋“œ์˜ ์ •๋Ÿ‰์  ๊ถŒ๊ณ  ์ œ๊ณต์ด๋ผ๋Š” 4๊ฐ€์ง€ ๊ฒฐํ•ฉ์„ ๊ฐ–๋Š”๋‹ค. ์ด ๊ฒฐํ•ฉ์€ ์„ ํ–‰ ๊ธฐ์ˆ ์— ์กด์žฌํ•˜์ง€ ์•Š์œผ๋ฉฐ, ์ •์‹ ๊ณผ ์ง„๋‹จ์˜ ๊ฐ๊ด€์„ฑ๊ณผ ์ •ํ™•์„ฑ์„ ๋น„์•ฝ์ ์œผ๋กœ ํ–ฅ์ƒ์‹œํ‚จ๋‹ค. The present invention possesses four combined elements: (1) it is the first psychiatric diagnostic system to integrate eight heterogeneous data sources via Bayesian Multi-Module Fusion; (2) it produces simultaneous posterior probabilities for eighteen DSM-5-TR diagnoses; (3) it mathematically guarantees monotonically non-decreasing confidence as data is added (H(D|Tier_n+1) โ‰ค H(D|Tier_n)); and (4) it provides quantitative recommendations for Tier upgrades. This combination does not exist in the prior art and dramatically improves the objectivity and accuracy of psychiatric diagnosis.

10์‚ฐ์—…์ƒ ์ด์šฉ ๊ฐ€๋Šฅ์„ฑIndustrial Applicability

10.1 ์ ์šฉ ์‹œ์žฅ10.1 Target Markets

10.2 ๊ทœ์ œ ๊ฒฝ๋กœ10.2 Regulatory Pathway

FDA 510(k) Class II ์˜๋ฃŒ๊ธฐ๊ธฐ๋กœ์˜ ๋“ฑ์žฌ๊ฐ€ ๊ฐ€๋Šฅํ•˜๋ฉฐ, ์ถ”๊ฐ€ ์ž„์ƒ ์‹œํ—˜์„ ํ†ตํ•œ PMA(Premarket Approval) ํ›„ ์ง„๋‹จ ๋ณด์กฐ ๋„๊ตฌ๋กœ ์ž„์ƒ ์‚ฌ์šฉ ๊ฐ€๋Šฅ. EU์—์„œ๋Š” MDR Class IIa ์˜๋ฃŒ๊ธฐ๊ธฐ๋กœ ๋ถ„๋ฅ˜ ๊ฐ€๋Šฅ. The system is amenable to FDA 510(k) Class II medical-device clearance and, following PMA via additional clinical trials, may be used clinically as a diagnostic-aid tool. In the European Union, classification as an MDR Class IIa medical device is feasible.

10.3 BCN+PhD ๋…์ ์„ฑ10.3 BCN-Plus-PhD Exclusivity

๋ณธ ๋ฐœ๋ช…์˜ ๋ชจ๋“ˆ๋ณ„ ์šฐ๋„ ํ•จ์ˆ˜์™€ ์‹ ๋ขฐ๋„ ๊ฐ€์ค‘์น˜๋Š” Board Certified in Neurofeedback (BCN) + PhD + 3๋…„ ์ž„์ƒ ๋ฐ์ดํ„ฐ๋ฅผ ๊ฒฐํ•ฉํ•œ ์˜์—… ๋น„๋ฐ€์ด๋ฉฐ, ์ผ๋ฐ˜ AI/ML ์—”์ง€๋‹ˆ์–ด๊ฐ€ ๋ชจ๋ฐฉํ•˜๊ธฐ ๋งค์šฐ ์–ด๋ ต๋‹ค. ๋ฐœ๋ช…์ž์˜ Harvard Mind, Brain & Education ๋ฐฉ๋ฌธ ํ•™์ž ์ž๊ฒฉ(Kurt Fischer ์‚ฌ์‚ฌ)๊ณผ Instructional Design PhD ๊ฒฐํ•ฉ์€ ๋ณธ ๋ฐœ๋ช…์˜ ์•Œ๊ณ ๋ฆฌ์ฆ˜ ์„ค๊ณ„ ์‹ ๋ขฐ์„ฑ์„ ํ•™์ˆ ์ ์œผ๋กœ ๋’ท๋ฐ›์นจํ•œ๋‹ค. The per-module likelihood functions and reliability weights of this invention are trade secrets combining Board Certified in Neurofeedback (BCN), PhD, and three years of clinical data; they are exceedingly difficult for general AI/ML engineers to replicate. The inventor's Harvard Graduate School of Education Visiting Scholar credentials (mentored by Kurt Fischer) and Instructional Design PhD academically reinforce the design reliability of the algorithm.

11๊ด€๋ จ ๋…ผ๋ฌธReferences

๋ณธ ๋ฐœ๋ช…์˜ ์ด๋ก ์ ยท์ž„์ƒ์  ๊ทผ๊ฑฐ๊ฐ€ ๋˜๋Š” ํ•ต์‹ฌ ๋…ผ๋ฌธ ๋ฐ ์ž๋ฃŒ. ํด๋ฆญํ•˜๋ฉด ์™ธ๋ถ€ ์ถœ์ฒ˜๋กœ ์ด๋™ํ•ฉ๋‹ˆ๋‹ค. Key papers and resources providing the theoretical and clinical basis for this invention. Click links to access external sources.

A. DSM-5-TR ๋ฐ RDoC ํ”„๋ ˆ์ž„์›ŒํฌA. DSM-5-TR and RDoC Framework
B. ๋ฒ ์ด์ง€์•ˆ ์ง„๋‹จ ์ถ”๋ก B. Bayesian Diagnostic Inference
C. QEEG ์ž„์ƒ ์‘์šฉC. QEEG Clinical Applications
D. ๋””์ง€ํ„ธ ํ‘œํ˜„ํ˜• (Digital Phenotyping)D. Digital Phenotyping
E. HRV ๋ฐ ์ž์œจ์‹ ๊ฒฝ๊ณ„E. HRV and Autonomic Nervous System
F. ์–ผ๊ตดยท์Œ์„ฑ ์ •์„œ ๋ถ„์„F. Facial and Vocal Affect Analysis
G. ํƒ€์ดํ•‘ ํ–‰๋™ ํŒจํ„ด (Behavioral Biometrics)G. Typing Behavior Patterns (Behavioral Biometrics)
H. ERP ์ž„์ƒ ์‘์šฉH. ERP Clinical Applications
I. FDA / ๊ทœ์ œ ๊ฐ€์ด๋“œ๋ผ์ธI. FDA / Regulatory Guidelines