Deep DiveMedicine & Health
AI Predicts Patient Deterioration 17 Hours Before the Crisis
A deep learning model trained on 888 inpatient visits using continuous wearable vital signs predicted adverse clinical outcomes up to 17 hours in advance with 81.8% accuracy. Published in Nature Communications with prospective validation.
By Sean K.S. Shin
This blog summarizes research trends based on published paper abstracts. Specific numbers or findings may contain inaccuracies. For scholarly rigor, always consult the original papers cited in each post.
Hospital patients deteriorate in patterns. Heart rate drifts upward hours before sepsis declares itself. Respiratory rate subtly increases before a patient needs intubation. Blood pressure variability changes before hemodynamic collapse. These patterns exist in the data β the challenge is that they are too subtle and too embedded in noise for human clinicians monitoring dozens of patients simultaneously to detect reliably. Scheid et al. (2025), published in Nature Communications, trained a deep learning model on continuous wearable sensor data from hospitalized patients and demonstrated that adverse outcomes could be predicted up to 17 hours before they occurred.
The Research Landscape
The Problem of Delayed Detection
Current hospital monitoring relies on periodic vital sign checks (every 4-8 hours on general wards) and threshold-based alarms in higher-acuity settings. Both approaches have fundamental limitations:
Periodic checks miss trends β spot measurements cannot capture the trajectories that signal evolving deterioration. Threshold alarms generate fatigue β nurses face hundreds of alerts per shift, most clinically insignificant. Early warning scores are blunt instruments β calculated from spot-check data, they detect deterioration already underway rather than its subtle precursors.
Continuous Wearable Monitoring
Continuous wearable sensors change the data landscape fundamentally. Rather than 3-6 data points per day for each vital sign, a wearable generates thousands β capturing not just values but variability, trends, circadian patterns, and subtle physiological dynamics that are invisible in spot-check data.
The Scheid et al. study used data from 888 inpatient visits where patients wore continuous monitoring devices that captured heart rate, respiratory rate, oxygen saturation, and other physiological signals at high temporal resolution throughout their hospital stay.
Model Architecture and Results
The deep learning model processed continuous streams of wearable vital sign data, learning temporal patterns associated with subsequent adverse outcomes. The key results:
Prediction window: up to 17 hours. The model identified patients at risk of adverse clinical outcomes (including clinical deterioration events requiring escalation of care) up to 17 hours before the event occurred. This lead time is clinically transformative β 17 hours is enough time to order labs, start antibiotics, adjust fluid management, transfer to a higher level of care, or activate a rapid response team before the patient is in crisis.
Accuracy: 81.8%. The model achieved 81.8% accuracy in predicting adverse outcomes. This metric should be interpreted carefully. In the context of clinical prediction models, 81.8% is a strong result for a prospectively validated model. However, accuracy alone does not capture the full clinical picture β sensitivity (how many true deteriorations are caught), specificity (how many false alarms are generated), and positive predictive value (when the model alarms, how often is it right) are equally important for clinical deployment.
Prospective validation. The study included prospective validation β testing the model on patients whose data were collected after the model was developed. This is a critical methodological feature that many clinical AI studies lack. Retrospective-only studies are prone to data leakage, temporal confounding, and overfitting to historical patterns that may not generalize forward.
Publication venue. Nature Communications provides rigorous peer review for this type of interdisciplinary work, bridging machine learning methodology and clinical validation.
How This Differs from Existing Approaches
The distinction from conventional early warning systems is threefold. First, the model uses continuous rather than intermittent data, capturing temporal dynamics that spot-check scores miss. Second, it learns nonlinear multivariate patterns rather than applying fixed thresholds to individual parameters. Third, the 17-hour prediction window extends substantially beyond what scores like NEWS typically achieve (which detect deterioration already in progress, offering minutes to low single-digit hours of warning).
Critical Analysis: Claims and Evidence
<
| Claim | Source | Evidence Level | Verdict |
|---|
| Model predicts adverse outcomes up to 17 hours in advance | Scheid et al. (2025) | Prospectively validated | β
Reported with prospective validation |
| Accuracy of 81.8% | Scheid et al. (2025) | Single-site study | β
Reported; generalizability to other sites untested |
| Continuous wearable data captures dynamics missed by spot checks | Study rationale | Well-established principle | β
Supported by prior literature |
| Model outperforms conventional early warning scores | Implied comparison | Not directly head-to-head in this study | β οΈ Likely true but formal comparison needed |
| Approach is ready for clinical deployment | Not claimed | Implementation barriers exist | β οΈ Validation is necessary but not sufficient for deployment |
Open Questions
Generalizability across hospitals. The model was developed and validated at a single institution. Patient populations, clinical workflows, wearable device brands, nursing practices, and electronic health record systems vary substantially across hospitals. Multi-site validation is essential before broad deployment.False alarm rate. An 81.8% accuracy figure does not specify the false positive rate. If the model generates one false alarm for every true alarm, clinical adoption will face the same alarm fatigue problem it aims to solve. The trade-off between sensitivity (catching every deterioration) and specificity (avoiding false alarms) must be tuned for the clinical context.Clinical workflow integration. A prediction is only useful if it reaches the right clinician at the right time and triggers an appropriate response. How should a 17-hour-ahead alert be presented? To whom? With what recommended actions? The human factors and implementation science aspects are as challenging as the machine learning.Patient populations. Hospital patients are heterogeneous. A model trained primarily on surgical patients may not perform well on medical patients, and vice versa. Performance stratification by diagnosis, age, baseline acuity, and comorbidity profile is needed.Wearable tolerance and data quality. Hospitalized patients may remove wearable sensors during bathing, imaging, or procedures, creating data gaps. Patient movement generates motion artifacts. The model must be robust to missing and noisy data in real-world deployment.What This Means for Hospital Medicine
The Scheid et al. study demonstrates that deep learning applied to continuous wearable vital sign data can predict patient deterioration with a lead time that is clinically actionable. Seventeen hours is not a marginal improvement over existing approaches β it represents a qualitative shift from reactive to proactive care.
The path from a validated prediction model to a deployed clinical system, however, involves challenges that are not primarily technical. They are organizational (integrating alerts into existing workflows), human (ensuring clinicians trust and act on AI-generated alerts), regulatory (FDA clearance for clinical decision support), and economic (demonstrating that earlier intervention reduces costs and improves outcomes enough to justify the monitoring infrastructure).
The study is best understood as a strong proof of concept that the signal exists in continuous wearable data and that deep learning can extract it. The next phase β multi-site validation, implementation trials, and outcome studies β will determine whether the prediction translates into better patient outcomes.
Explore related digital health and clinical AI research through ORAA ResearchBrain.
λ©΄μ±
μ‘°ν: μ΄ κ²μλ¬Όμ μ 보 μ 곡 λͺ©μ μΌλ‘ λ¨μΌ μμ AI μ°κ΅¬λ₯Ό κ²ν νλ€. μμΈ‘ λͺ¨λΈμ μ±λ₯μ μ§λ¨κ³Ό νκ²½μ λ°λΌ λ¬λΌμ§ μ μλ€. λͺ¨λ μ£Όμ₯μ μΈμ© μ μ μλ³Έ μΆνλ¬Όκ³Ό λμ‘°νμ¬ κ²μ¦ν΄μΌ νλ€.
AI, μκΈ° λ°μ 17μκ° μ μ νμ μν μ
νλ₯Ό μμΈ‘νλ€
μ
μ νμμ μν μ
νλ μΌμ ν ν¨ν΄μ λ°λ₯Έλ€. ν¨νμ¦μ΄ λͺ
νν λνλκΈ° μ μκ° μ λΆν° μ¬λ°μκ° μμν μμΉνλ€. νμκ° κΈ°κ΄ μ½κ΄μ νμλ‘ νκΈ° μ μ νΈν‘μκ° λ―ΈμΈνκ² μ¦κ°νλ€. νμνμ λΆκ΄΄κ° μΌμ΄λκΈ° μ μ νμ λ³λμ±μ΄ λ³ννλ€. μ΄λ¬ν ν¨ν΄μ λ°μ΄ν° μμ μ‘΄μ¬νμ§λ§, μμ λͺ
μ νμλ₯Ό λμμ λͺ¨λν°λ§νλ μμμκ° μμ μ μΌλ‘ κ°μ§νκΈ°μλ λ무 λ―Έλ¬νκ³ μ‘μ μμ κΉμ΄ 묻ν μλ€λ κ²μ΄ λ¬Έμ μ΄λ€. Scheid et al. (2025)μ΄ Nature Communicationsμ λ°νν μ°κ΅¬λ μ
μ νμμ μ°μ μ¨μ΄λ¬λΈ μΌμ λ°μ΄ν°λ₯Ό μ΄μ©ν΄ λ₯λ¬λ λͺ¨λΈμ νλ ¨μμΌ, μ΄μ κ²°κ³Όκ° λ°μνκΈ° μ΅λ 17μκ° μ μ μ΄λ₯Ό μμΈ‘ν μ μμμ μ
μ¦νμλ€.
μ°κ΅¬ λ°°κ²½
μ§μ° κ°μ§μ λ¬Έμ
νμ¬ λ³μ λͺ¨λν°λ§μ (μΌλ° λ³λμμ) μ£ΌκΈ°μ μΈ νλ ₯μ§ν μΈ‘μ (4~8μκ° κ°κ²©)κ³Ό κ³ μ€μ¦λ νκ²½μμμ μκ³κ° κΈ°λ° κ²½λ³΄μ μμ‘΄νλ€. λ μ κ·Ό λ°©μ λͺ¨λ κ·Όλ³Έμ μΈ νκ³λ₯Ό μ§λλ€.
μ£ΌκΈ°μ μΈ‘μ μ μΆμΈλ₯Ό λμΉλ€ β λ¨νΈμ μΈ‘μ κ°μ μ
νκ° μ§νλκ³ μμμ μ리λ κΆ€μ μ ν¬μ°©ν μ μλ€. μκ³κ° 경보λ 경보 νΌλ‘λ₯Ό μ λ°νλ€ β κ°νΈμ¬λ κ΅λ 근무 μ€ μλ°± 건μ 경보λ₯Ό λ°μΌλ©°, κ·Έ λλΆλΆμ μμμ μΌλ‘ μ μλ―Ένμ§ μλ€. μ‘°κΈ° 경보 μ μλ μ‘°μ‘ν λꡬμ΄λ€ β λ¨νΈμ μΈ‘μ λ°μ΄ν°λ‘ μ°μΆλλ―λ‘, λ―Έλ¬ν μ μ‘° μ§ν보λ€λ μ΄λ―Έ μ§ν μ€μΈ μ
νλ₯Ό κ°μ§νλ€.
μ°μ μ¨μ΄λ¬λΈ λͺ¨λν°λ§
μ°μ μ¨μ΄λ¬λΈ μΌμλ λ°μ΄ν° νκ²½μ κ·Όλ³Έμ μΌλ‘ λ³νμν¨λ€. κ° νλ ₯μ§νμ λν΄ ν루 3~6κ°μ λ°μ΄ν° ν¬μΈνΈκ° μλ μμ² κ°λ₯Ό μμ±νμ¬, λ¨μ μμΉλΏλ§ μλλΌ λ³λμ±, μΆμΈ, μΌμ£ΌκΈ° ν¨ν΄, κ·Έλ¦¬κ³ λ¨νΈμ μΈ‘μ λ°μ΄ν°μμλ 보μ΄μ§ μλ λ―Έλ¬ν μ리μ μνκΉμ§ ν¬μ°©νλ€.
Scheid et al.μ μ°κ΅¬λ νμκ° μ
μ κΈ°κ° λ΄λ΄ μ°μ λͺ¨λν°λ§ κΈ°κΈ°λ₯Ό μ°©μ©ν 888건μ μ
μ λ°©λ¬Έ λ°μ΄ν°λ₯Ό μ¬μ©νμλ€. ν΄λΉ κΈ°κΈ°λ μ¬λ°μ, νΈν‘μ, μ°μ ν¬νλ λ° κΈ°ν μ리μ μ νΈλ₯Ό λμ μκ° ν΄μλλ‘ μμ§νμλ€.
λͺ¨λΈ ꡬ쑰 λ° κ²°κ³Ό
λ₯λ¬λ λͺ¨λΈμ μ¨μ΄λ¬λΈ νλ ₯μ§νμ μ°μ λ°μ΄ν° μ€νΈλ¦Όμ μ²λ¦¬νμ¬ μ΄νμ μ΄μ κ²°κ³Όμ κ΄λ ¨λ μκ°μ ν¨ν΄μ νμ΅νμλ€. μ£Όμ κ²°κ³Όλ λ€μκ³Ό κ°λ€.
μμΈ‘ κ°λ₯ μκ°: μ΅λ 17μκ°. λͺ¨λΈμ μ΄μ μμ κ²°κ³Ό(μΉλ£ κ°νκ° νμν μμμ μ
ν μ¬κ±΄ ν¬ν¨)κ° λ°μνκΈ° μ΅λ 17μκ° μ μ μν νμλ₯Ό μλ³νμλ€. μ΄ μ ν μκ°μ μμμ μΌλ‘ λ§€μ° μ€μνλ€ β 17μκ°μ κ²μ¬λ₯Ό μ§μνκ³ , νμμ λ₯Ό μμνλ©°, μμ‘ κ΄λ¦¬λ₯Ό μ‘°μ νκ³ , μμ μΉλ£ λ¨κ³λ‘ μ ννκ±°λ, νμκ° μκΈ°μ μ²νκΈ° μ μ μ μ λμνμ κ°λνκΈ°μ μΆ©λΆν μκ°μ΄λ€.
μ νλ: 81.8%. λͺ¨λΈμ μ΄μ κ²°κ³Ό μμΈ‘μμ 81.8%μ μ νλλ₯Ό λ¬μ±νμλ€. μ΄ μμΉλ μ μ€νκ² ν΄μν΄μΌ νλ€. μμ μμΈ‘ λͺ¨λΈμ λ§₯λ½μμ 81.8%λ μ ν₯μ μΌλ‘ κ²μ¦λ λͺ¨λΈλ‘μ μ°μν κ²°κ³Όμ΄λ€. κ·Έλ¬λ μ νλλ§μΌλ‘λ μ 체μ μΈ μμ μν©μ νμ
ν μ μλ€ β λ―Όκ°λ(μ€μ μ
νλ₯Ό μΌλ§λ λ§μ΄ κ°μ§νλκ°), νΉμ΄λ(μΌλ§λ λ§μ μμμ± κ²½λ³΄κ° λ°μνλκ°), μμ± μμΈ‘λ(λͺ¨λΈμ΄ 경보λ₯Ό μΈλ¦΄ λ μ€μ λ‘ λ§λ λΉλ) μμ μμ μ μ©μ μν΄ λλ±νκ² μ€μνλ€.
μ ν₯μ κ²μ¦. μ΄ μ°κ΅¬μλ μ ν₯μ κ²μ¦μ΄ ν¬ν¨λμλλ°, μ΄λ λͺ¨λΈμ΄ κ°λ°λ μ΄νμ μμ§λ νμ λ°μ΄ν°λ‘ λͺ¨λΈμ ν
μ€νΈνλ κ²μ΄λ€. μ΄λ λ§μ μμ AI μ°κ΅¬κ° κ²°μ¬νκ³ μλ μ€μν λ°©λ²λ‘ μ νΉμ§μ΄λ€. νν₯μ μ°κ΅¬λ§μΌλ‘λ λ°μ΄ν° λμΆ, μκ°μ κ΅λ, κ·Έλ¦¬κ³ λ―Έλμ μΌλ°νλμ§ μμ μ μλ μμ¬μ ν¨ν΄μ λν κ³Όμ ν©μ΄ λ°μνκΈ° μ½λ€.
κ²μ¬ νμ μ§. Nature Communicationsλ μ΄λ¬ν μ νμ νμ κ° μ°κ΅¬μ λν΄ μ격ν λλ£ μ¬μ¬λ₯Ό μ 곡νλ©°, κΈ°κ³ νμ΅ λ°©λ²λ‘ κ³Ό μμ κ²μ¦μ μ°κ²°νλ€.
κΈ°μ‘΄ μ κ·Όλ²κ³Όμ μ°¨λ³μ
κΈ°μ‘΄ μ‘°κΈ° κ²½κ³ μμ€ν
κ³Όμ μ°¨λ³μ μ μΈ κ°μ§μ΄λ€. 첫째, μ΄ λͺ¨λΈμ κ°νμ λ°μ΄ν°κ° μλ μ°μμ λ°μ΄ν°λ₯Ό μ¬μ©νμ¬, μΌνμ± μ μ μΈ‘μ μ΄ λμΉλ μκ°μ μνμ ν¬μ°©νλ€. λμ§Έ, κ°λ³ νλΌλ―Έν°μ κ³ μ λ μκ³κ°μ μ μ©νλ λ°©μμ΄ μλλΌ λΉμ ν λ€λ³λ ν¨ν΄μ νμ΅νλ€. μ
μ§Έ, 17μκ°μ μμΈ‘ μ°½μ NEWSμ κ°μ μ μ 체κ³κ° μΌλ°μ μΌλ‘ λ¬μ±νλ μμ€(μ΄λ―Έ μ§ν μ€μΈ μ
νλ₯Ό κ°μ§νμ¬ μ λΆμμ ν μλ¦Ώμ μ΄λ° μκ°λμ κ²½κ³ λ₯Ό μ 곡νλ μμ€)μ ν¬κ² λμ΄μ λ€.
λΉνμ λΆμ: μ£Όμ₯κ³Ό κ·Όκ±°
<
| μ£Όμ₯ | μΆμ² | κ·Όκ±° μμ€ | νκ° |
|---|
| λͺ¨λΈμ΄ μ΅λ 17μκ° μ μ μ΄μ κ²°κ³Όλ₯Ό μμΈ‘νλ€ | Scheid et al. (2025) | μ ν₯μ κ²μ¦ μλ£ | β
μ ν₯μ κ²μ¦κ³Ό ν¨κ» λ³΄κ³ λ¨ |
| μ νλ 81.8% | Scheid et al. (2025) | λ¨μΌ κΈ°κ΄ μ°κ΅¬ | β
λ³΄κ³ λ¨; ν κΈ°κ΄μΌλ‘μ μΌλ°ν κ°λ₯μ±μ λ―Έκ²μ¦ |
| μ°μ μ¨μ΄λ¬λΈ λ°μ΄ν°κ° μΌνμ± μΈ‘μ μ΄ λμΉλ μνμ ν¬μ°©νλ€ | μ°κ΅¬ κ·Όκ±° | μΆ©λΆν ν립λ μμΉ | β
μ ν λ¬Ένμ μν΄ μ§μ§λ¨ |
| λͺ¨λΈμ΄ κΈ°μ‘΄ μ‘°κΈ° κ²½κ³ μ μλ₯Ό λ₯κ°νλ€ | μ묡μ λΉκ΅ | λ³Έ μ°κ΅¬μμ μ§μ μ μΈ λλ©΄ λΉκ΅ μμ | β οΈ μ¬μ€μΌ κ°λ₯μ±μ΄ λμΌλ 곡μμ λΉκ΅ νμ |
| λ³Έ μ κ·Όλ²μ΄ μμ νμ₯ μ μ©μ μ€λΉλμ΄ μλ€ | μ£Όμ₯λμ§ μμ | ꡬνμ μ₯λ²½μ΄ μ‘΄μ¬ν¨ | β οΈ κ²μ¦μ νμ쑰건μ΄λ λ°°ν¬λ₯Ό μν μΆ©λΆμ‘°κ±΄μ μλ |
λ―Έν΄κ²° κ³Όμ
λ³μ κ° μΌλ°ν κ°λ₯μ±. μ΄ λͺ¨λΈμ λ¨μΌ κΈ°κ΄μμ κ°λ° λ° κ²μ¦λμλ€. νμ μ§λ¨, μμ μν¬νλ‘μ°, μ¨μ΄λ¬λΈ κΈ°κΈ° λΈλλ, κ°νΈ κ΄ν, μ μ μλ¬΄κΈ°λ‘ μμ€ν
μ λ³μλ§λ€ μλΉν λ€λ₯΄λ€. κ΄λ²μν λ°°ν¬ μ μ λ€κΈ°κ΄ κ²μ¦μ΄ νμμ μ΄λ€.μ€κ²½λ³΄μ¨. 81.8%μ μ νλ μμΉλ μμμ±λ₯ μ λͺ
μνμ§ μλλ€. λͺ¨λΈμ΄ μ€μ 경보 νλλΉ μ€κ²½λ³΄λ₯Ό νλμ© λ°μμν¨λ€λ©΄, μμ λμ
μ λͺ¨λΈμ΄ ν΄κ²°νκ³ μ νλ 경보 νΌλ‘ λ¬Έμ μ μ§λ©΄νκ² λλ€. λ―Όκ°λ(λͺ¨λ μ
ν ν¬μ°©)μ νΉμ΄λ(μ€κ²½λ³΄ ννΌ) κ°μ κ· νμ μμμ λ§₯λ½μ λ§κ² μ‘°μ λμ΄μΌ νλ€.μμ μν¬νλ‘μ° ν΅ν©. μμΈ‘μ μ μ ν μκ°μ μ μ ν μμμμκ² μ λ¬λμ΄ μ μ ν λ°μμ μ λ°ν λλ§ μ μ©νλ€. 17μκ° μ 경보λ μ΄λ»κ² μ μλμ΄μΌ νλκ°? λꡬμκ²? μ΄λ€ κΆκ³ μ‘°μΉμ ν¨κ»? μΈμ μμΈκ³Ό ꡬν κ³Όνμ μΈ‘λ©΄μ κΈ°κ³ νμ΅λ§νΌμ΄λ λμ μ μ΄λ€.νμ μ§λ¨. μ
μ νμλ μ΄μ§μ μ΄λ€. μ£Όλ‘ μΈκ³Ό νμλ₯Ό λμμΌλ‘ νλ ¨λ λͺ¨λΈμ λ΄κ³Ό νμμμ μ’μ μ±λ₯μ 보μ΄μ§ λͺ»ν μ μμΌλ©°, κ·Έ λ°λλ λ§μ°¬κ°μ§μ΄λ€. μ§λ¨λͺ
, μ°λ Ή, κΈ°μ μ€μ¦λ, λλ° μ§ν νλ‘νμΌμ λ°λ₯Έ μ±λ₯ κ³μΈ΅νκ° νμνλ€.μ¨μ΄λ¬λΈ λ΄μ± λ° λ°μ΄ν° νμ§. μ
μ νμλ λͺ©μ, μμ κ²μ¬, λλ μ²μΉ μ€μ μ¨μ΄λ¬λΈ μΌμλ₯Ό μ κ±°ν μ μμ΄ λ°μ΄ν° κ³΅λ°±μ΄ λ°μνλ€. νμμ μμ§μμ λμ μ‘μμ μμ±νλ€. λͺ¨λΈμ μ€μ λ°°ν¬ νκ²½μμ κ²°μΈ‘ λ°μ΄ν°μ μ‘μμ΄ λ§μ λ°μ΄ν°μ λν΄ κ°κ±΄ν΄μΌ νλ€.λ³μ μνμ λν μμ¬μ
Scheid et al.μ μ°κ΅¬λ μ°μ μ¨μ΄λ¬λΈ νλ ₯ μ§ν λ°μ΄ν°μ μ μ©λ λ₯λ¬λμ΄ μμμ μΌλ‘ μ€ν κ°λ₯ν μ ν μκ°μ λκ³ νμ μ
νλ₯Ό μμΈ‘ν μ μμμ μ
μ¦νλ€. 17μκ°μ κΈ°μ‘΄ μ κ·Όλ²μ λΉν΄ λ―Έλ―Έν κ°μ μ΄ μλλΌ, λ°μμ μΉλ£μμ λ₯λμ μΉλ£λ‘μ μ§μ μ νμ μλ―Ένλ€.
κ·Έλ¬λ κ²μ¦λ μμΈ‘ λͺ¨λΈμμ μ€μ λ‘ λ°°ν¬λ μμ μμ€ν
μΌλ‘ λμκ°λ κ³Όμ μλ μ£Όλ‘ κΈ°μ μ μ΄μ§ μμ μ¬λ¬ κ³Όμ κ° μλ°λλ€. μ΄λ¬ν κ³Όμ λ μ‘°μ§μ (κΈ°μ‘΄ μν¬νλ‘μ μλ¦Ό ν΅ν©), μΈμ (μμμκ° AI μμ± μλ¦Όμ μ λ’°νκ³ μ΄μ λ°λΌ νλνλλ‘ λ³΄μ₯), κ·μ μ (μμ μμ¬κ²°μ μ§μμ λν FDA μΉμΈ), κ²½μ μ (μ‘°κΈ° κ°μ
μ΄ λͺ¨λν°λ§ μΈνλΌλ₯Ό μ λΉνν λ§νΌ μΆ©λΆν λΉμ©μ μ κ°νκ³ κ²°κ³Όλ₯Ό κ°μ νλ€λ κ²μ μ
μ¦) μΈ‘λ©΄μΌλ‘ λλλ€.
μ΄ μ°κ΅¬λ μ°μ μ¨μ΄λ¬λΈ λ°μ΄ν°μ μ νΈκ° μ‘΄μ¬νλ©° λ₯λ¬λμ΄ μ΄λ₯Ό μΆμΆν μ μλ€λ κ°λ ₯ν κ°λ
μ¦λͺ
(proof of concept)μΌλ‘ μ΄ν΄νλ κ²μ΄ κ°μ₯ μ μ νλ€. λ€μ λ¨κ³μΈ λ€κΈ°κ΄ κ²μ¦, ꡬν μν, κ²°κ³Ό μ°κ΅¬λ₯Ό ν΅ν΄ μ΄ μμΈ‘μ΄ μ€μ λ‘ νμ κ²°κ³Ό κ°μ μΌλ‘ μ΄μ΄μ§ μ μλμ§κ° κ²°μ λ κ²μ΄λ€.
κ΄λ ¨ λμ§νΈ ν¬μ€ λ° μμ AI μ°κ΅¬λ ORAA ResearchBrainμ ν΅ν΄ νμν μ μλ€.