Saved in:
| Hovedforfatter: | |
|---|---|
| Format: | Recurso digital |
| Sprog: | |
| Udgivet: |
Zenodo
2026
|
| Online adgang: | https://doi.org/10.5281/zenodo.19658529 |
| Tags: |
Tilføj Tag
Ingen Tags, Vær først til at tagge denne postø!
|
Indholdsfortegnelse:
- <p><em><span>The integration of digital agents into clinical environments represents a paradigm shift in medical diagnostics, directly addressing systemic challenges such as specialist shortages and cognitive fatigue. This paper presents a comprehensive diagnostic ecosystem designed for chest radiography (CXR) analysis, employing a dual-stage deep learning architecture. Utilizing a dataset exceeding 50,000 anonymized images, the system pipeline first validates morphological integrity via YOLOv8, followed by a multi-class pathology classification (Normal, COVID-19, Viral Pneumonia, Lung Opacity) using Convolutional Neural Networks. To bridge the trust deficit often associated with automated systems, spatial heat-mapping via Grad-CAM was implemented, providing visual anatomical rationales for algorithmic determinations. Empirical evaluations demonstrate an aggregate diagnostic accuracy of over 92%, compressing the analytical timeframes from an average of 12 minutes per patient to roughly 5 seconds. The resultant ecosystem, highly accessible via a unified Django web dashboard and cross-platform Telegram bot infrastructure, serves not merely as a clinical triage tool but as a continuous pedagogical instrument, refining the observational heuristics of diagnostic professionals in real-time.</span></em></p>