A Real-Time Standalone Artificial Intelligence System Outside the Picture Archiving and Communication System for Detecting Brain Metastases Using Magnetic Resonance Imaging: A Feasibility Study
Year: 2026
Type: article
Abstract: Background/Objectives: Small brain metastases on magnetic resonance (MR) imaging can be overlooked, necessitating careful and time-consuming inspection. We evaluated the feasibility of a real-time standalone artificial intelligence (AI) system for detecting lesions operating outside the picture archiving and communication system with a low implementation barrier. Methods: In this retrospective study, 212 cases were divided into training (n = 149), validation (n = 32), and testing (n = 31) datasets. Training and validation datasets were used for deep learning with You Only Look Once 11x. AI detection was performed on a laptop for test images imported from non-embedded devices, including a video capture card connected via a monitor cable, a webcam, and a smartphone using a wireless streaming system. Detection rates for these image acquisition methods were compared with those of the original images. Results: All reported detection rates are lesion-level evaluations. The AI detection rates for the original images and those acquired via the video capture card, webcam and smartphone were 91.2%, 91.0%, 89.5%, and 79.6%, respectively. The AI detection rate for the video capture card and webcam did not differ significantly from that for the original images (mean difference, −0.19% and −1.84%, respectively; both p-values > 0.05), whereas the detection rate for the smartphone was significantly lower than that for the original images (mean difference, −11.58%; p-value < 0.0001). Conclusions: Real-time standalone AI detection systems using devices such as video capture cards or webcams offer low implementation barriers. These findings demonstrate technical feasibility and potential for real-time diagnostic assistance.
Source: Clinics and Practice
Language: English
Subfield: Health Informatics
Field: Medicine
Domain: Health Sciences
Open Access status: gold
Estimated APC paid: $1,930
OpenAlex ID: W7225798956