A lightweight attention-driven YOLOv5m model for improved brain tumor detection DOI Creative Commons
Shakhnoza Muksimova, Sabina Umirzakova,

Sevara Mardieva

et al.

Computers in Biology and Medicine, Journal Year: 2025, Volume and Issue: 188, P. 109893 - 109893

Published: Feb. 22, 2025

Brain tumors are regarded as one of the most lethal, devastating, and aggressive diseases, significantly reducing life expectancy affected individuals. For this reason, in pursuit advancing brain tumor diagnostics, study introduces a significant enhancement to YOLOv5m model by integrating an Enhanced Spatial Attention (ESA) layer, tailored specifically for analysis magnetic resonance imaging (MRI) scans. Traditional detection methods, heavily reliant on expert interpretation MRI, fraught with challenges such high variability risk human error. Our innovative approach leverages ESA layer acutely focus salient features, improving method ability differentiate between common classes tumors-meningioma, pituitary, glioma tumors. By processing spatial features enhanced precision, minimizes false positives maximizes reliability. Validated against comprehensive dataset 3064 T1-weighted contrast-enhanced MRI images from 233 patients, our modified architecture demonstrates superior performance metrics compared standard model, highlighting its potential robust tool clinical applications automated precise diagnosis.

Language: Английский

A lightweight attention-driven YOLOv5m model for improved brain tumor detection DOI Creative Commons
Shakhnoza Muksimova, Sabina Umirzakova,

Sevara Mardieva

et al.

Computers in Biology and Medicine, Journal Year: 2025, Volume and Issue: 188, P. 109893 - 109893

Published: Feb. 22, 2025

Brain tumors are regarded as one of the most lethal, devastating, and aggressive diseases, significantly reducing life expectancy affected individuals. For this reason, in pursuit advancing brain tumor diagnostics, study introduces a significant enhancement to YOLOv5m model by integrating an Enhanced Spatial Attention (ESA) layer, tailored specifically for analysis magnetic resonance imaging (MRI) scans. Traditional detection methods, heavily reliant on expert interpretation MRI, fraught with challenges such high variability risk human error. Our innovative approach leverages ESA layer acutely focus salient features, improving method ability differentiate between common classes tumors-meningioma, pituitary, glioma tumors. By processing spatial features enhanced precision, minimizes false positives maximizes reliability. Validated against comprehensive dataset 3064 T1-weighted contrast-enhanced MRI images from 233 patients, our modified architecture demonstrates superior performance metrics compared standard model, highlighting its potential robust tool clinical applications automated precise diagnosis.

Language: Английский

Citations

2