Scholarlink

Pneumonia diagnosis is significant given its high mortality rate; hence, there is a need for precision medicine to ensure effective treatment. Diagnosing pneumonia can be challenging due to its varied patterns, heterogeneous multilobular inflammatory infection in the lung, complex overlapping patches, and severity influenced by the patient’s medical history. These challenges are further exacerbated by the computerized AI-driven diagnostic technique are reliance on black-and-white X-ray images, inconsistencies in X-ray intensity, obtaining a comprehensive pneumonia dataset that captures diverse opacity patterns, early diagnostic markers, variations in body shape, age, geographic factors, and patient positioning remains a significant hurdle. In this study, simplified lightweight deep architecture is suggested for Chest X-Rays classification as healthy or bacterial or viral pneumonia-affected using spatial characteristics learnt on publically accessible large-scale datasets. The classification performance of the proposed model is compared and analyzed with pre-trained CNN models like EfficientNetB7, MobileNetV2, NasNetMobile, and XceptionNet. The suggested streamlined architecture due to its simple architecture and layer-by-layer detailed feature development and a great vision architecture successfully identified pneumonia from thoracic Images with 93.15% accuracy.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Deep Architecture for Multiclass Pneumonia Diagnosis from Chest X-Ray Images

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Kanchan Dabre,
  • Satishkumar L. Varma,
  • Pankaj B. Patil

摘要

Pneumonia diagnosis is significant given its high mortality rate; hence, there is a need for precision medicine to ensure effective treatment. Diagnosing pneumonia can be challenging due to its varied patterns, heterogeneous multilobular inflammatory infection in the lung, complex overlapping patches, and severity influenced by the patient’s medical history. These challenges are further exacerbated by the computerized AI-driven diagnostic technique are reliance on black-and-white X-ray images, inconsistencies in X-ray intensity, obtaining a comprehensive pneumonia dataset that captures diverse opacity patterns, early diagnostic markers, variations in body shape, age, geographic factors, and patient positioning remains a significant hurdle. In this study, simplified lightweight deep architecture is suggested for Chest X-Rays classification as healthy or bacterial or viral pneumonia-affected using spatial characteristics learnt on publically accessible large-scale datasets. The classification performance of the proposed model is compared and analyzed with pre-trained CNN models like EfficientNetB7, MobileNetV2, NasNetMobile, and XceptionNet. The suggested streamlined architecture due to its simple architecture and layer-by-layer detailed feature development and a great vision architecture successfully identified pneumonia from thoracic Images with 93.15% accuracy.