Scholarlink

Acute Lymphocytic Leukemia (ALL) must be diagnosed promptly and accurately to be treated. This paper introduces a Lightning CNN-based deep learning model for blood cell classification into basophil, erythroblast, monocyte, myeloblast, and segmented neutrophil. Here, a 750 blood smear image-based model attained an accuracy rate of 98.87% and macro F1-score of 97.88%, with high reliability. To enhance interpretability, we employ Explainable AI (XAI) techniques—LIME identifies salient image regions affecting predictions, and SHAP offers feature importance scores. These techniques make transparent and trustworthy AI-based leukemia detection possible to aid clinicians in medical decision-making. Our findings show that deep learning combined with XAI enhances diagnostic accuracy and model interpretability, making AI-based blood cell classification a powerful medical tool.

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

Blood Cell Cancer Prediction Using Lightning CNN with XAI Interpretability (LIME and SHAP)

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • T. P. Pattanaik,
  • Soumya Ranjan Nayak,
  • Alok Kumar Jagadev,
  • G. Palai,
  • K. P. Swain

摘要

Acute Lymphocytic Leukemia (ALL) must be diagnosed promptly and accurately to be treated. This paper introduces a Lightning CNN-based deep learning model for blood cell classification into basophil, erythroblast, monocyte, myeloblast, and segmented neutrophil. Here, a 750 blood smear image-based model attained an accuracy rate of 98.87% and macro F1-score of 97.88%, with high reliability. To enhance interpretability, we employ Explainable AI (XAI) techniques—LIME identifies salient image regions affecting predictions, and SHAP offers feature importance scores. These techniques make transparent and trustworthy AI-based leukemia detection possible to aid clinicians in medical decision-making. Our findings show that deep learning combined with XAI enhances diagnostic accuracy and model interpretability, making AI-based blood cell classification a powerful medical tool.