An explainable deep learning-based feature fusion model for acute lymphoblastic leukemia diagnosis and severity
Hajra Khan1, Muhammad Zaheer Sajid2, Nauman Ali Khan1
1Department of Computer Software Engineering, Military College of Signals, National University of Science and Technology, Islamabad, Pakistan.
Frontiers in Medicine
|January 1, 2026
Summary
A new deep learning model, XIncept-ALL, accurately detects and classifies acute lymphoblastic leukemia (ALL) severity. This AI tool aids in early diagnosis and treatment planning for this critical blood cancer.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Hematology
Background:
- Acute lymphoblastic leukemia (ALL) is a prevalent childhood blood cancer requiring precise diagnosis.
- Current diagnostic methods can be time-consuming and may benefit from automated analysis.
- Early detection and severity classification are crucial for effective ALL treatment and improved patient outcomes.
Purpose of the Study:
- To develop and evaluate a novel deep learning framework, XIncept-ALL, for automated detection and classification of ALL severity.
- To enhance the model's robustness and generalization through feature fusion and data augmentation.
- To provide interpretable insights into the model's predictions using Grad-CAM visualizations.
Main Methods:
- A deep learning framework (XIncept-ALL) was created by integrating InceptionV3 and Xception networks.
- Data auto-augmentation techniques were used to address class imbalance and prevent overfitting.
- Grad-CAM was employed for visualizing discriminative regions in ALL cell images.
- An XGBoost classifier was used for classifying extracted features into four severity levels: Benign, Early, Pre, and Pro.
Main Results:
- The XIncept-ALL model achieved a high average accuracy of 99.5% on an external dataset.
- The model demonstrated superior performance in classifying ALL severity levels.
- Grad-CAM visualizations confirmed that the model focuses on clinically relevant features, enhancing interpretability.
Conclusions:
- The XIncept-ALL framework offers an efficient, scalable, and practical approach for AI-driven medical image analysis in ALL.
- The model shows significant potential for supporting clinical decision-making and improving patient outcomes through accurate and interpretable diagnostics.
- This advancement contributes to the development of reliable computer-aided diagnostic systems for acute lymphoblastic leukemia.
