An anatomically enhanced and clinically validated framework for lung abnormality classification using deep features and KL divergence

Abstract

Detecting lung abnormalities via chest X-rays is challenging due to understated tissue variations often ignored by traditional methods. Augmentation techniques like rotation or flipping risk distorting critical anatomical features, actually leading to misdiagnosis. This paper proposes a novel two-stage ASCE (Anatomical Segmentation and Color-Based Enhancement) framework for precise and efficient classification of lung abnormalities while preserving anatomical integrity. Stage 1 classifies Normal vs. Pneumonia with 95 % accuracy, an AUC of 0.98, and an F1-score of 0.92. Stage 2 distinguishes Pneumonia into Viral and Bacterial subtypes with 100 % accuracy and F1-score. This approach integrates segmentation and tissue-specific color enhancements with Kullback-Leibler (KL) divergence, quantifying deviations from healthy lung regions for improved classification. The lightweight pipeline ensures computational efficiency (∼0.06s/image) and clinical interpretability by preserving diagnostic features, enhancing visibility, and enabling quantitative analysis.1.Preserving Anatomical Structures: The methodology ensures that diagnostic features are preserved and highlighted with Anatomy-Preserved Segmentation2.Enhancing Diagnostic Visibility: The system employs targeted colour-based enhancement that improves the visibility of potential abnormalities3.Quantitative Analysis with Kullback-Leibler (KL) divergence: The model enhances precise identification of abnormal tissue by comparing the probability distributions of healthy lungs and abnormal areas.

Keywords: Anatomical Segmentation and Color-Based Enhancement; Anatomical segmentation; Chest X-rays; Color based enhancement; Deep learning; KL divergence; Lung Abnormality; Pneumonia Detection.

Graphical Abstract: High-level overview of the ASCE preprocessing pipeline. This figure illustrates the sequential stages involved in the Anatomical Segmentation and Color-Based Enhancement process, from lung segmentation to the application of KL Divergence for feature enhancement. Each stage is marked to highlight its contribution to enhancing the detection accuracy of lung abnormalities.

Fig. 1

Comprehensive Workflow for Chest X-Ray Analysis from Preprocessing to Classification. This diagram illustrates the sequence of steps applied to chest X-ray datasets, including lung and tissue segmentation, colour mapping, and texture preservation, followed by feature extraction using the ResNet50 architecture, concluding in the classification of images into categories such as Pneumonia, and Normal and then Viru vs Bacteria.

Fig. 2

Visual plan of the ASCE preprocessing pipeline. The left column shows the original chest X-ray images, the center column displays the lung masks generated through anatomical segmentation, and the right column depicts the final enhanced output after applying tissue-specific color mapping and structural sharpening. This enhancement improves the visual separability of lung tissues such as alveoli and blood vessels, facilitating better feature extraction.

Fig. 3

Workflow of the proposed ASCE (Anatomical Segmentation and Colour-Based Enhancement) framework. The pipeline consists of five major stages: (1) Data Preparation, including validation of image-mask-annotation pairs; (2) Anatomical Segmentation and tissue-specific Colour-Based Enhancement to highlight alveoli, bronchi, and blood vessels; (3) Feature Extraction using ResNet50 followed by KL divergence computation against a healthy reference vector; (4) Classification using a soft-voting ensemble (Random Forest + XGBoost) that incorporates both deep features and KL divergence; and (5) Evaluation using standard performance metrics.

Fig. 4

a. Confusion Matrix for Stage 1: Normal vs. Pneumonia, b. Confusion Matrix for Stage 2: Virus vs. Bacteria.

Fig. 5

a. ROC Curve for Stage 1: Binary Classification, b. Precision-Recall Curve for Stage 1: Binary Classification, c. ROC Curve for Stage 2: Multi-Class Classification, d.Precision-Recall Curve for Stage 2: Multi-Class Classification.