Automated Adrenal Gland Disease Classes Using Patch-Based Center Symmetric Local Binary Pattern Technique with CT...

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Medicine
• Computational Pathology

Background:

• Incidental adrenal masses are frequently detected on abdominal CT scans, necessitating accurate diagnosis for effective treatment and prognosis.
• Distinguishing between various adrenal pathologies, such as pheochromocytoma, adenoma, and metastasis, is crucial but challenging.
• Automated classification methods can aid radiologists in the interpretation of adrenal CT images.

Purpose of the Study:

• To develop and evaluate a novel, handcrafted machine learning approach for automated classification of adrenal gland CT images.
• To assess the performance of different classifiers in distinguishing normal adrenal glands from pathological conditions.
• To establish a highly accurate and efficient method for screening adrenal gland diseases using CT data.

Main Methods:

• A dataset of 759 adrenal gland CT slices from 96 subjects was curated and labeled into four classes: normal, pheochromocytoma, lipid-poor adenoma, and metastasis.
• Image features were extracted using the center symmetric local binary pattern (CS-LBP) method on fixed-size patches.
• Neighborhood Component Analysis (NCA) was employed for feature selection, followed by classification using k-nearest neighbor (kNN), support vector machine (SVM), and neural network (NN) models.

Main Results:

• The proposed method achieved high classification accuracies: 99.87% with kNN, 99.21% with SVM, and 98.81% with NN.
• The kNN classifier demonstrated superior performance, with no instances of pathological images being misclassified as normal.
• The CS-LBP-based approach exhibited low time complexity, making it suitable for rapid analysis.

Conclusions:

• The developed automated classification system for adrenal gland pathologies on CT images is highly accurate and efficient.
• The CS-LBP method combined with kNN shows significant potential for clinical application in screening adrenal gland diseases.
• This AI-driven approach can assist in the diagnostic workflow for incidental adrenal findings on CT scans.