An extensive study for binary characterisation of adrenal tumours.

Area of Science:

• Medical Imaging
• Oncology
• Computer-Aided Diagnosis

Background:

• Adrenal tumors present diagnostic challenges due to overlapping imaging features.
• Current diagnostic methods like biopsy carry risks of injury and complications.
• Accurate characterization of adrenal tumors is crucial for appropriate treatment planning.

Purpose of the Study:

• To develop a non-invasive method for binary characterization of adrenal tumors using dynamic computed tomography (CT) images.
• To exclude the need for additional imaging modalities and invasive biopsy procedures.
• To identify the most effective image features for adrenal tumor classification.

Main Methods:

• Utilized dynamic CT images from a dataset of 8 adrenal tumor subtypes.
• Investigated histogram, grey level co-occurrence matrix (GLCM), and wavelet-based features.
• Employed optimized neural networks and four classification algorithms for binary classification.
• Evaluated performance using accuracy, sensitivity, specificity, and Area Under the Curve (AUC).

Main Results:

• The proposed framework achieved success rates of 80.7% for accuracy, 75% for sensitivity, 82.22% for specificity, and 78.61% for AUC.
• Identified specific image features that are most effective for adrenal tumor identification.
• Demonstrated the feasibility of classifying adrenal tumors without invasive procedures.

Conclusions:

• The developed framework offers an efficient and accurate approach for the non-invasive binary characterization of adrenal tumors.
• This method has the potential to reduce reliance on invasive biopsy, thereby minimizing patient risk.
• Further research can build upon these findings to enhance diagnostic capabilities in adrenal tumor management.