Texture Feature-Based Classification on Transrectal Ultrasound Image for Prostatic Cancer Detection

Area of Science:

• Medical Imaging
• Oncology
• Computer-Aided Diagnosis

Background:

• Prostate cancer is a prevalent malignancy in men, where early detection significantly improves treatment outcomes.
• Ultrasound imaging offers a viable modality for early prostate cancer detection, but subjective interpretation of images limits diagnostic accuracy.
• Objective characterization of prostate tissue from ultrasound images is needed to enhance the detection of malignant tumors.

Purpose of the Study:

• To develop and evaluate a novel transrectal ultrasound image analysis method for characterizing prostate tissue.
• To assess the potential of the proposed method in distinguishing malignant from benign prostate lesions.
• To improve the accuracy of prostate cancer detection through automated image processing and classification.

Main Methods:

• Image preprocessing involved optical density conversion of transrectal ultrasound images.
• Texture features were extracted using local binarization and Gaussian Markov random fields, followed by linear combination.
• Support Vector Machine (SVM) classifier was employed for the classification of fused texture features.

Main Results:

• The proposed method was applied to a dataset of 342 transrectal ultrasound images.
• The system achieved an overall accuracy of 70.93%, with a sensitivity of 70.00% and specificity of 71.74%.
• Experimental results demonstrated a degree of success in differentiating cancerous from non-cancerous prostate tissues.

Conclusions:

• The developed transrectal ultrasound image analysis method shows promise for aiding in the detection of prostate cancer.
• Automated feature extraction and classification can partially overcome the limitations of subjective interpretation in ultrasound diagnostics.
• Further research and refinement of the method could lead to improved diagnostic performance for prostate cancer.