An Iterative Algorithm for Semisupervised Classification of Hotspots on Bone Scintigraphies of Patients with Prostate...

Area of Science:

• Medical Imaging
• Oncology
• Artificial Intelligence

Background:

• Prostate cancer (PCa) frequently metastasizes to bone, necessitating accurate imaging for diagnosis and monitoring.
• Bone scintigraphy is a sensitive and widely used imaging technique for detecting bone metastases.
• Current assessment of bone scans is subjective and time-consuming, lacking standardized quantification methods.

Purpose of the Study:

• To develop and evaluate a new semi-supervised algorithm for reducing false positives in automatically detected hotspots in bone scintigraphy images.
• To create a tool that assists physicians in quantifying bone scans, evaluating disease progression, and monitoring treatment response in PCa patients.

Main Methods:

• A novel semi-supervised algorithm was developed for false positive reduction in bone scintigraphy.
• The algorithm operates iteratively and addresses the challenge of limited annotated data.
• The method was tested on a dataset of bone scans manually labeled against patient medical records.

Main Results:

• The algorithm achieved a classification sensitivity of 63%, specificity of 58%, and a false negative rate of 37%.
• The proposed method demonstrated superiority when compared to other state-of-the-art classification algorithms.

Conclusions:

• The developed semi-supervised algorithm shows promise for improving the accuracy and efficiency of bone scan analysis in prostate cancer.
• This tool has the potential to aid physicians in making faster, more precise assessments of bone metastases and treatment efficacy.