Related Experiment Video
Updated: May 12, 2026
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
Evaluating the Impact of Sex-Specific Artificial Intelligence on Bone Metastasis Detection: VSBONE Bone Scan Index
Aliya Toilybayeva1, Reiko Ideguchi2, Takashi Kudo2
1Department of Radioisotope Medicine, Atomic Bomb Disease Institute, Nagasaki University Graduate School of Biomedical Sciences, Nagasaki, JPN.
Objective(S):
The present study compared the diagnostic accuracy of two different versions of an AI-assisted computer-aided diagnosis (CAD) software (VSBONE versions 2.1 and 3.0, Nihon Medi-Physics, Tokyo, Japan) for detecting bone metastases using bone scintigraphy images, with a focus on evaluating the impact of the new sex-specific analysis feature in version 3.0 on the overall performance and diagnostic accuracy in clinical decision-making.
Methods:
This retrospective study analyzed 1,421 bone scintigraphy images from 1,119 patients at Nagasaki University Hospital. Patients were initially classified into metastatic and non-metastatic groups based on expert visual interpretation of planar images. Final classifications were made after single-photon emission computed tomography (SPECT)/CT and clinical review, resulting in metastatic, non-metastatic, and initially missed metastases groups. Bone Scan Index (BSI) and Hotspot number (HSn) were quantified using VSBONE BSI versions 2.1 and 3.0.
Results:
Overall diagnostic performance improved with version 3.0 compared to version 2.1. In male patients, specificity increased (from 83.9% to 86.4%), with a slight decrease in sensitivity (from 82.0% to 79.7%), in the Youden index remained similar (from 0.660 to 0.662). In female patients, diagnostic accuracy improved, with notable increases in sensitivity (from 77.3% to 79.1%) and specificity (from 73.9% to 86.7%), and a marked increase in the Youden index (from 0.512 to 0.658).
Conclusion:
The present study showed that version 3.0 improved diagnostic quality and accuracy, particularly in the sex-specific analysis (female patients). These findings support the integration of sex-specific AI in CAD software to enhance diagnostic accuracy, reduce interpretation errors, and promote more equitable patient care.
