Related Experiment Video
Updated: Apr 29, 2026
Automated Joint Space Detection Improves Bone Segmentation Accuracy
Published on: November 28, 2025
[Usefulness of spatially adaptive noise reduction processing in computer-assisted diagnosis system for bone
Hajime Ichikawa1, Masahisa Onoguchi, Koichi Okuda
1Department of Radiology, Toyohashi Municipal Hospital.
Objectives:
The goal of this study was to assess the diagnostic accuracy of Pixon-processed images in comparison with raw images for computer-assisted interpretation of bone scintigraphy (BONENAVI).
Methods:
Whole-body scans of 57 patients with prostate cancer who had undergone bone scintigraphy for suspected bone metastases were obtained approximately 3 h after intravenous injection of 740 MBq (99m)Tc-methylene diphosphonate. We obtained two image sets: raw images and images processed using the Pixon method. Artificial neural network (ANN) values, bone scan index (BSI), number of hotspots and regional ANN value of two images set were automatically calculated by the BONENAVI software. Areas under the receiver operator characteristic curves (AUC) were calculated in patient-based and lesion-based analyses.
Results:
In ten cases with bone metastases, ANN, BSI and number of hotspots for processed images were equivalent to those in the raw images. However, in 47 cases without bone metastases, ANN, BSI and number of hotspots for processed images showed significantly lower values than those for the raw images (p<0.05). Sensitivity, specificity and accuracy of the raw images were 90.2, 44.7 and 65.9%, and those of the processed images were 90.2, 57.4 and 72.7%, respectively. The AUC for processed images was equivalent to that for raw images.
Conclusions:
Specificity and accuracy in the detection of bone metastases showed the Pixon-processed images to have high diagnostic performance. We conclude that the precision of computer-assisted interpretation of bone scintigraphy can be enhanced by using Pixon processing.
