Related Experiment Video
Updated: May 3, 2026

Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
Rapid detection of bone metastasis at thoracoabdominal CT: accuracy and efficiency of a new visualization algorithm
Daniel F Toth1, Michael Töpker, Marius E Mayerhöfer
1From the Departments of Radiology (D.F.T., M.T., M.E.M., J.F., U.A., M.W., C.C., C.J.H., H.R.) and Nuclear Medicine (G.K.), Medical University of Vienna, Waehringer Guertel 18-20, A-1090 Vienna, Austria; and Department of Radiology, Duke University School of Medicine, Durham, NC (G.D.R.).
Purpose:
To retrospectively assess the use of a combination of cancellous bone reconstructions (CBR) and multiplanar reconstructions (MPRs) for the detection of bone metastases at thoracoabdominal computed tomography (CT) compared with the use of MPRs alone.
Materials And Methods:
The study was approved by the local institutional review board. Included were 156 consecutive patients with confirmed cancer who underwent a whole-body positron emission tomography (PET)/CT examination for clinical purposes (93 male and 63 female patients; mean age ± standard deviation, 59.8 years ± 14.9; range, 11-85 years). Only the CT images were processed with the CBR algorithm, which segments the bones and removes the cortical layer from the images. The PET images served as part of the reference standard. Images from 15 patients were used as a training set. Four radiologists independently evaluated images of half of the remaining 141 patients by using CBRs and MPRs together, and the other half by using MPRs only. Radiologists were blinded to patient names, and patient order was randomized. Results for detection rates and reporting time were recorded and compared with a standard of reference for each patient that was created by one senior radiologist and one nuclear medicine specialist by using all available CT and PET data, CBRs, and follow-up examinations. General estimation equations were used for statistical analysis.
Results:
There were 349 lesions found in 103 patients, with 203 classified as malignant. Each patient was assessed by two readers per method, leading to a total of 698 lesions. The detection rate for all bone lesions was 35% (247 of 698) for MPRs and 74% (520 of 698) when CBRs and MPRs were used together, which was significantly higher (P < .001). The average reading time decreased from 85 to 43 seconds (P < .001) when both reconstructions were used.
Conclusion:
Advanced visualization of cancellous bone significantly increased the detection of bone metastases and reduced the time for interpretation.
More Related Videos
12:23Multi-modal Imaging of Angiogenesis in a Nude Rat Model of Breast Cancer Bone Metastasis Using Magnetic Resonance Imaging, Volumetric Computed Tomography and Ultrasound
Published on: August 14, 2012
08:04Monitoring Tumor Metastases and Osteolytic Lesions with Bioluminescence and Micro CT Imaging
Published on: April 14, 2011