In bone scintigraphy, bone scan index (BSI) has recently been shown to be useful for quantitative assessment of bone metastasis in patients with prostate cancer. We report our experience with the use of a new BSI analysis software for technetium-99m hydroxymethylene diphosphonate
99mTc-HMDP) bone scanning.
Deep learning is accomplished using a database of identified bone anatomy structures of 246 patients, and accordingly, the skeleton is classified into 12 segments: skull, sternum, cervical spine, thoracic spine, lumbar spine, sacrum, clavicle, scapula, humerus, ribs, pelvis, and femur.
It also identifies abnormal accumulation through deep learning using a database of identified sites of abnormal accumulation in 896 patients with prostate cancer as supervised data. Each bone area in the 12 segments is calculated for each patient, and the local BSI is calculated from the area of each bone metastasis in each segment. The sum of local BSI is defined as BSI, and the number of bone metastases is defined as the number of hot spots.
The analysis was performed on a 64-bit Windows 10 personal computer with Intel® CoreTMi7-8550U and 16.0 GB RAM. The time required for analyzing bone scintigraphy for a patient was approximately 20 seconds.
Some cases actually used are presented in the text.
View full abstract