Fully automated analysis for bone scintigraphy with artificial neural network: usefulness of bone scan index (BSI) in breast cancer

Anri Inaki; Kenichi Nakajima; Hiroshi Wakabayashi; Takafumi Mochizuki; Seigo Kinuya

doi:10.1007/s12149-019-01386-1

Fully automated analysis for bone scintigraphy with artificial neural network: usefulness of bone scan index (BSI) in breast cancer

Ann Nucl Med. 2019 Oct;33(10):755-765. doi: 10.1007/s12149-019-01386-1. Epub 2019 Jul 17.

Authors

Anri Inaki¹, Kenichi Nakajima², Hiroshi Wakabayashi³, Takafumi Mochizuki⁴, Seigo Kinuya⁵

Affiliations

¹ Department of Nuclear Medicine, Kanazawa University Hospital, 13-1 Takara-machi, Kanazawa, 920-8641, Japan. henri@staff.kanazawa-u.ac.jp.
² Department of Functional Imaging and Artificial Intelligence, Kanazawa University, 13-1 Takara-machi, Kanazawa, 920-8641, Japan.
³ Department of Nuclear Medicine, Kanazawa University Hospital, 13-1 Takara-machi, Kanazawa, 920-8641, Japan.
⁴ Kanazawa Advanced Medical Center, 13-1 Takara-machi, Kanazawa, 920-8641, Japan.
⁵ Department of Nuclear Medicine, Faculty of Medicine, Institute of Medical, Pharmaceutical and Health Sciences, Kanazawa University, 13-1 Takara-machi, Kanazawa, 920-8641, Japan.

PMID: 31317398
DOI: 10.1007/s12149-019-01386-1

Abstract

Objective: Artificial neural network (ANN) technology has been developed for clinical use to analyze bone scintigraphy with metastatic bone tumors. It has been reported to improve diagnostic accuracy and reproducibility especially in cases of prostate cancer. The aim of this study was to evaluate the diagnostic usefulness of quantitative bone scintigraphy with ANN in patients having breast cancer.

Patients and methods: We retrospectively evaluated 88 patients having breast cancer who underwent both bone scintigraphy and ¹⁸F-fluorodeoxyglucose (FDG) positron-emission computed tomography/X-ray computed tomography (PET/CT) within an interval of 8 weeks between both examinations for comparison. The whole-body bone images were analyzed with fully automated software that was customized according to a Japanese multicenter database. The region of interest for FDG-PET was set to bone lesions in patients with bone metastasis, while the bone marrow of the ilium and the vertebra was used in patients without bone metastasis.

Results: Thirty of 88 patients had bone metastasis. Extent of disease, bone scan index (BSI) which indicate severity of bone metastasis, the maximum standardized uptake value (SUVmax), metabolic tumor volume (MTV), total lesion glycolysis (TLG), and serum tumor markers in patients with bone metastasis were significantly higher than those in patients without metastasis. The Kaplan-Meier survival curve showed that the overall survival of the lower BSI group was longer than that with the higher BSI group in patients with visceral metastasis. In the multivariate Cox proportional hazard model, BSI (hazard ratio (HR): 19.15, p = 0.0077) and SUVmax (HR: 10.12, p = 0.0068) were prognostic factors in patients without visceral metastasis, while the BSI was only a prognostic factor in patients with visceral metastasis (HR: 7.88, p = 0.0084), when dividing the sample into two groups with each mean value in patients with bone metastasis.

Conclusion: BSI, an easily and automatically calculated parameter, was a well prognostic factor in patients with visceral metastasis as well as without visceral metastasis from breast cancer.

Keywords: Artificial neural network; Bone scan; Breast cancer with bone metastasis; FDG-PET.

MeSH terms

Automation
Bone Neoplasms / diagnostic imaging
Bone Neoplasms / secondary
Bone and Bones / diagnostic imaging*
Breast Neoplasms / diagnostic imaging*
Breast Neoplasms / pathology
Female
Fluorodeoxyglucose F18
Humans
Image Processing, Computer-Assisted / methods*
Middle Aged
Neural Networks, Computer*
Positron Emission Tomography Computed Tomography*
Retrospective Studies
Whole Body Imaging

Substances

Fluorodeoxyglucose F18