Usefulness of novel computer-assisted diagnosis algorithm with bone scintigraphy in CRPC patients with bone metastases.

1248 Purpose: Image-based quantitative analysis in bone scintigraphy may be useful in predicting disease progression in patients with castration-resistant prostate cancer (CRPC). The aim of this study was to evaluate clinical usefulness of novel computer-assisted diagnosis algorithm by using deep learning with bone scintigraphy. Materials and Methods: CRPC patients with bone metastasis who received Ra-223 treatment from January 2017 to April 2019 were included. All patients underwent bone scintigraphy, diffusion-weighted whole-body imaging with background body signal suppression (DWIBS) - MRI and F-18 FDG-PET/CT in the treatment. The number of abnormal hotspots was automatically calculated by using computer-assisted diagnosis algorithm, "bone scan index automated calculation software by Tc-99m-HMDP (BSIACS)". Two nuclear medicine specialist finally diagnosed bone metastasis with reference to these imaging examinations. The correlations between these parameters derived from bone scintigraphy and final diagnoses of specialists were analyzed. Results: A total of 1355 lesions were included in this study. The sensitivity of BSIACS in diagnosing bone metastasis was 88.7%, the specificity was 98.8 %, positive predictive value was 95.6 %, negative predictive value was 97.1 %, and accuracy was 96.4%. Lesions showing false negatives included: 30 small lesions and 4 osteolytic lesions. A statistically significant correlation was observed between manually counted lesion numbers and BSIACS with a Spearman9s correlation coefficient (r) of 0.944 and a p-value of 0.0001. Conclusions: The novel computer-assisted diagnosis algorithm using BSIACS may be effective for evaluating disease condition in patients with CRPC.