K Doi¹*, Y Anetai², H Takegawa², Y Koike², S Nakamura², T Nishio¹, (1) Department of Medical Physics and Engineering, Osaka University Graduate School of Medicine, Suita city, Osaka, JP, (2) Department of Radiation Oncology, Kansai Medical University Hospital, Hirakata city, Osaka, Japan

• 
  K Doi

• 
  
• 
  
• 
  
• 
  
• 
  
• 
  
• 
  

TU-D930-IePD-F8-4 (Tuesday, 7/12/2022) 9:30 AM - 10:00 AM [Eastern Time (GMT-4)]

Exhibit Hall | Forum 8

Purpose: A bone metastasis lesion (BML) is detected by the CT examination, whereas this is difficult because of the blurry outlines or contrast of regions. Thus, this depends on the empirical manner of manual contrast adjusting to extract, leading doctors to overlook the BML. If bone metastasis (BM) progresses, it can cause skeletal-related events (SREs), which is possible to reduce the patient’s quality of life and overall survival [1]. From the above, despite BM being one of the crucial cancers, its lesion can be detected delay. Therefore, we have developed the method to automatically extract the BML from CT images in the early stage.

Methods: In this study, diffusion equation (DEQ) was numerically applied to image processing using the Crank-Nicolson method. DEQ has been applied to the process of the image smoothing [2]. DEQ works on a similar level of the gradient of the images to smooth, resulting in the emphasis of high-low regions of the calculated image. By using this, the method to extract the high-CT value regions from the CT image was developed. We investigated optimized calculation conditions (table 1), which also extracted obvious BM false-positive regions. Thus, the method to reduce noises was also developed by canceling calculation direction dependence (CDD). CDD was canceled combined by Hadamard product (HP), and the ratio of noise reduction was evaluated.

Results: High-CT value regions suspected of osteosclerosis and osteoblastic BML was extracted via DEQ without HP (figure 1). In addition, the noise caused by the former method was decrease 26% through additional HP processing (figure 2). The details of CDD are showed in figure 3.

Conclusion: We developed the mathematical method to extract regions suspected of osteosclerosis from the CT image. However, an improvement in the accuracy of diagnosis is still required.

Funding Support, Disclosures, and Conflict of Interest: This work was partly supported by QLEAR fellowship(Grant Number: Q21-024).

Keywords

CAD, Numerical Analysis, Image Analysis

Taxonomy

IM/TH- Image Analysis (Single Modality or Multi-Modality): Computer-aided decision support systems (detection, diagnosis, risk prediction, staging, treatment response assessment/monitoring, prognosis prediction)

Contact Email