Predicting malignancy of breast lesions seen in CT using Textural Analysis and Machine Learning

Congress:

ECR 2020

Poster Number:

C-14762

Type:

Scientific Exhibit

Keywords:

Artificial Intelligence and Machine Learning, Artificial Intelligence, Breast, CT, Computer Applications-Detection, diagnosis, Cancer, Retrospective, Not applicable, Performed at one institution

Authors:

M. A. Alkubeyyer¹, F. Almohideb ¹, R. N. ALJURAYYAN², S. Alsultan¹; ¹Riyadh/SA, ²Riyadh 11411/SA

DOI:

10.26044/ecr2020/C-14762

DOI-Link:

https://dx.doi.org/10.26044/ecr2020/C-14762

Fig. 1: Benign breast lesions dataset distribution of variable pathologies

Fig. 4: sensitivity , specificity , F-Measure and AUROC of 87%, 84%, 86% and 87% for...

Table 1: Extracted features per breast lesion region of interest (ROI).

Purpose

Breast cancer is the most commonly diagnosed cancer among women and the second leading cause of cancer-related mortality(1)(2). With the widespread use of CT in the current years for various indications, It is important for general radiologists to pay attention to the breasts encountered in CT to characterize breast lesions as benign or malignant. Texture analysis might help to discriminate between malignant and benign breast lesion encountered in CT chest using first and second-order statistics (Haralick features) and machine learning.

Methods and materials

We retrospectively extracted enhanced CT images of 77 lesions (39 malignant versus 38 benign) which are pathologically proven at our institute (King Saud University Medical City) aquired between 2014 to 2018. All CT images were obtained using (Discovery CT750 HD GE Healthcare, Waukesha, WI) with scanning parameters of 120 kV, 150 mA slice thickness of 3.75 mm( matrix 512 x 512). Images were reconstructed with filtered back projection technique. All CT images are enhanced post IV contrast obtained at portovenous phase. All malignant lesions pathologically...

Results

The proposed classifier showed good performance in discriminating malignant from benign lesions with a sensitivity, specificity, F-Measure, and AUROC of 87%, 84%, 86% and 87% for malignant breast lesion versus 84%, 87 %, 85% and 87% for benign lesions. Figure 4 shows plot of AUROC.

Conclusion

We presented a machine learning algorithm based on Jrip (RIPPER) classifier with acceptable accuracy for discriminating malignant breast lesions from benign ones encountered in enhanced CT images based on texture analysis. Nevertheless, there are some limitations to this study: This is a single institute study with a low sample study which might not be generalized in practice. Although invasive ductal carcinoma is the most encountered malignant breast tumor, other malignant breast lesions such as invasive lobar carcinoma(ILC), Ductal carcinoma in situ (DCIS), Medullary carcinoma (MC)...

Personal information and conflict of interest

M. A. Alkubeyyer; Riyadh/SA - nothing to disclose F. Almohideb; Riyadh/SA - nothing to disclose R. N. Aljurayyan; Riyadh/SA - nothing to disclose S. Alsultan; Riyadh/SA - nothing to disclose

References

World Health Organization. Global Health Observatory. Geneva: World Health Organization; 2018. who.int/gho/database/en/. Accessed June 21, 2018. Ghoncheh M, Pournamdar Z, Salehiniya H. Incidence and Mortality and Epidemiology of Breast Cancer in the World. Asian Pac J Cancer Prev 2016;17(S3):43–46. Rueden, C. T.; Schindelin, J. & Hiner, M. C. et al. (2017), "ImageJ2: ImageJ for the next generation of scientific image data", BMC Bioinformatics 18:529, PMID 29187165, doi:10.1186/s12859-017-1934-z (on Google Scholar) Schneider, C. A.; Rasband, W. S. & Eliceiri, K. W. (2012), "NIH Image to ImageJ:...