Predicting location of breast lesions in supine position from prone MRI data using machine learning.

Congress:

ECR 2019

Poster Number:

C-1813

Type:

Scientific Exhibit

Keywords:

Image registration, Cancer, Localisation, MR, Computer applications, Breast, Artificial Intelligence

Authors:

L. Cherkezyan PhD¹, E. Aribal², O. Buğdaycı², C. Caluser³, Y. Lei⁴, Z. Anwar⁵; ¹Evanston/US, ²Istanbul/TR, ³Glen Ellyn, IL/US, ⁴Chicago/US, ⁵Napperville/US

DOI:

10.26044/ecr2019/C-1813

DOI-Link:

https://dx.doi.org/10.26044/ecr2019/C-1813

Conclusion

Our ML-based model shows promise in predicting lesion location in supine position from prone MRI images with a mean error of <17mm, which is likely to improve with adding more data sets and predictors. Future inclusion of automated detection of landmarks will further expedite the calculation of the supine target predicted position.

The obtained results support the feasibility of using trained algorithms to predict the position of small prone MRI-detected breast lesions in supine position and provide guidance for supine breast ultrasound exams or other supine procedures.