Computer Vision News - June 2022

45 AI in Breast Cancer Screening tracking during the procedure to increase accuracy and reduce procedural time . The Breast Imaging Reporting and Data System (BI-RADS) was developed to give specific guidelines for reporting findings in breast screening images. These guidelines apply to mammography, MRI, and ultrasound , and they account for the lesions size, volume, shape, homogeneity, restriction, and other features. All these parameters can be automatically extracted from the imaging data, and presented to the radiologist for consideration, saving even more precious time and increasing physician confidence . As described above, breast cancer takes a large toll on the medical industry. Numerous scans are conducted daily, and early detection is the main factor affecting survival rate. Introducing AI modules into the screening and diagnosis pipeline can reduce scan and radiologist review time, lower costs, and increase detection and survival rate . There are many methods for AI implementation, and RSIP Vision can assist in the development of any of the aforementioned solutions. screening and detection of breast cancer. Breast MRI exams are often long - there are multiple sequences to acquire, and breathing causes movement which needs to be accounted for. This issue can be addressed by accurate fusion of the acquired sequences . Deep learning methods can be applied and register each sequence to a predetermined baseline sequence, like Diffusion Weighted Imaging (DWI) toT2, anddiscard the relative motion between and during acquisition. Additionally, resolution can be enhanced by training deep neural networks to reconstruct MRI images from suboptimal acquisitions . Both applications can reduce acquisition time as the need for repeat acquisition is reduced. In MRI, like mammography, the resulting image needs to be reviewed by the radiologist. Computer vision systems canbe developed to support the analysis of breast MRI images. These systems can detect and segment suspicious lesions, calculate their intensity in multiple sequences, and report these results to the radiologist, to be used as the baseline for interpretation and diagnosis. Ultrasound: Breast US is performed to further examine women with dense breast tissue or with findings in other imaging exams. Automated breast ultrasound (ABUS) is a relatively new approach, where a large transducer is placed on the breast and automatically acquires multiple images to cover the entire breast. Also here, AI can be used for automatic detection of abnormalities , and with sufficient data even classification of malignancy . This can reduce the miss rate significantly. Additionally, ultrasound is often used for biopsy guidance, and Artificial Intelligence methods can be applied to provide needle guidance and Breast MRI - MIP with subtraction showing enhanced procedure in left breast