Development and preclinical evaluation of a patient‐specific high energy x‐ray phase sensitive breast tomosynthesis system

Abstract

Background: This article reports the first x-ray phase sensitive breast tomosynthesis (PBT) system that is aimed for direct translation to clinical practice for the diagnosis of breast cancer. Purpose: To report the preclinical evaluation and comparison of the newly built PBT system with a conventional digital breast tomosynthesis (DBT) system. Methods and materials: The PBT system is developed based on a comprehensive inline phase contrast theoretical model. The system consists of a polyenergetic microfocus x-ray source and a flat panel detector mounted on an arm that is attached to a rotating gantry. It acquires nine projections over a 15°angular span in a stop-and-shoot manner. A dedicated phase retrieval algorithm is integrated with a filtered back-projection method that reconstructs tomographic slices. The American College of Radiology (ACR) accreditation phantom, a contrast detail (CD) phantom and mastectomy tissue samples were imaged at the same glandular dose levels by both the PBT and a standard of care DBT system for image quality characterizations and comparisons. Results: The PBT imaging scores with the ACR phantom are in good to excellent range and meet the quality assurance criteria set by the Mammography Quality Standard Act. The CD phantom image comparison and associated statistical analyses from two-alternative forced-choice reader studies confirm the improvement offered by the PBT system in terms of contrast resolution, spatial resolution, and conspicuity. The artifact spread function (ASF) analyses revealed a sizable lateral spread of metal artifacts in PBT slices as compared to DBT slices. Signal-to-noise ratio values for various inserts of the ACR and CD phantoms further validated the superiority of the PBT system. Mastectomy sample images acquired by the PBT system showed a superior depiction of microcalcifications vs the DBT system. Conclusion:The PBT imaging technology can be clinically employed for improving the accuracy of breast cancer screening and diagnosis.

Assessing detectability improvement of a self-supervised noise reduction algorithm for phase-sensitive breast tomosynthesis phantom images

“…For this study, four different phantoms were imaged: an ACR Mammography Phantom (Mammo 156, Gammex, Inc.) [3], a 6x6 contrast detail phantom (CIRS Inc.) [4], a 7x7 contrast detail phantom (CIRS Inc.) [1], and a 9x10 contrast detail phantom (Mammo 180, Gammex, Inc.) [5]. All the phantoms simulate the x-ray attenuation properties similar to 50% glandular and 50% adipose breast tissue.…”

A phase sensitive x-ray breast tomosynthesis system: Preliminary patient images with cancer lesions

“…The PBT system was designed and developed in an effort led by academic researchers in collaboration with an industrial partner. The system has been characterized for preclinical image quality performance (Ghani et al 2021) and is currently undergoing testing in a clinical trial. The complete technical details and a picture of the system can be found elsewhere (Ghani et al 2021), while some of the major components of the PBT system are summarized in table 1.…”

“…The system has been characterized for preclinical image quality performance (Ghani et al 2021) and is currently undergoing testing in a clinical trial. The complete technical details and a picture of the system can be found elsewhere (Ghani et al 2021), while some of the major components of the PBT system are summarized in table 1. For a PBT scan, the patient breast is compressed between two compression plates such that the source to breast isocenter distance is 70 cm, and the breast isocenter to detector distance is 84 cm, resulting in a geometric magnification (M) of 2.2.…”

Fuzzy Weighted Median Filter With Unsharp Masking for Enhancement of DBT Images in Breast Cancer Detection

“…Algorithms or methods will be used to remove noise from DBT images. Some noise filtering techniques will be used, while thresholding methods will be used to enhance the DBT images (9)(10)(11)(12)(13)(14)(15)(16). Many researchers, however, use unsharp masking to aid their proposed model, such as a deep learning model, and use it as the main proposed enhancement algorithm.…”