Use this URL to cite or link to this record in EThOS:
Title: Breast imaging beyond 2D : from screening to treatment
Author: Pöhlmann, Stefanie
ISNI:       0000 0005 0288 9648
Awarding Body: University of Manchester
Current Institution: University of Manchester
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Access from Institution:
Background: Breast cancer is the most common cancer in women worldwide, affecting one in eight. It can be detected by X-ray mammography, which is used in screening programs to find cancer at an early stage when treatment is often effective. Aim: The aim of this work was to overcome some of the limitations associated with mammography and interpret information "beyond 2D" from Digital Breast Tomosynthesis (DBT) and surface imaging in a clinically relevant way. Methods: Unmet user requirements in current breast imaging practice were identified and the properties of DBT and surface images were investigated with quantitative image analysis in mind. New imaging analysis methodology was developed with applications in cancer risk estimation, cancer assessment and treatment planning. Three separate studies were conducted to assess feasibility and clinical performance. Breast density, a well established risk factor for breast cancer, was calculated from mammograms and DBT images of 35 women with different methods and compared. 3D segmentations from DBT were validated with histo-pathological assessment of 20 breast tumours. The feasibility of using Microsoft Kinect, an inexpensive depth sensor for gaming applications, was demonstrated using a phantom, then assessed in 10 women undergoing reconstructive surgery. Results: DBT provides spatial information about breast composition in reconstructed image volumes, but resolution is highly anisotropic and images exhibit artefacts. It is currently not possible to distinguish between different breast tissue types and imaging artefacts overlapping anatomical structures for the breast as a whole. The Microsoft Kinect can map the surface of the breast with adequate precision (error < 1 mm), but performance varies with imaging conditions. Although it is not accurate to analyse the composition of the breast from DBT reconstructions due to significant artefacts, good correlation (p=0.95) with the mammographic gold standard can be achieved when analysing DBT projections with volumetric methods in combination with pre-processing. We have found that it is possible to distinguish between tissue structure and artefacts in the context of extracting 3D morphology of breast lesions and achieved similar correlation with histological sizing than expert annotation (p=0.69) and better agreement with tumour volumes estimated from pathology assessment (95% limits of agreement -16 to 11 ml). Using the Kinect for planning implant based breast reconstruction showed similar accuracy than achieved by experienced breast surgeons or using mastectomy specimen weight as guidance (standard error 120 ml, 141 ml and 142 ml, respectively). Conclusion: Quantitative assessment of both DBT and surface imaging could improve breast imaging practice and help clinicians to estimate breast cancer risk reliably, assess tumour size and morphology and plan breast reconstruction.
Supervisor: Taylor, Christopher ; Astley, Susan Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: Breast Reconstruction ; Microsoft Kinect ; Image Segmentation ; Breast Density ; Breast Cancer ; Breast Imaging ; Tomosynthesis