Louise Wilkinson

An Ontology to Support Adaptive Training for Breast Radiologists

Medical education and training increasingly rely on computer-based tools. A number of initiatives incorporate digital libraries in tools to train radiologists. Our research involves the use of an informatics infrastructure to access a database of annotated images. We argue that an intelligent training tool requires a rich annotation of images in the database. In order to allow for the flexible querying of the database and intelligent feedback to trainees, those annotations must be organised using a clear and explicit model of the relevant concepts: an ontology. The paper reviews existing work on ontologies for mammography and outlines a new approach which is (a) derived from a detailed analysis of a large number of cases and (b) rich enough to meet the requirements of a training tool.

Design and validation of realistic breast models for use in multiple alternative forced choice virtual clinical trials

A novel method has been developed for generating quasi-realistic voxel phantoms which simulate the compressed breast in mammography and digital breast tomosynthesis (DBT). The models are suitable for use in virtual clinical trials requiring realistic anatomy which use the multiple alternative forced choice (AFC) paradigm and patches from the complete breast image. The breast models are produced by extracting features of breast tissue components from DBT clinical images including skin, adipose and fibro-glandular tissue, blood vessels and Coopers ligaments. A range of different breast models can then be generated by combining these components. Visual realism was validated using a receiver operating characteristic (ROC) study of patches from simulated images calculated using the breast models and from real patient images. Quantitative analysis was undertaken using fractal dimension and power spectrum analysis. The average areas under the ROC curves for 2D and DBT images were 0.51  ±  0.06 and 0.54  ±  0.09 demonstrating that simulated and real images were statistically indistinguishable by expert breast readers (7 observers); errors represented as one standard error of the mean. The average fractal dimensions (2D, DBT) for real and simulated images were (2.72  ±  0.01, 2.75  ±  0.01) and (2.77  ±  0.03, 2.82  ±  0.04) respectively; errors represented as one standard error of the mean. Excellent agreement was found between power spectrum curves of real and simulated images, with average β values (2D, DBT) of (3.10  ±  0.17, 3.21  ±  0.11) and (3.01  ±  0.32, 3.19  ±  0.07) respectively; errors represented as one standard error of the mean. These results demonstrate that radiological images of these breast models realistically represent the complexity of real breast structures and can be used to simulate patches from mammograms and DBT images that are indistinguishable from patches from the corresponding real breast images. The method can generate about 500 radiological patches (~30 mm  ×  30 mm) per day for AFC experiments on a single workstation. This is the first study to quantitatively validate the realism of simulated radiological breast images using direct blinded comparison with real data via the ROC paradigm with expert breast readers.

Assessing mammography film-reader performance

Mammography film-readers in the National Health Service Breast Screening Programme are required to read 5,000 screening mammograms per year, and to participate in Personal Performance in Mammographic Screening (PERFORMS). It is known that individual cancer detection rates are higher when individuals perform high volumes of breast imaging annually, and that poor performance may be identified by PERFORMS. We have looked more closely at details of reader performance to guide individuals on how practice may be improved.

Design and validation of biologically inspired spiculated breast lesion models utilizing structural tissue distortion

The use of conventional clinical trials to optimise technology and techniques in breast cancer screening carries with it issues of dose, high cost and delay. This has motivated the development of Virtual Clinical Trials (VCTs) as an alternative in-silico assessment paradigm. However, such an approach requires a set of modelling tools that can realistically represent the key biological and technical components within the imaging chain. The OPTIMAM image simulation toolbox provides a complete validated end-to-end solution for VCTs, wherein commonly-found regular and irregular lesions can be successfully and realistically simulated. As spiculated lesions are the second most common form of solid mass we report on our latest developments to produce realistic spiculated lesion models, with particular application in Alternative Forced Choice trials. We make use of sets of spicules drawn using manually annotated landmarks and interpolated by a fitted 3D spline for each spicule. Once combined with a solid core, these are inserted into 2D and tomosynthesis image segments and blended using a combination of elongation, rotational alignment with background, spicule twisting and core radial contraction effects. A mixture of real and simulated images (86 2D and 86 DBT images) with spiculated lesions were presented to an experienced radiologist in an observer study. The latest observer study results demonstrated that 88.4% of simulated images of lesions in 2D and 67.4% of simulated lesions in DBT were rated as definitely or probably real on a six-point scale. This presents a significant improvement on our previous work which did not employ any background blending algorithms to simulate spiculated lesions in clinical images.

Impact of two types of image processing on cancer detection in mammography

The impact of image processing on cancer detection is still a concern to radiologists and physicists. This work aims to evaluate the effect of two types of image processing on cancer detection in mammography. An observer study was performed in which six radiologists inspected 349 cases (a mixture of normal cases, benign lesions and cancers) processed with two types of image processing. The observers marked areas they were suspicious were cancers. JAFROC analysis was performed to determine if there was a significant difference in cancer detection between the two types of image processing. Cancer detection was significantly better with the standard setting image processing (flavor A) compared with one that provides enhanced image contrast (flavor B), p = 0.036. The image processing was applied to images of the CDMAM test object, which were then analysed using CDCOM. The threshold gold thickness measured with the CDMAM test object was thinner using flavor A than flavor B image processing. Since Flavor A was found to be superior in both the observer study and the measurements using the CDMAM phantom, this may indicate that measurements using the CDMAM correlate with change in cancer detection with different types of image processing.

Quantitative study: should vacuum-assisted biopsy be the first biopsy approach in breast interventional techniques in stereotactic guided microcalcifications rather than 14 gauge core needle biopsy?

Stereotactic guided 14 gauge core needle biopsy (14GCNB) and vacuum-assisted biopsy (VAB) are the two commonly used biopsy methods for obtaining an accurate diagnosis for microcalcifications. Retrospective review of 399 patients who underwent biopsy for breast microcalcification during screening assessment from April 2012 to March 2013 was used to evaluate the performance and cost-effectiveness of both methods.

Impact of CAD with full field digital mammography on workflow and cost

The cost-effectiveness of single reading with CAD as an alternative to double reading was assessed in a national screening programme using CAD with full field digital mammography The impact of CAD on the time taken to read screening films (n=5710) and on the proportion of films referred for arbitration and for assessment was measured (n=3064) No evidence was found of a change in the time taken to read films and no evidence of a change in rates of referral or recall Estimates of the cost implications were made under three different scenarios for screening units We conclude that single reading with CAD is likely to be a cost-effective alternative to double reading in terms of radiologist time Published data however shows increased recall rates using CAD and no significant increase in sensitivity for CAD use over single reading Any decision to introduce CAD instead of double reading should take into account the impact of sensitivity and specificity on women attending for screening.