Evaluation of computer-aided detection (CAD) prompting techniques for mammography.

Computer-aided detection (CAD) systems, in which abnormalities are automatically detected and their locations presented to the radiologist as prompts, are increasingly being used to improve reader performance. The performance of CAD systems can be evaluated in two ways: by measuring the performance of the algorithms, or by monitoring the performance of readers using the system. All aspects of evaluation need careful consideration to avoid potential bias. This paper examines a variety of different approaches to evaluation and discusses their relative strengths and weaknesses.

Computer-aided detection in mammography.

Mammographic film reading for breast screening is a highly demanding visual task involving a detailed visual search for signs of abnormality, which are infrequent and often small or subtle. False-negative cases, in which a cancer is missed by a film reader, are known to occur. Although double reading has proved effective in reducing errors, there is a national shortage of film readers in the screening programme, and recent extensions to the programme have exacerbated this problem. The use of computer-aided detection (CAD) systems could potentially provide a solution by improving individual performance to the extent that double reading is no longer necessary. In this paper, we describe how CAD works, review the relevant literature and examine the strengths and weaknesses of the approach.

Computer-based detection and prompting of mammographic abnormalities.

Mammographic film reading is a highly demanding task, particularly in screening programmes where the reader must perform a detailed visual search of a large number of images for early signs of abnormality, which are often subtle or small, and which occur very infrequently. False negative cases, where signs of abnormality are missed by a film reader, are known to occur. Computer based algorithms can be used to detect abnormal patterns in images, but it is not possible to reliably detect all signs of abnormality in mammograms, so screening cannot yet be fully automated. The most successful detection algorithms are, however, incorporated in computer-aided detection (CAD) systems which indicate potentially abnormal locations to the reader in a process known as prompting. CAD systems have the capacity to reduce the frequency of false negative errors by ensuring that suspicious regions of the images are thoroughly searched and by increasing the weighting attached to subtle signs that may otherwise have been dismissed. One of the areas currently being researched is the effect of prompting on human performance. This is complex, since readers are presented with prompts generated by multiple detection algorithms, each of which has a different sensitivity and specificity. This paper reviews progress in abnormality detection, the strengths and the weaknesses of CAD, and the methodologies used to evaluate CAD in a clinical setting.

Computer-assisted mammographic imaging.

Computer-assisted mammography imaging comprises computer-based analysis of digitized images resulting in prompts aiding mammographic interpretation and computerized stereotactic localization devices which improve location accuracy. The commercial prompting systems available are designed to draw attention to mammographic abnormalities detected by algorithms based on symptomatic practise in North America. High sensitivity rates are important commercially but result in increased false prompt rates, which are known to distract radiologists. A national shortage of breast radiologists in the UK necessitates evaluation of such systems in a population breast screening programme to determine effectiveness in increasing cancer detection and feasibility of implementation.

Model-based detection of spiculated lesions in mammograms.

Computer-aided mammographic prompting systems require the reliable detection of a variety of signs of cancer. In this paper we concentrate on the detection of spiculated lesions in mammograms. A spiculated lesion is typically characterized by an abnormal pattern of linear structures and a central mass. Statistical models have been developed to describe and detect both these aspects of spiculated lesions. We describe a generic method of representing patterns of linear structures, which relies on the use of factor analysis to separate the systematic and random aspects of a class of patterns. We model the appearance of central masses using local scale-orientation signatures based on recursive median filtering, approximated using principal-component analysis. For lesions of 16 mm and larger the pattern detection technique results in a sensitivity of 80% at 0.014 false positives per image, whilst the mass detection approach results in a sensitivity 80% at 0.23 false positives per image. Simple combination techniques result in an improved sensitivity and specificity close to that required to improve the performance of a radiologist in a prompting environment.

3-D histo-radiographic comparison of surgical clearances of microcalcified lesions in breast localization biopsies.

There is controversy as to the value of the radiological or pathological estimation of surgical clearance of microcalcifying breast lesions. An important part of this issue has been addressed by coordinated three-dimensional radiographic and histological examination of a prospective consecutive series of 40 benign and malignant mammographically detected lesions in surgical breast biopsy specimens containing microcalcifications, including 20 cases of ductal carcinoma in situ. They were radiographed from four viewpoints by means of rotation in a radiolucent tetrahedral container. The planes of histological examination were then chosen to correspond to the radiographic view showing the minimum separation of the edge of the specimen and the outermost microcalcification. There was a close correlation (Spearman ranked) between the least tetrahedral radiographic distance and the corresponding histological distance separating the surgical margin of excision. There were, however, incompatible Wilcoxon signed ranking orders when comparing the least tetrahedral distance or the histological distance with all four single radiographic views, including the conventional specimen radiographic view. Two-dimensional specimen mammography and standardized histological examination are suboptimal and may thus have contributed to confusion as to the value of determining adequate surgical excision of ductal carcinoma in situ of the breast. Although labour-intensive, use of four-view radiography and choice of the appropriate plane of histological examination give a better correlation of the radiographic estimates of surgical clearance with histology than single-view specimen radiography and arbitrary histological sectioning.

New avenues in 3-D computerised (stereopathological) imaging of breast cancer (review).

Multimodal methods of three-dimensional (3-D) imaging of breast cancer are described. These involve scanning confocal microscopy, using 50 MHz acoustic or near-infrared images, four-view (tetrahedral) radiography and x-ray projection microscopy. Computerised volume data from these techniques can be used to produce three-dimensional images of tissue ranging from 500 microns to approximately 4 mm in thickness. Preliminary findings indicate that stereoscopic images or 3-D computerised reconstructions are capable of advancing the understanding of the structure of ductal carcinoma in situ, lesions simulating microinvasive breast carcinoma, surgical clearance of high-grade calcifying ductal carcinoma in situ, and the 3-D growth patterns of invasive forms of breast carcinoma. In the future computerised image fusion techniques seem likely to be able to take advantage of multimodal imaging of breast cancers, thus correcting primary imaging artefacts, improving robustness, and combining complementary information. In addition, the use of computerised tetrahedral radiography may change the intraoperative assessment of breast cancers, which mostly depend at present upon subsequent laboratory procedures that take days to perform.

Human papilloma viruses: a study of their prevalence in the normal larynx.

The association of human papilloma viruses (HPV) with laryngopharyngeal squamous cell carcinoma is under investigation. The suitability of control tissue in the reported series, invariably obtained from histologically normal tissue adjacent to a squamous cell carcinoma or from patients with benign laryngopharyngeal disease, is questionable. The present study determined the prevalence of HPV in a series of normal larynges. Twelve autopsy larynges were collected. DNA was obtained by SDS proteinase K digestion. Evidence of HPV infection was documented by the polymerase chain reaction using oligonucleotide primers complementary to sequences in the E6 region of HPV types 11, 16 and 18. Four female and eight male larynges, mean age 65 years (SD = 16 years) were collected 72 hours postmortem (median value). HPV type 11 was isolated from three specimens. A 25 per cent prevalence rate for HPV 11 was found. No other HPV types were isolated.