Computer-aided diagnosis in radiology: potential and pitfalls

https://doi.org/10.1016/S0720-048X(99)00016-9Get rights and content

Abstract

Computer-aided diagnosis (CAD) may be defined as a diagnosis made by a physician who takes into account the computer output as a second opinion. The purpose of CAD is to improve the diagnostic accuracy and the consistency of the radiologists’ image interpretation. This article is to provide a brief overview of some of CAD schemes for detection and differential diagnosis of pulmonary nodules and interstitial opacities in chest radiographs as well as clustered micro-calcifications and masses in mammograms. ROC analysis clearly indicated that the radiologists’ performances were significantly improved when the computer output was available. An intelligent CAD workstation was developed for detection of breast lesions in mammograms. Results obtained from the first 10 000 cases indicated the potential of CAD in detecting approximately one-half of ‘missed’ breast cancer.

Introduction

Over the last decade or so, many investigators have carried out basic studies and clinical applications toward the development of modern computerized schemes for detection and characterization of lesions in radiologic images, based on computer vision and artificial intelligence. These methods and techniques are generally called computer-aided diagnosis (CAD) schemes. The development of CAD has now reached a new phase, since the first commercial unit for detection of breast lesions in mammograms by R2 Technology, Inc., Los Altos, CA, was approved in June, 1998, by the Food and Drug Administration (FDA) for marketing and sale for clinical use in the United States. It is likely that this will expedite clinical applications of the CAD schemes and also further the development of various CAD schemes in different diagnostic examinations. Our purpose in this article is to provide a brief overview of some of the current CAD schemes developed at the University of Chicago for chest radiography and mammography, and to discuss some issues related to the future potential and pitfalls of CAD.

Section snippets

Basic concept of computer-aided diagnosis

CAD may be defined as a diagnosis made by a physician who takes into account the results of the computer output as a ‘second opinion’ [1], [2], [3], [4], [5], [6]. In radiology, the computer output is derived from quantitative analysis of diagnostic images. It is important to note that the computer is used only as a tool to provide additional information to clinicians, who will make the final decision as to the diagnosis of a patient. Therefore, the basic concept of CAD is clearly different

Computer-aided diagnosis in chest radiography

It is a difficult task for radiologists to detect some lung lesions, such as nodules. It is well documented that radiologists may miss up to 30% of lung nodules in chest radiographs, because of the camouflaging effect of the normal anatomic background. It is expected, therefore, that the computer prompt by indicating potential sites of nodules would improve their detection accuracy. The computerized scheme for detection of nodules [15], [17] is shown in Fig. 1. The digital or digitized chest

Computer-aided diagnosis in mammography

Mammography is considered the most reliable method for early detection of breast cancers at present. However, it is difficult for radiologists to detect breast lesions such as clustered microcalcifications and masses on mammograms. It is well known that radiologists may miss 15–30% of these lesions. Therefore, the computer output indicating the potential sites of lesions may be useful to assist radiologists’ interpretation of mammograms, especially in mass screening, due to the fact that the

Discussion and conclusion

We believe that there is a strong evidence of the potential benefit of CAD in the detection and characterization of some lesions in chest radiography and mammography. However, it is important to be cautious about potential pitfalls associated with the use of the computer output. Advances in science and technology can bring many benefits, but also can be harmful if not used properly. Potential pitfalls of CAD can occur for all four of the possible outcomes, namely, false positives and false

Acknowledgements

Authors are grateful to numerous current and former members in the Rossmann Lab for their contribution to the development of CAD schemes, and also Mrs E. Lanzl for improving the manuscript. This work was supported by the NIH (USPHS Grants CA 60187, CA 64370, CA 62625, AR 42739, HL 52567, T32 CA 09649, RR 11459), US Army Medical Research and Materiel Command (DAMD 17-94-J-4071, DAMD 17-96-1-6058, DAMD 17-96-1-6228, DAMD 17-96-1-6229, DAMD 17-97-1-7202), the American Cancer Society (FRA-390), and

References (37)

  • F.F. Yin et al.

    Computerized detection of masses in digital mammograms: Analysis of bilateral subtraction images

    Med. Phys.

    (1991)
  • H.P. Chan et al.

    Image feature analysis and computer-aided diagnosis in digital radiography. 1. Automated detection of microcalcifications in mammography

    Med. Phys.

    (1987)
  • H.P. Chan et al.

    Improvement in radiologists’ detection of clustered microcalcifications on mammograms: The potential of computer-aided diagnosis

    Invest. Radiol.

    (1990)
  • R.M. Nishikawa et al.

    Computer-aided detection of clustered microcalcifications: An improved method for grouping detected signals

    Med. Phys.

    (1993)
  • W. Zhang et al.

    An improved shift-invariant artificial neural network for computerized detection of clustered microcalcifications in digital mammograms

    Med. Phys.

    (1996)
  • Y. Jiang et al.

    Malignant and benign clustered microcalcifications: Automated feature analysis and classification

    Radiology

    (1996)
  • M.L. Giger et al.

    Image feature analysis and computer-aided diagnosis in digital radiography. 3. Automated detection of nodules in peripheral lung fields

    Med. Phys.

    (1988)
  • T. Kobayashi et al.

    Effect of a computer-aided diagnosis scheme on radiologists’ performance in detection of lung nodules on radiographs

    Radiology

    (1996)
  • Cited by (208)

    View all citing articles on Scopus
    View full text