Hello. Sign In
Standards Store

Artificial Neural Networks in Cancer Diagnosis, Prognosis, and Patient Management

2001 Edition, June 22, 2001

Complete Document

Detail Summary

Active, Most Current

Additional Comments:
ISBN: 978-0-8493-9692-2
Price (USD)
Add to Cart

Product Details:

  • Revision: 2001 Edition, June 22, 2001
  • Published Date: June 22, 2001
  • Status: Active, Most Current
  • Document Language: English
  • Published By: CRC Press (CRC)
  • Page Count: 217
  • ANSI Approved: No
  • DoD Adopted: No

Description / Abstract:


The potential value of artificial neural networks (ANNs) as a predictor of malignancy has now been widely recognised. The concept of ANNs dates back to the early part of the 20th century; however, their latest resurrection started in earnest in the 1980s when they were applied to many problems in the areas of pattern recognition, control, and optimisation. Here we present a series of articles that emphasise the keen interest displayed by the scientific community in the application of neural networks in the management of human cancers, and also reflect the recent intense activity in this field. The neural systems use prognostic cancer markers as input neurons and provide an ideal means of combining the input signals so that the output neurons can provide a predictive basis on which to determine the course of patient management. In essence, the two characteristics that determine the reliability of neural networks are the discriminant analysis of the input variables and the correct classification of tumours with regard to the accuracy of the predictive output, whether in the form of diagnosis, its spread to regional lymph nodes, or patient survival. This versatility of the artificial neural networks is aptly encapsulated by the Tamil quotation.

One of the main objectives of this book is to ensure that the material reported in it emanates from a variety of institutions, in order that an objective and unbiased view of the different approaches undertaken by researchers in those institutions can be presented. Thus, the book contains chapters relating to different types of cancer. These have been written by leading international researchers in the field of artificial neural networks, as well as by leading experts in oncology, physicians, pathologists, and surgeons from Europe and North and South America.

The subject of artificial neural networks has been extensively treated in the literature and therefore the object of this book is not to present yet another monograph in this area. The aim, however, is to focus solely on their applicability in the field of oncology which, as witnessed by the growing number of publications, is rapidly developing into a major area of research in its own right. Chapter 1, therefore, presents a very brief introduction to ANNs whilst attempting to emphasise their direct relationship to the inherent problems in cancer diagnosis, prognosis, and patient management.

Breast cancer has been the focus of much research lately, especially with regard to the identification of molecular cancer markers. In Chapter 2, Angus et al. discuss some issues relating to prognosis in breast cancer patients. They highlight the clinical importance of molecular, cytological, and histological prognostic markers, including the proteins associated with metastatic potential. They also present an ANN analysis of the significance of the expression of the metastasis-associated h-mts1 and n m 2 3 genes and their individual relationship to nodal spread of breast cancer. This is then followed in Chapter 3 by a discussion of the use of image cytometric measurements of DNA ploidy, the S-phase cell fraction and nuclear pleomorphism as prognostic aids. The practicalities and putative benefits of analysing these data using neural systems are then scrutinised.

The application of ANNs in lung cancer is the subject of Chapters 4 and 5. Chapter 4 is a contribution by Esteva and colleagues. The chapter presents the general issues involved in the prognostic analysis of patients with carcinoma of the lung, and assesses the accuracy of applying artificial neural networks to the prognosis of post-surgical outcome of lung cancer patients. Chapter 5 by Jefferson et al., on the other hand, focuses on the use of a genetic algorithm neural network for prognosis in surgically treated nonsmall cell lung cancer.

In Chapter 6, Speight and Hammond deal with the use of machine learning in screening for oral cancer. The chapter summarises the investigations into the application of ANNs in the selection of high-risk groups of patients, comparing their predictive performance with other machine learning techniques and evaluating the potential performance of machine learning for the detection of high-risk individuals. This chapter also discusses the merits of utilising machine learning for the prediction of risk of oral, and other cancers, as an adjunct to population screening.

The contribution by Wayman and Griffin in Chapter 7 describes the application of ANNs in the prediction of outcome for cancer of the oesophago-gastric junction. As the decision on whether a surgical procedure is necessary or not is made on the interpretation of pre-operative assessments of tumour stage and patient fitness, the correlation between pre- and postoperative findings is frequently poor. Thus, results relating to the application of ANNs to patients undergoing potentially curative resection of adenocarcinoma of the oesophago-gastric junction, both pre- and post-operatively, are reported in this chapter.

It is inevitable, as with any novel and fast growing area of research within a relatively tightly knit and closely collaborating community of researchers, that a degree of overlap will occur. This is particularly evident in the chapters on urological oncology, an area which has perhaps attracted a more concerted effort on the part of clinicians, engineers, and biomedical computing scientists. This is reflected in the number of chapters in this book that deal with this specific area of cancer from both a diagnosis and prognosis viewpoint.

Chapters 8 to 12 therefore represent a comprehensive treatment of this area of machine learning in urological oncology. Chapters 8 (Douglas and Moul) and 9 (Niederberger and Ridout) review the basic concepts of artificial neural networks and summarise their application in renal cell carcinoma (RCC), prostate, bladder, and testicular cancers. These reviews provide an insight into the difficult task of preoperative diagnosis of RCC, issues in prostate cancer diagnosis, outcome prediction, and patients' quality of life. These parameters are increasingly being regarded as important elements in the choice of treatment. The combined application of image analysis procedures and ANNs to identify bladder cells expressing the tumour antigen p300, as well as automated cytology-based interpretation in bladder cancer, are emphasised. Testicular cancer is reviewed, especially from the angle of pathologic stage I vs. stage II prediction in patients with clinical stage I nonseminomatous testicular cancer.

In Chapter 10, Stamey et al. give a detailed description of ProstAsure™, a neural-based technique to predict the risk of prostate cancer in men with a PSA ≤ 4.0 ng/ml, and predict tumour recurrence following radical prostatectomy. Hamdy reports, in Chapter 11, on the use of a neural network to predict prognosis and outcome in prostate cancer. Conventional input variables (age, stage, bone scan findings, grade, PSA, and type of treatment) are used in addition to data derived from the immunohistochemical staining performed on tissue specimens for the proto-oncogene bcl-2 and tumour suppressor gene p53. The presence of increased abnormal expression of these genes has been associated with disease progression in prostate cancer. Hamdy shows that ANNs can be used to assess the benefits, or otherwise, of such newer experimental tumour markers.

Finally, in the field of urological oncology, Chapter 12's authors Qureshi and Mellon present a study relating to the prediction of clinical outcome for patients with bladder cancer utilising prognostic markers identified at initial presentation. In this study, they seek to assess the ability of an ANN to predict the recurrence and stage progression in bladder cancer. This is carried out in a group of patients with newly diagnosed Ta/T1 bladder cancer, and 12-month cancer-specific survival in a group of patients with primary T2-T4 bladder cancer by using clinical, pathological and molecular prognostic indicators. In addition, the networks' predictions are compared with those of four consultant urologists supplied with the same data

The final chapter is a contribution on skin cancer by Hintz-Madsen and his colleagues, where the authors describe a comprehensive study of the applications of ANNs in the diagnosis of melanoma. The accurate detection of this malignancy is heavily dependent upon the precise analysis of skin pigmentation, tumour shape, and colour.

There are several areas of neoplasia that have not been investigated by using the ANN tool. The editors hope this volume will not only emphasise the value of this tool in the study of this disease process but will provide a nucleus around which future work might be planned and executed.

Apart from minor cosmetic changes we have restrained from attempting to change either the structure or contents of each chapter. We sincerely felt that the views and technical contents presented by the authors should give an honest insight into the modes and practices of their efforts in this particularly sensitive, and perhaps contentious, area of research.

The experts who have contributed to this work have certainly made this book move from the possible to the achievable. We wish to thank them most sincerely. Without the time and effort they spent on their respective chapters, the objectives that we set for ourselves would not have been accomplished. Our research students have, over the years, been a tremendous source of inspiration. They have been instrumental in developing ideas and algorithms which, in many instances, seemed initially to be impractical. We are indebted to them for their efforts, original ideas, and many stimulating discussions. Our thanks also go to Sonia Clarke whose editing skills and careful attention to detail have made our task much easier than originally expected. We wish to thank Dr. M.S. Lakshmi for providing the Tamil quotation and its translation. Finally, we also wish to acknowledge the Cancer Research Campaign of the United Kingdom for supporting many aspects of this work.