The use of artificial neural networks in decision support in cancer: A systematic review
Introduction
In the last decade, the use of artificial intelligence (AI) has become widely accepted in medical applications. This is manifested by an increasing number of medical devices currently available on the market with embedded AI algorithms, together with an accelerating pace of publication in medical journals, with over 500 academic publications each year featuring Artificial Neural Networks (ANNs) (Gant, Rodway, & Wyatt 2001). Claimed advantages of neural network methods include:
- •
Ease of optimisation, resulting in cost-effective and flexible non-linear modelling of large data sets.
- •
Accuracy for predictive inference, with potential to support clinical decision making.
- •
These models can make knowledge dissemination easier by providing explanation, for instance, using rule extraction or sensitivity analysis (Lisboa, 2002).
The published literature suggests that ANN models have been shown to be valuable tools in reducing the workload on the clinicians by detecting artefact and providing decision support, potentially with the ability to automatically re-estimate the model on-line. However, there are relatively few published clinical trials, and even fewer testing the clinical value of ANNs against established linear-in-the-parameters statistical methods (Lisboa, 2002).
There are two recurring concerns on ANNs. The first is the use of first principle statistical methods to control model complexity, which has been addressed by regularisation methods and with the use of cross-validation (Biganzoli et al., 1998, Lisboa et al., 2003, Ripley, 1996, Ripley and Ripley, 2001). The second key issue is transparency, i.e. explaining what influences the network predictions and how to resolve outcome predictions in terms of readily understood clinical statements. This is partly addressed by rule-extraction algorithms.
Notwithstanding these concerns, an interesting feature of neural network decision support in medicine is the routine clinical use of a range of systems, from the commercial-C.Net (Nabney, Evans, Tenner, & Gamlyn, 2001) and BioSleep (Tarassenko, McGrogan, & Braithwaite, 2002)—to research prototypes (Lisboa et al., 2000, Taktak et al., 2004) without listing in PubMed of supportive clinical trials. The situation is not specific to neural networks, but extends particularly to web-based decision support tools such as www.adjuvantonline.com, marking a departure from algorithms for clinical routine assessments, e.g. the Glasgow Coma Score for severity of illness in critical care and Nottingham Prognostic Index for breast cancer, both of which have undergone rigorous multi-centre clinical trials evidenced in the literature, if not altogether without controversy.
The use of unstructured approaches to clinical evaluation of new medical research is a trend, which has proved hard to change. Already in 1994 a paper entitled ‘the scandal of poor medical research’ (Altman, 1994) highlighted the need to proper study design bordering on the unethical typically through the application of such bad scientific methodology as to be sometimes called ‘torturing the data’ until they confess to the desired result (Mills, 1993).
Therefore, it is important to define and keep to a staged framework to design a sequence of studies each with a clear-cut purpose, ranging from the exploratory to the definitive, where the chief aim of each step in this chain is to support the next developmental step until a power calculation is possible which will determine the sample size, along with clinical protocol and study design for a multi-centre randomised clinical trial. Such a framework has been published (Campbell et al., 2000) and adapted for the development of intelligent decision support in an earlier review (Lisboa, 2002). This review will note the current trends in the studies that reach journals in the medical or medically related science literature, highlight points of good and poor practice, and draw conclusions for study design to improve the likelihood of studies being appropriately followed-up in the future.
Section snippets
Literature search
A systematic literature search was conducted using Pubmed for entries during the period 1994–2003 with the keywords ‘neural networks’. The search was limited to clinical trials and randomised controlled trials (RCTs). Results of the search are summarised in Fig. 1. The search was repeated using the keywords (neural networks) and (cancer) from 1994 to the current date. There were 396 hits in total with only 27 either CTs or RCTs and the abstracts of the resulting hits were analysed. The
Review of papers related to cancer listed in Pubmed
The majority of clinical trial studies benchmarked the ANNs performance against traditional screening methods. In prostate cancer, this involves the use of prostate specific antigen (PSA) serum marker, digital rectal examination, Gleason sum, age and race (Gamito et al., 2000, Remzi et al., 2003, Stephan et al., 2003, Tewari et al., 2001). Some studies have compared ANNs with statistical methods (Chan et al., 2003, Finne et al., 2000, Matsui et al., 2002, Remzi et al., 2003). Remzi demonstrated
Implications for study design
It is well documented that hundreds of papers are published in the medical literature, at a vast mean cost per published paper, yet few results find their way into improving healthcare practices in routine clinical use. There are reasons for this, partly the unavoidable result that not all interesting new methods turn out to fulfil their early promise. However, more often than not it is methodological shortcomings that mortally damage the future worth of the paper. Some of the reasons for this
Ethical and legal issues
A final consideration with particular implications for the evaluation of biomedical decision support systems concerns the legal and ethical foundation to judge whether the ‘duty of care’ has been breached. The principles involved hark back to the ‘Bolam test’ which refers to the skill of an ordinary competent practitioner. This test offers considerable latitude in the exercise of clinical discretion, a leniency founded on confidence in the doctor's training (Gant et al., 2001). Similar
Conclusions
A review of PubMed listed publications involving clinical trials of neural network systems identified trends in areas of clinical promise, specifically in the diagnosis, prognosis and therapeutic guidance for cancer, but also the need for more extensive application of rigorous methodologies. This has implications for study design, to address some of the more common pitfalls of empirical models for medical diagnosis, particularly those relying on generic non-linear function approximations, which
Acknowledgements
This work is supported by the BIOPATTERN EU Network of Excellence. EU Contract 508803.
References (57)
- et al.
Artificial neural networks for decision-making in urologic oncology
European Urology
(2003) A review of analytical techniques for gait data. Part 2: Neural network and wavelet methods
Gait Posture
(2001)- et al.
Predicting the outcome of prostate biopsy in screen-positive men by a multilayer perceptron network
Urology
(2000) - et al.
Hybrid artificial neural network segmentation and classification of dynamic contrast-enhanced MR imaging (DEMRI) of osteosarcoma
Magnetic Resonance Imaging
(1998) A review of evidence of health benefit from artificial neural networks in medical intervention
Neural Networks
(2002)- et al.
A Bayesian neural network approach for modelling censored data with an application to prognosis after surgery for breast cancer
Artificial Intelligence in Medicine
(2003) - et al.
Bias reduction in skewed binary classification with Bayesian neural networks
Neural Networks
(2000) - et al.
Neural network using combined urine nuclear matrix protein-22, monocyte chemoattractant protein-1 and urinary intercellular adhesion molecule-1 to detect bladder cancer
The Journal of Urology
(2003) - et al.
An artificial neural network to predict the outcome of repeat prostate biopsies
Urology
(2003) - et al.
A non-invasive test for the pre-cancerous breast
European Journal of Cancer
(1995)
Editorial: The scandal of poor medical research
British Medical Journal
What do we mean by validating a prognostic model?
Statistics in Medicine
Feed forward neural networks for the analysis of censored survival data: A partial logistic regression approach
Statistics in Medicine
Neural network for pattern recognition
Using artificial neural networks to screen cervical smears: How new technology enhances health care
Framework for design and evaluation of complex interventions to improve health
British Medical Journal
Detection of prostate cancer by integration of line-scan diffusion, T2-mapping and T2-weighted magnetic resonance imaging; a multichannel statistical classifier
Medical Physics
Artificial neural networks: Opening the black box
Cancer
The diagnostic value of computer-assisted primary cervical smear screening: A longitudinal cohort study
Modern Pathology
Detection of basal cell carcinoma using electrical impedance and neural networks
IEEE Transactions on Biomedical Engineering
Use of artificial neural networks in the clinical staging of prostate cancer: Implications for prostate brachytherapy
Techniques in Urology
Artificial neural networks: Practical considerations for clinical applications
Automated epiluminescence microscopy—Tissue counter analysis using CART and 1-NN in the diagnosis of melanoma
Skin Research and Technology
A computer-aided diagnostic system to characterize CT focal liver lesions: Design and optimization of a neural network classifier
IEEE Transaction on Information Technology in Biomedicine
Cited by (384)
A machine-learning based approach to estimate acoustic macroscopic parameters of porous concrete
2024, Construction and Building MaterialsArtificial Intelligence applications in healthcare: A bibliometric and topic model-based analysis
2024, Intelligent Systems with ApplicationsClassification-based deep neural network vs mixture density network models for insulin sensitivity prediction problem
2023, Computer Methods and Programs in BiomedicineLocating and deploying essential goods and equipment in disasters using AI-enabled approaches: A systematic literature review
2023, Progress in Disaster ScienceRecent advancements in deep learning based lung cancer detection: A systematic review
2022, Engineering Applications of Artificial Intelligence