academic article review

academic article review

The Impact of Artificial Intelligence on Healthcare

1. Introduction

The effect of AI in the health kiosk scenario is to provide a type of low-level healthcare to a greater portion of the populace. Such technology can alleviate the many economically based ailments the world faces. In the nearer term, simple analysis programs are enabling non-specialists to give a preliminary diagnosis. There are a myriad of such programs in existence for all variety of medical issues. While the quality of such a tool varies greatly, it has been shown in some cases to provide a diagnosis as good as that of a human. A concrete example of cost reduction and increased efficacy in the medical sector is the use of AI in culling and diagnosis of abnormalities in medical imaging. This has traditionally been a highly error-prone and manual process. High-quality diagnosis is a complex and painstaking endeavor. For this reason, it is highly suitable for automation.

The profusion of data in the medical sector and the high cost of healthcare are strong motivators. AI has the promise of addressing both problems. It can process the copious amount of data that the health sector is generating, and, by developing algorithms, it can provide a tentative analysis of an ailment and suggest a course of action. Treatment can be meted out at lower cost and/or with higher efficacy. A likely scenario in the future is that a patient will visit a health kiosk and receive an analysis that rivals that of a human doctor, with much reduced cost. If a serious ailment is diagnosed, the patient may then consult with a human specialist.

The research of artificial intelligence (AI) was initially yielding results that were fairly abstract and theoretical. Nevertheless, we have recently seen a trend towards bringing the technology to bear on real problems. There is a growing awareness that AI carries a great potential for increasing the effectiveness of the medical sector.

2. Benefits of Artificial Intelligence in Healthcare

Diagnosis is the area in which AI is seeing the most growth. A number of AI applications have been developed to diagnose cancer and other diseases. For example, researchers at Houston Methodist Cancer Centre have developed an AI software to identify patients with oesophageal cancer. The machine learning software was trained with 79 patients, and when tested it managed to identify all 79 patients, whereas previous methods using bronchoscopy only identified 65. A team of Japanese researchers have also developed a method to analyse camera images in an effort to identify early-stage stomach cancer. Although it has only been tested on early-stage stomach cancer, it has a success rate of 80%, and is far less invasive than standard endoscopy, and so could be a lot less stressful for the patient. In the UK, researchers at University College London have used an unsupervised machine learning algorithm to develop an AI platform which analyses MRI scans to predict the growth rate of brain tumours. Another very important factor in diagnosis is in the analysis of various risk factors. For example, researchers at Stanford University have developed an AI platform which can predict the onset of Alzheimer’s disease by studying the patterns of an individual’s natural speech. It was a study of 40 patients over 5 years, and during this time 14 of the patients developed Alzheimer’s. The AI managed to identify these 14 patients, and also identified 13 of the remaining 26 patients who didn’t develop Alzheimer’s. Although this diagnostic method is years away from being used practically, it has the potential to identify the disease much earlier than current methods by which time it is often too late to treat the disease effectively.

3. Challenges and Concerns

The inherent challenge of predicting and treating the health of a complex adaptive system, such as the human body, has perpetually haunted AI researchers. Despite the immense promise and potential of AI in healthcare, the growth and widespread application of AI systems in clinical/health research practice has been slow and complex. There are high expectations that AI systems will be able to use patient data with the intention of tracking health outcomes, and possibly even assist physicians in clinical practice. Such applications have the potential to increase the quality of healthcare, but have raised several concerning issues. A foremost concern is that AI systems using machine learning will have imperfect learning from data that is not quite accurate and that may lead to amplified health disparities. There are concerns that the use of patient data and the interaction with AI systems could impair the patient-physician relationship. Other concerns are the increased complexity and specialization of AI systems which could lead to testing and implementation crises, and that the AI health industry might operate under different standards and regulations to other areas of healthcare and clinical practice. These concerns underscore a great need for the rigorous evaluation of the impact of AI in health, through assessment of different AI applications, comparison with existing alternatives, and consideration of AI as a suite of technologies as opposed to a single intervention.

4. Case Studies: Successful Implementation of AI in Healthcare

Various case studies have shown the potential for AI to improve the standard of healthcare. For example, at Memorial Sloan Kettering Cancer Centre (MSKCC), IBM’s Watson computer is being trained to help diagnose and treat cancer. When a patient is diagnosed with cancer, the standard protocol is to conduct genetic testing to see whether they have a mutation that makes them eligible for treatment with a specific drug. In many cases, genetic mutations can be very complex and it can be difficult for physicians to stay on top of every possible treatment option. In an attempt to make this process more efficient, MSKCC has trained Watson to process and understand genetic data, and to provide information and recommendations to oncologists to match patients with an appropriate treatment strategy. MSKCC believes that this will be a much more efficient way of matching patients with treatment strategies, and that Watson has the potential to match patients with clinical trials that they would not have been aware of. Although it is still early days for this project, it is an encouraging example of AI being used to augment and assist medical decision making.

5. Conclusion and Future Directions

Looking towards the future, certain projected changes in the healthcare and the biomedical ecosystem will impact the utilization and development of AI in clinical contexts. The convergence of healthcare and biomedical research toward ‘precision medicine’ will rely on the ability to create patient taxonomies that subdivide diagnostic categories into more specific patient subgroups. This shift aligns well with AI’s ability to make predictions at the level of an individual patient, rather than at the group level. Dissemination of AI technologies in clinical context will benefit from a new generation of physicians who have been raised using digital tools and are comfortable with the concepts of automation and data-driven decision making. Finally, the translational pull from industry is to create more ‘validated’ predictive models and decision support tools that can be sold as part of a clinical product.

In conclusion, AI applications have the ability to transform healthcare by automating and augmenting cognitive tasks. The availability of vast amounts of data in healthcare, combined with the ever-increasing power of computing, will continue to drive AI research and applications. We foresee great progress in the diagnosis and management of disease, ultimately leading to highly individualized patient care. The potential of AI to reduce the financial burden of healthcare through the automation of administrative and clinical processes is immense. Successful implementation of AI systems in healthcare will depend on the ability to integrate AI processes into existing clinical workflow. This will require re-engineering clinical documentation and decision-making processes. Regulatory and reimbursement issues will need to be addressed in order to provide the necessary changes for incentives to adopt AI technologies.