How can AI be used in healthcare?
The use of Artificial Intelligence (AI) in healthcare is revolutionizing the way medical professionals provide care to patients. From diagnosis to treatment, AI can help doctors and other healthcare professionals make more informed decisions, ultimately leading to better patient outcomes. In this article, Christos Chatzichristos explores some specific cases of AI in healthcare and provides ideas and thoughts on the evolution of the healthcare sector as it uses AI. He ends with a critical conclusion, because in a sensitive domain such as healthcare, the risks cannot be denied.
Medical imaging
AI-powered computer vision algorithms have achieved remarkable accuracy in tasks such as object detection, facial recognition, and image classification. This success in computer vision has also translated to the field of medical imaging, where AI is being used to improve diagnosis and treatment planning.
For example, AI algorithms can analyze medical images such as X-rays, CT scans, and MRIs to detect and classify abnormalities, assisting radiologists in making more accurate diagnoses. AI is also being used to develop predictive models for diseases such as cancer, enabling early detection and more personalized treatment plans.
In a study published in Nature Medicine in 2020, researchers developed an AI system that can collaborate with human experts to accurately identify skin cancer based on images of moles and lesions. The system was able to identify skin cancer with a similar level of accuracy as dermatologists, but at a much faster rate. Similarly, in a recent study published in JAMA Oncology, researchers developed an AI system that can diagnose breast cancer by analyzing digital images of breast tissue. The system achieved a similar level of accuracy as human pathologists, but again, much faster.
Another example of AI in medical imaging is in the detection of diabetic retinopathy. This condition is the leading cause of blindness in working-age adults. Google's DeepMind team developed an AI algorithm that can detect diabetic retinopathy with a high degree of accuracy. The algorithm was trained on a dataset of over 128,000 images and can detect the condition with a sensitivity of 90%.
AI will continue to play an increasingly important role in medical imaging
in the future. AI algorithms continue to improve, they will be able to
analyze medical images with greater speed and accuracy, which will
improve diagnosis, treatment planning, and patient outcomes.
Key areas in which AI is transforming healthcare
Personalized treatment
AI can also be used to provide personalized treatment to patients. One of the challenges in healthcare is that not all patients respond to treatment in the same way. AI can help here: by analyzing patient data, including genetic information and medical history, AI algorithms can identify which treatments are likely to be the most effective for individual patients.
For example, the IBM Watson Health team developed an AI-powered platform that can analyze genetic data to help doctors develop personalized treatment plans for cancer patients. The platform can identify genetic mutations that may make a patient's cancer more resistant to certain treatments and recommend alternative treatments that may be more effective.
Medical Chatbots are computer programs that use natural language processing to interact with users. In healthcare, chatbots are used to provide patients with medical advice and guidance. They can be particularly useful for patients who are unable to see a doctor in person, such as those who live in remote areas or who have mobility issues. For example, the Babylon Health chatbot uses AI to provide medical advice to patients. The chatbot can ask patients questions about their symptoms and medical history, and provide personalized advice based on their answers. If needed, the chatbot connects patients with a doctor for a virtual consultation.
Even in countries like Belgium, where access to a doctor is easier, chatbots can potentially reduce the burden on healthcare providers by providing basic information and advice, freeing up their time to focus on more complex cases. Furthermore, basic healthcare-related questions that are related to health-literacy and would not justify an appointment with a doctor can be easily addressed.
“ We must not use the existence of AI tools as an excuse to decrease quality of healthcare and resources. ”
Drug discovery
Developing new drugs is a complex and time-consuming process. AI can help to speed up the process by analyzing vast amounts of data to identify potential drug candidates and by predicting the efficacy of drug compounds, based on large datasets of chemical structures and biological activity.
AI models was developed that can predict which compounds are likely to interact with specific protein targets in the body,
a crucial step in the process to discover new drugs. DeepDTA, as the
models is called, predicts the binding affinity between small molecules
(drugs) and target proteins. The model was trained on a large dataset of
known drug-target interactions, and the researchers significantly
demonstrated its effectiveness in predicting new interactions.
Drug repurposing, also known as drug repositioning, is the process of discovering new therapeutic uses for existing drugs. Artificial intelligence (AI) has become a powerful tool for drug repurposing, as it can analyze large amounts of data and identify potential new uses for existing drugs. By using AI algorithms to analyze gene expression patterns, drug targets, and other factors, researchers can identify existing drugs that may be effective against a range of diseases, as it has been widely been done during COVID-19 pandemics. This approach has the potential to significantly accelerate the drug development process and lead to the discovery of new treatments for a variety of conditions.
Furthermore AI can be leveraged in order to improve existing methods.
For example, as it has been shown by researchers from Ghent University,
if AI is combined with population pharmacokinetic (PopPK) modeling has
the potential to significantly improve drug development and personalized
medicine. PopPK takes into account the variability in drug absorption,
distribution, metabolism, and excretion that can occur between
individuals, and can be used to optimize drug dosing and minimize the
risk of toxicity. AI algorithms can be used to identify
important predictors of drug concentrations, such as age, weight, and
genetic factors, and to develop more accurate and precise models of drug
behavior. AI can also be used to simulate drug concentrations
in virtual populations, allowing researchers to explore different dosing
regimens and identify optimal treatment strategies. By better
understanding how drugs behave in different patient populations, and by
using AI to optimize dosing and predict outcomes, researchers can
develop more effective and safer treatments for a variety of conditions.
As the field continues to advance, we can expect to see more
innovative approaches and breakthroughs in the discovery of new drugs.
Monitoring and predictive analytics
Machine learning algorithms can analyze large amounts of patient data, such as medical records, test results, and patient history as well as longitudinal data captured everyday with the aid of wearables. This data is used to identify patterns and trends that can be used to predict patient outcomes and to monitor patients' health over time.
One of the most important benefits of AI in patient predictive analysis and monitoring is its ability to identify patients who are at high risk for certain health conditions or complications. For example, AI algorithms can analyze patient data to identify individuals who are at high risk for heart disease or diabetes and can recommend interventions to prevent or manage these conditions.
In addition, AI can be used to monitor patients over time, providing continuous data and analysis that can help healthcare providers detect subtle changes in a patient's health before they become significant problems. For example, AI can analyze data from wearable devices to detect changes in the activity of EEG the heart rate and the movement of epileptic patients to detect seizure events and create accurate patient diaries that can provide support to clinical decision-making, insights that can help them make more informed decisions about patient care.
Furthermore, AI algorithms have been employed to analyze patient data to identify potential drug interactions or adverse reactions or can provide recommendations for personalized treatment plans based on individual patient characteristics. Finally, AI can be used to improve patient engagement and education, providing patients with personalized information and support that can help them better understand and manage their health. For example, AI can provide patients with personalized health recommendations based on their individual health profile or can provide patients with educational resources and support to help them manage chronic conditions or other health issues.
Brain-computer interfaces (BCI)
Brain-Computer Interface (BCI) technology is an innovative field that is rapidly advancing in healthcare. BCI enables direct communication between the brain and an external device, allowing for monitoring and control of brain activity. This technology has the potential to significantly improve the lives of patients with neurological conditions, such as epilepsy, Parkinson's disease, or paralysis caused by spinal cord injuries. BCI can be used to restore communication and movement in patients who have lost the ability to control their limbs due to neurological damage.
A very promising use case of BCI technology in healthcare is the development of Brain-Machine Interfaces for patients with paralysis caused by spinal cord injuries. Such interfaces allow patients to control prosthetic limbs using their thoughts. By using electrodes implanted in the patient's brain, the system can detect the patient's intention to move and translate it into movements of the prosthetic limb. This technology has the potential to restore mobility and independence in patients who have lost the ability to move their limbs due to spinal cord injuries. In fact, a team of researchers at the University of Pittsburgh successfully implanted a BCI system in a patient with quadriplegia, allowing him to control a robotic arm using his thoughts.
Very recently researchers were able to reconstruct high-quality images from brain recordings obtained via functional MRI (fMRI), even though brain signals are very complex underlying representations and data annotations are scarce. This research, to decode visual stimuli from brain recordings, aims to deepen our understanding of the human visual system and build a solid foundation for bridging human vision and computer vision. In simple words, the researchers managed to reproduce, with high accuracy, the images that a human was looking at with the use of AI.
A critical conclusion
The integration of AI into healthcare is revolutionizing the
industry. From improving the accuracy of diagnoses, to streamlining
administrative processes, the possibilities are endless. The examples
outlined in this blog post are just a few of the many ways in which AI
is being set to use in healthcare.
As it has been explained there are several potential benefits to using AI in healthcare, but we must keep in mind that there are also some risks that need to be considered, while implementing this revolutionary change of healthcare.
- AI systems can perpetuate existing biases and discrimination, especially if they are trained on biased or incomplete datasets. For example, an AI system that is trained on historical healthcare data can be biased against certain demographics and as such leads to unfair treatment or diagnosis of certain groups of people.
- AI systems are only as good as the data they are trained on and the algorithms they use. If an AI system is not properly designed or trained, it can produce inaccurate or unreliable results, which can lead to incorrect diagnoses or treatment recommendations.
- AI systems in healthcare often rely on sensitive personal data, such as medical records, which might be vulnerable to breaches or hacks and personal health information can be compromised.
- While AI systems can automate many aspects of healthcare, they cannot replace human judgement and expertise. Over-reliance on AI systems without adequate human oversight leads to errors or missed diagnoses. We must not use the existence of AI tools as an excuse to decrease quality of healthcare and resources.
The use of AI in healthcare is a relatively new field, and there are still many challenges that need to be addressed, such as how to ensure transparency and accountability in the development and deployment of AI. Nevertheless, the potential benefits of ethically implemented AI are too great to ignore. As AI technology continues to advance, we can only expect to see even more innovative medical applications of this powerful tool.
References & sources
Tschandl, P., Rinner, C., Apalla, Z., Argenziano, G., Codella, N., Halpern, A., ... & Kittler, H. (2020). Human–computer collaboration for skin cancer recognition. Nature Medicine, 26(8), 1229-1234.
Ehteshami Bejnordi, B., Veta, M., Johannes van Diest, P., van Ginneken, B., Karssemeijer, N., Litjens, G., & Hermsen, M. (2021). Development and validation of a deep learning system for histopathologic diagnosis of breast cancer. JAMA Oncology, 7(3), 331-339.
Ting, D. S., Cheung, C. Y., Lim, G., Tan, G. S., Quang, N. D., Gan, A., ... & Wong, T. Y. (2017). Development and validation of a deep learning system for diabetic retinopathy and related eye diseases using retinal images from multiethnic populations with diabetes. JAMA, 318(22), 2211-2223.Babylon Health. (n.d.). How it works.
Coughlin, S. S. (2019). Health chatbots for medical advice: Design, implementation, and evaluation. Journal of Medical Systems, 43(8), 239.
Öztürk H., Özgür A., Ozkirimli E. (2018) , DeepDTA: deep drug–target binding affinity prediction, Bioinformatics, Volume 34, Issue 17, 01 September 2018, Pages i821–i829,
Verhaeghe J., Dhaese S., De Corte T. ... & Van Hoecke S. (2022) Development
and evaluation of uncertainty quantifying machine learning models to
predict piperacillin plasma concentrations in critically ill patients. BMC Med Inform Decis Mak. 25;22(1)
Parisi D., Melissa A. , Sveshnikova A., Bolz S., Moreau Y., Schroeder M. (2020) , Drug
repositioning or target repositioning: A structural perspective of
drug-target-indication relationship for available repurposed drugs, Computational and Structural Biotechnology Journal, Volume 18
Chatzichristos, C., Swinnen, L., Macea, J., Bhagubai, M., Van Paesschen, W., De Vos, M. (2022). Multimodal detection of typical absence seizures in home environment with wearable electrodes. Frontiers in Signal Processing, 2, Art.No. 1014700. doi: 10.3389/Frsip.2022.1014700
Wodlinger, B., Downey, J. E., Tyler-Kabara, E. C., Schwartz, A. B., Boninger, M. L., & Collinger, J. L. (2014). Ten-dimensional anthropomorphic arm control in a human brain–machine interface: difficulties, solutions, and limitations. Journal of neural engineering, 12(1), 016011.
Chen, Z., Qing, J. , Xiang T. , Yue W. and Zhou, J. H. (2022). Seeing Beyond the Brain: Conditional Diffusion Model with Sparse Masked Modeling for Vision Decoding.
Christos Chatzichristos
I am Christos Chatzichristos, currently a post-doctoral researcher at KU Leuven. My educational background revolves around electrical and computer engineering, with a specialization in Biomedical Applications and an emphasis on signal processing during both my Master's and Ph.D. studies. During my doctoral journey, I witnessed the profound impact of neural networks on the field of signal processing, marking the inception of my foray into the realm of AI applications. I hold a strong belief in fostering broad interdisciplinary collaborations, as I believe that research today cannot thrive in isolation within a single domain. Artificial intelligence stands as a potent tool to expedite healthcare research. However, to truly harness its potential, we must bridge the gap by facilitating healthcare professionals' understanding of fundamental AI concepts, just as they aid biomedical engineers in unraveling the mysteries of the human body. So, here's to using AI to accelerate healthcare research while ensuring that we all speak the same language – whether it's the language of algorithms or the language of anatomy!
Related courses
