Understanding Machine Learning and Artificial Intelligence and Their Growth in Medicine

Why are Machine Learning and Artificial Intelligence important? What has been their impact and where is the technology headed? Ranging from computational science, business, and computer science these fields have profoundly changed several academic disciplines and approaches to problem solving. Now, through imaging technology and disease prevention, AI and machine learning applications are continuing to increase within medicine (Deo, 2015). While the potential for these analysis mechanisms can help solve the medical challenges of today and bring about incredible positive change in the field of medicine, it is vital that such technology is used within the bounds of medical ethics.

From University of Wisconsin Madison, a map of a general neural network model.

Artificial intelligence is a rather broad field encompassing several different types of methods including machine learning, deep learning, and neural networks (Park et. al, 2019; Hao, 2018). Essentially, they are complex, evolving models built from a combination of mathematics, algorithms, and statistics (Deo, 2015; Hao, 2018). In other words, these algorithms interpret data that may have hidden correlations, numerical relationships, and motifs. They achieve these findings by first training on datasets and then evaluating on an experimental dataset. In the training phase, these models work through large datasets and create their own analysis. With more time and practice with the analysis that these algorithms complete, the greater the levels of accuracy and performance they reach. In this way, the machines “learn” by examples and self-correction.

Once the algorithms have finished their training, they can move onto their second phase, or be used to evaluate experimental datasets. In particular, the most prevalently used types of machine learning are supervised and unsupervised learning (Deo, 2015; Hao 2018). Supervised learning consists of having the computer identify or “classify” items (Deo, 2015).  In his article “Machine Learning in Medicine”, Physician Scientist Rahul Deo describes supervised learning as “classification, which involves choosing among subgroups to best describe a new data instance, and prediction” (Deo, 2015). Deo demonstrates examples in cardiology and radiology where analyzing certain tests such as electrocardiograms and X-rays are areas that machine learning techniques could be applied/are applied to (Deo, 2015). Unsupervised learning, on the other hand, is more generalized, in which the machine attempts to extrapolate subtle but distinct “patterns” within the dataset (Deo, 2019; Hao 2018).

From Rahul Deo’s paper Machine Learning in Medicine, the following image shows the applications of both types of machine learning techniques to understanding Myocardial Infarction.

Researchers across the globe are finding new applications of machine learning to improve therapies and curing diseases. One research group has focused on improving skin disease diagnosis using deep learning models (Liu, Bui, 2019). The “Deep Learning System (DLS)” that the research team developed  “achieve[d] an accuracy across 26 skin conditions on par with U.S. board-certified dermatologists” and may “augment the ability of general practitioners who did not have additional specialty training to accurately diagnose skin conditions”  (Liu, Bui,  2019). Another research group is using deep learning to solve diabetic retinopathy, a disease that can cause permanent loss of vision (Peng, Gulshan, 2016). With the goal of implementing their research towards aiding underserved populations and other disease treatment applications, their findings have shown that using 2D images is able to achieve diagnostic results comparable to eye-specialists with an accuracy of 0.95 relative to 0.91, respectively (metric used is known as F-score) (Peng, Gulshan, 2016). The research group still hopes that the technology achieves new heights by incorporating analysis of 3D images, as this is necessary in further improving diagnostics of dabetic retinopathy (Peng, Gulshan, 2016).

Nevertheless, machine learning and artificial intelligence also pose vital ethical considerations. In their article “Machine Learning in Medicine: Addressing Ethical Challenges”, authors Effy Vayena, Alessandro Blasimme, and I. Glen Cohenn emphasize how some algorithm’s analysis/reasoning cannot be traced, interfering with the necessary detailed knowledge patients deserve from doctors about medical procedures (Vayena, Blasimme, and Cohen, 2018).  Furthermore, these authors evaluate and elaborate on how, based on the nature of AI algorithms, human interaction and decision making could be reduced (Vayena, Blasimme, and Cohen, 2018). Thus, while technology can help medical diagnosis, it also can infringe upon the privacy and security required in taking care of medical data. To ensure ethical standards are maintained, algorithms must be closely monitored, and their use should be limited to specific, regulated, and verified cases. In the future, we need to ensure that technology does not compromise the foundational moral and ethical values of medicine and the healthcare industry, but also continue finding ways for it to create positive changes in medical treatment globally.

Edited by Thalia Le
Posted by Rachel Xue

References

Chen, P.-H. C., Liu, Y., & Peng, L. (2019, April 18). How to develop machine learning models for healthcare. Retrieved October 15, 2019, from https://www.nature.com/articles/s41563-019-0345-0.

Deo, R. C. (2015, November 17). Machine Learning in Medicine. Retrieved October 15, 2019, from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5831252/.

Ekins, S., Puhl, A. C., Zorn, K. M., Lane, T. R., Russo, D. P., Klein, J. J., … Clark, A. M. (2019, April 18). Exploiting machine learning for end-to-end drug discovery and development. Retrieved October 15, 2019, from https://www.nature.com/articles/s41563-019-0338-z.

Gulshan, V., Peng, L., Coram, M., Stumpe C. M., Wu, D., Narayanaswamy, A., … Webster, R. D. (2016, December 13). Accuracy of a Deep Learning Algorithm for Detection of Diabetic Retinopathy. Retrieved October 15, 2019, from https://jamanetwork.com/journals/jama/fullarticle/2588763.

Hao, K. (2019, September 2). What is machine learning? Retrieved October 15, 2019, from https://www.technologyreview.com/s/612437/what-is-machine-learning-we-drew-you-another-flowchart/.

Hardesty, L., & MIT News Office. (2016, October 27). Making computers explain themselves. Retrieved October 15, 2019, from http://news.mit.edu/2016/making-computers-explain-themselves-machine-learning-1028.

Liu, Y., & Bui, P. (2019, September 12).  Using Deep Learning to Inform Differential Diagnoses of Skin Diseases. Retrieved October 15, 2019, from https://ai.googleblog.com/2019/09/using-deep-learning-to-inform.html.

Liu, Y., Jain, A., Eng, C., Way, H.D., Lee, K., Bui, P., … Coz, D. (2019, September). PDF. Palo Alto.https://arxiv.org/ftp/arxiv/papers/1909/1909.05382.pdf

Lu, L. (2018, October 28). PDF. http://www.cs.jhu.edu/~lelu/publication/SNMMI2018_LeLu_02.pdf

Liu, Y., Gadepalli, K., Norouzi, M., Dahl, G. E., Kohlberger, T., Boyko, A., … Stumpe, M. C. (2017, March 8). Detecting Cancer Metastases on Gigapixel Pathology Images. Retrieved October 15, 2019, from https://arxiv.org/pdf/1703.02442.pdf.

Park, S. H., Do, K.-H. H., Kim, S. undefined, Park, J. undefined, & Lim, Y.-S. undefined. (2019, July 3). What should medical students know about artificial intelligence in medicine? Retrieved October 15, 2019, from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6639123/.

Patel, J. L., & Goyal, R. K. (2007, September 2). Applications of artificial neural networks in medical science. Retrieved October 15, 2019, from https://www.ncbi.nlm.nih.gov/pubmed/18690868.

Peng, L., & Gulshan, V. (2016, November 29). Deep Learning for Detection of Diabetic Eye Disease.  Retrieved October 15, 2019, from https://ai.googleblog.com/2016/11/deep-learning-for-detection-of-diabetic.html.

The whisper of schizophrenia: Machine learning finds ‘sound’ words predict psychosis. (2019, June 14). Retrieved October 15, 2019, from http://news.emory.edu/stories/2019/06/esc_schizophrenia_machine_learning/campus.html.

Tran, H., Chen, K., Lim, A. C., Jabbour, J., & Shumack, S. (2005, November). Assessing diagnostic skill in dermatology: a comparison between general practitioners and dermatologists. Retrieved October 15, 2019, from https://www.ncbi.nlm.nih.gov/pubmed/16197420.

Vayena, E., Blasimme, A., & Cohen, I. G. (2018, November 6). Machine learning in medicine: Addressing ethical challenges. Retrieved October 15, 2019, from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6219763/.

Image References

A Basic Introduction To Neural Networks. (n.d.). Retrieved March 8, 2020, from http://pages.cs.wisc.edu/~bolo/shipyard/neural/local.html

Deo, R. C. (2015, November 17). Machine Learning in Medicine. Retrieved October 15, 2019, from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5831252/.

Huang, L. (2019) “Artificial Neural Network with Chip” [Online image]. Flickr. https://www.flickr.com/photos/186021024@N08/49203125457

Leave a Reply

Your email address will not be published. Required fields are marked *