Artificial Intelligence and Machine Learning in Medical Imaging

Since the start of this year, artificial intelligence (AI) and machine learning (ML) have been widely used to aid in the technological processes required by many fields. AI is a tool that has the ability to simulate human thinking and behavior, while ML is a portion of AI in which machines learn from data to organize classifications or make predictions. ML can sometimes be managed by humans, but this is not a necessary factor. Using a training dataset in their development, AI and ML algorithms become refined using a validation dataset, then tested by a different institution for their performance in an independent test dataset.

AI has shown some of the most accurate responses in clinical applications, such as medical imaging. The medical image as imaging data, also known as radiomics, is basically the computerized analysis of medical imaging. Radiomics contains subunits (pixel/voxel), which can be used in mathematical features that might be related to disease behavior or outcomes. 

To produce a medical image, a lengthy process must occur: acquisition, reconstruction, interpretation of results, and digital space reporting in which the image can be communicated. In this procedure, a lot of data can be collected by AI and ML. For example, while analyzing a brain MRI utilizing machine learning, AI can identify and localize tissue changes that reflect an early ischaemic stroke, which may sometimes be much more difficult to recognize by a human reader. 

Aortic stenosis management also may be aided by the capabilities of AI and ML. AI applications of echocardiography, computed tomography, and MRI have the ability to provide granular assessment of annular conformation, leaflet mobility, and outflow tract to identify patients with less severe stenosis in which surgical intervention might be more advantageous than medical management. Identifying changes in left ventricular function and fibrosis or remodeling could play a crucial role in prompting earlier intervention. The enhanced reading performance of AI could be harnessed to improve patient selection for intervention by identifying structural or dynamic changes that correlate with worse outcomes. 

Accuracy is extremely important in the AI classification of aortic stenosis severity. This is to ensure that patients with severe diseases are correctly captured and those with mild diseases are not erroneously reclassified into a high-risk group. 

Another example of AI and ML in medical imaging is diagnosing autoimmune myocarditis, which is a very complicated and rare case of immunotherapy by using cardiac imaging at an earlier time to prevent morbidity. 

Artificial intelligence can also be used for cancer detection. It’s a high-yield niche where early exploration of AI and ML would most likely be used by radiologists. In cancer imaging, images captured from the patient are preprocessed and transformed as input to develop machine-learning algorithms and models. AI and ML are useful in ensuring that the images have similar image section thickness and pixel dimensions, mapping the input image data, and learning simple or complex mathematical functions related to a goal or output (such as a clinical or scientific observation). Machine learning algorithms can be built or trained with or without “real variables”, which are reference results validated by domain experts. 

The main goal for AI and ML is to use algorithms in order to identify patterns within medical images, especially those that are really hard to perceive with a human eye, and therefore make predictions that can be useful to the clinical decision-making procedure. AI and ML have both the massive ability to improve work efficiency, reduce errors, and enhance diagnostic performance if used wisely.

References

Pesapane, Filippo, et al. “Artificial Intelligence in Medical Imaging: Threat or Opportunity? Radiologists Again at the Forefront of Innovation in Medicine.” European Radiology Experimental, vol. 2, no. 1, Dec. 2018, p. 35. DOI.org (Crossref), https://doi.org/10.1186/s41747-018-0061-6.

Trivizakis, Eleftherios, et al. “Artificial Intelligence Radiogenomics for Advancing Precision and Effectiveness in Oncologic Care (Review).” International Journal of Oncology, vol. 57, no. 1, May 2020, pp. 43–53. DOI.org (Crossref), https://doi.org/10.3892/ijo.2020.5063.

Van Griethuysen, Joost J. M., et al. “Computational Radiomics System to Decode the Radiographic Phenotype.” Cancer Research, vol. 77, no. 21, Nov. 2017, pp. e104–07. DOI.org (Crossref), https://doi.org/10.1158/0008-5472.CAN-17-0339.

Zhang, Lifei, et al. “ibex : An Open Infrastructure Software Platform to Facilitate Collaborative Work in Radiomics.” Medical Physics, vol. 42, no. 3, Mar. 2015, pp. 1341–53. DOI.org (Crossref), https://doi.org/10.1118/1.4908210.