
The featured image was bought from Istock.com. Copyrights.
The latest developments in artificial intelligence (AI), including Machine Learning (ML) and Deep Learning (DL), have led to the development of several tools designed to help physicians in their daily tasks.
Machine Learning and Deep Learning
Machine learning, a subcategory of artificial intelligence, is a system that uses algorithms to analyze and evaluate data in order to offer hypotheses and predict results. The machine is fed large amounts of data and algorithms that teach the system to perform a task.
Deep learning is a type of machine learning based on a system of probabilities. In addition, DL can be represented as a mathematical system, called artificial neural networks, inspired by the human brain. O6ver the years, researchers have improved the concept through various techniques, and today’s computers have higher computing power. Using large amounts of data, the system can provide statements, decisions or predictions with a certain degree of accuracy.
Decision-Support and Diagnostic Tools
Image segmentation
DL is being used for medical imaging analysis, especially for segmentation. Most medical treatments or diagnoses require the identification of areas of interest on the images. These actions are largely performed manually by a radiologist who outlines the affected area on the image. They take a long time to complete and are expensive. Recently, DL has proven to be a reliable technique in the field of image processing, mainly due to convolutional neural networks—artificial neural network arrangements inspired by the visual cortex in animals. DL has the advantage of automating this step, making disease detection and diagnosis a faster process for the patient.

Figure 1 Knee segmentation—sagittal section
Automatic annotation
Significant progress in computer vision and automatic annotation has been made in recent years. Thanks to complex algorithms, the program can annotate medical imagery automatically, as in the following example showing an X-ray of a thorax where the image is annotated and described with accuracy: disease or lesion detection, location, and severity, as well as identification of other affected areas or organs.

Figure 2 Example of annotations—Chest X-ray
Computer-assisted detection and diagnosis
Different types of computer-assisted detection are integrated in the PACS system, a computer program used by doctors to view medical images. Radiologists can view up to 100 images per day, using computer-aided detection for faster and more efficient analysis. Recent studies have shown good performance when computer-aided detection is used for breast cancer—which affects approximately 25,000 women in Canada [2]—lung cancer, and Alzheimer’s disease [1].

Figure 3 Example of detection and annotation—Breast cancer
Integrating big data with precision medicine
Precision medicine is the study of the patient’s environment, lifestyle and genome in order to prevent illness or provide effective treatments adapted to each individual. This is made possible with mobile devices, various health-related mobile applications, smart watches, or companies like 23andme and BiogeniQ. By combining all the data with modern medicine, DL can analyze and create an accurate and personalized health profile for each patient.
Limitations and Considerations in Applying AI to Radiology
In conclusion, despite a number of promising studies, several problems exist and must be resolved before artificial intelligence can replace radiologists.
DL is dependent on the quality as well as the availability of large amounts of data. High standards and protocols must be enforced in hospitals and clinics to ensure good image quality. There are also the ethical and legal considerations that do not always facilitate the use of clinical data to create commercial products. There are still questions that need answering: How does AI work? Can it detect rare diseases efficiently? Ultimately, systems that are being developed are always validated by radiologists to avoid the risk of errors and the risk of misinforming the patient.
Additional Information
For more information on this area of research at ÉTS, please refer to the following articles:

Marie-Anne Valiquette
Marie-Anne Valiquette obtained a Bachelor's degree in Mechanical Engineering at the École de technologie supérieure (ÉTS) in Montreal. She lives in Silicon Valley, California where she studies artificial intelligence through online platforms like Udacity and deeplearning.ai.
Program : Mechanical Engineering
Field(s) of expertise :
