Powered by
05:51 CET
EU 7 - Artificial intelligence and radiation protection
EuroSafe Imaging Artificial Intelligence and Machine Learning Physics in Medical Imaging
Sunday, March 3, 08:30 - 10:00
Room: N
Type of session: EuroSafe Imaging Session
Topic: EuroSafe Imaging, Artificial Intelligence and Machine Learning, Physics in Medical Imaging
Moderators: G. Frija (Paris/FR), C. Hoeschen (Magdeburg/DE)

A-1011
08:30
Chairpersons' introduction (part 1)
G. Frija; Paris/FR
Learning Objectives

1. To understand the mechanisms of how artificial intelligence (AI) can help to reduce necessary doses for imaging procedures relying on ionising radiation.
2. To appreciate which AI methods in medical imaging might be helpful for reducing dose for especially sensitive body regions or highly susceptible patients in general.
3. To learn how results from AI evaluation of radiographic images might help to influence decisions on the most appropriate methods or radiation protection measures on an individual patient basis in an early phase of patient diagnosis and therapy.

Abstract

Artificial intelligence (AI) is a vehicle for improving patient outcomes, and thus it will change the practice of radiology in the next few years. AI will be a method of demonstrating the clinical value of radiology and will allow for faster and more accurate image assessment and hence diagnosis. AI will also support the training of radiologists, improve clinical knowledge, and contribute to research in medical imaging. Furthermore, AI could have knock-on impacts on radiation protection by reducing the incidence of unnecessary procedures, reducing the doses administered, increasing the image quality, and improving patient safety in general. The transformative potential of AI for the radiology profession cannot be ignored: rather practitioners must be prepared to embrace it. This session will provide an overview of how AI can enhance radiation protection in medical imaging.

A-1012
08:33
Chairpersons' introduction (part 2)
C. Hoeschen; Magdeburg/DE
Learning Objectives

1. To understand the mechanisms of how artificial intelligence (AI) can help to reduce necessary doses for imaging procedures relying on ionising radiation.
2. To appreciate which AI methods in medical imaging might be helpful for reducing dose for especially sensitive body regions or highly susceptible patients in general.
3. To learn how results from AI evaluation of radiographic images might help to influence decisions on the most appropriate methods or radiation protection measures on an individual patient basis in an early phase of patient diagnosis and therapy.

Abstract

Using artificial intelligence in medical imaging for radiation protection purposes has three main pillars. The first pillar is that artificial intelligence can be used for investigating parameters of medical imaging procedures based on ionising radiation like image quality evaluation, exposure description and prediction of optimal imaging parameters in such sense depending on patient characteristics and diagnostic task. The second pillar is that artificial intelligence can be used to optimise image quality per dose based on advanced methods for image reconstruction, scatter reduction and thus noise and artefact reduction. The third pillar is the characterisation of patient radiation sensitivity and susceptibility for choosing the right radiation therapy or the optimal imaging procedure on individual patient base. Such radiation protection related aspects will be described in the session, and the difference will be explained in the introduction by the chairs.

A-1013
08:35
Artificial intelligence: a tool for quality and safety improvement in radiation protection
G. Frija; Paris/FR
Learning Objectives

1. To learn about current trends of AI in imaging.
2. To understand how these trends could improve quality and safety.
3. To appreciate how AI could affect radiation protection.

Abstract

The use of artificial intelligence (AI) in medicine has attracted enormous attention, and there is a lot of research going on to implement AI in imaging. Current applications include, for example, the detection of breast cancer on mammography images, the detection of lung nodules in CT scans, and detecting pneumonia in chest x-rays. In addition, applications beyond image interpretation are on the rise. The development of AI tools in medical imaging will certainly also have an input on medical radiation protection. However, currently, there is hardly any research work on the use of AI as a tool for quality and safety improvement in imaging, although there are a number of potential applications that could help reduce doses for imaging procedures relying on ionising radiation, especially for sensitive body regions or highly susceptible patients in general. Algorithms can be used to improve image quality and dose optimisation in CT. AI-based organ recognition combined with dose estimation algorithms can provide patient-specific organ doses. This talk will present current trends of AI in radiology and provide examples how these AI systems could contribute to strengthening medical radiation protection for the benefit of patients.

A-1014
08:55
Artificial intelligence for scatter reduction and optimising imaging procedures
C. Hoeschen; Magdeburg/DE
Learning Objectives

1. To learn about actual scatter reduction techniques in imaging based on ionising radiation, their advantages and drawbacks as well as the potential of optimising imaging trajectories.
2. To understand how AI-based procedures can provide a fast solution for scatter estimation and reduction as well as for image trajectory determination.
3. To appreciate how AI can help to reduce patient dose in medical imaging with easy to implement applications.

Abstract

Medical imaging based on ionising radiation is strongly suffering from two major drawbacks which result in the need for exposure levels mainly to patients but also to staff in interventional procedures higher than it might be necessary from the physical imaging process. This is the case since noise and artefacts are two of the most relevant factors deteriorating sufficient image quality for diagnostic or interventional purposes. Noise and artefacts both are strongly related to scatter that is generated in the patient body during the interaction of the X-rays with the material. Therefore ways will be discussed how the scatter contribution of the X-rays caused by patients can be estimated and thus be reduced by methods based on artificial intelligence as they have been shown by various groups and its advantages and disadvantages will be highlighted. By such scatter reduction better image quality can be gained and thus exposure levels can be reduced. In addition, this talk will briefly introduce another concept to use x-rays as efficient as possible based on optimising the imaging geometry based on artificial intelligence approaches.

A-1015
09:15
Artificial intelligence for intelligent reconstruction methods for radiation protection measures
C. T. Whitlow; Winston-Salem/US
Learning Objectives

1. To discuss radiation exposure due to computed tomography associated with common diagnostic tests and in the setting of screening exams.
2. To describe efforts aimed at reducing radiation dose and improving image quality via conventional approaches (e.g. iterative reconstruction).
3. To introduce methods for improving image quality via novel artificial intelligence (AI)-based approaches, with quantitative results characterising noise reduction and example cases from clinical application.

A-1016
09:40
Using artificial intelligence for optimising procedures reflecting radiosusceptibility of patients
C. Hoeschen; Magdeburg/DE
Learning Objectives

1. To learn about radiosusceptibility and why it is important to look for it in images.
2. To understand what role AI could play in the future for optimising procedures, especially interventional procedures, from the point of view of radiation protection.
3. To appreciate the potential for new research topics bridging the gap between AI and radiation protection with respect to the individual patient.

Abstract

Radiation protection of patients is an important task in general. It has been found out that it might be even more relevant to certain patients than to others since, obviously, not all patients are prone to negative effects of ionizing radiation in the same way. Such differences are described as individual radiation sensitivity if one refers to short-term effects, especially so-called tissue reactions and as individual radiation susceptibility if one refers to long-term effects like especially cancer induction. Obviously, it is even more important to protect especially the healthy tissue of those patients who are more radiation sensitive or susceptible than those who are not. Some of the potential reasons for individual radiation sensitivity or susceptibility will be described as well as indications that and how specific analysis of imaging data could provide insights into such individual patient configuration and why this might be helpful for radiation protection. Methods based on artificial intelligence would be most promising for this task. There are also medical imaging applications with pretty high doses and especially pretty localized high doses like in interventional procedures. Thus, it would be feasible to optimize not only radiation therapy for individuals that are very sensitive but also for example interventional procedures. For planned CT examinations, it might be better to think about low-dose procedures or even of a replacement of the procedure by means of MRI if feasible in patients with extremely high risk. However, there are some ethical aspects of such approaches.

09:50
This website uses cookies. Learn more