91黑料

  1. Home
  2. Insights
  3. Blog
  4. The rise and role of AI in medical imaging

The rise and role of AI in medical imaging

Artificial intelligence (AI) is rapidly becoming integral to healthcare and clinical research as it offers efficient solutions for the massive amounts of data generated, processed and utilised in the industry. AI can be particularly useful in medical imaging solutions, enhancing processes such as screening, patient eligibility, and disease detection and response monitoring. When integrated appropriately into medical imaging workflows, leading-edge technology like AI and automation can help mitigate some of the industry’s persistent challenges by alleviating burden on understaffed sites and improving efficiency and patient care. 

The significance of AI integration in medical imaging

AI systems can offer opportunities to support human expertise and drive efficiencies that benefit clinical trials.1, 2, 3 The potential benefits extend to participating clinical sites, pharmaceutical industry stakeholders and, ultimately, the patients. With strategic integration at key points, AI systems can drive quality and cost-effectiveness by mitigating the potential for human error, accelerating processes for faster results and potentially increasing the imaging biomarkers available to assess response. Sponsors can then make timelier and better informed decisions to support clinical milestones by getting high-quality data in hand faster thanks to AI-enhanced optimisations.

AI for streamlining image acquisition and quality control

AI integration during image acquisition workflows can bring valuable benefits to the irreplaceable human expertise required in medical imaging. At this stage, AI is already being integrated in several ways to help save time by reducing scan times and decrease ionising radiation exposure. Importantly, it is also contributing to the increasing availability of imaging tools and making them more accessible to further enhance the process and facilitate more efficient workflows. 

When uploading images from clinical sites, there is potential for human error. Using AI systems to assist in data transfer quality control can reduce the QC burden on personnel and reduce the potential for related delays. AI enables faster validation of patient information and image/scanning sequence type and can verify the data received against the imaging protocol. It can identify incomplete or missing images as well as prevent the transfer of protected health information (PHI) that may have mistakenly been included in the image data. The system will automatically flag errors for the appropriate personnel to review and remedy in a timely manner. Automating this process removes the burden from personnel and reduces the overall time commitment to the process to avoid delays.

Efficiency in AI-enhanced image assessment

Radiologists are responsible for the critical image assessment. Their human expertise is required to evaluate the safety and efficacy of new treatments, though the advancements in AI2, 3, 4 now offer radiologists a helping hand to save time during interpretation and post-processing. AI can also provide more precise biomarkers that are relevant to the clinical endpoints across indications, ranging from oncology, cardiovascular, metabolic, MASH, musculoskeletal and more.

The partial automation of these assessments can reduce reader fatigue and limit the potential variability between readers by standardising assessments. This is especially important considering that most reading processes include multiples reviewers plus adjudication to achieve consensus on the reads.

Depending on the specific criteria and anatomy of the assessments, the pool of available specialist readers can vary, and AI-assisted image assessment can benefit the timelines for both readers and the overall process. When put into context of the number of overall images that are generated per patient per visit over the course of a study, even small optimisations can compound to expedite the assessment process while ensuring data quality and integrity. For example, a liver oncology study may receive hundreds of images of a patient’s liver over multiple visits and automated deep learning-based segmentation models can be up to 20 times faster than manual annotation in such studies. In turn, this leads to faster assessments and data analysis that can significantly reduce study timelines.

Broader applications of AI in medical imaging

There has been significant progress in the last decade in the use of AI across therapeutic areas. For example, AI in medical imaging is currently being successfully leveraged in oncology, cardiometabolic and musculoskeletal areas. There has been notable progress in AI integration in oncology for automation of the segmentation of different tumours and lesions specifically using U-Net architecture, a deep-learning semantic segmentation technique. Once lesions are segmented, tumours’ bi-dimensional measurements can now be easily automated. 

Similar automation opportunities in liver tissue segmentation and the quantification of disease markers in conditions like metabolic dysfunction-associated steatohepatitis (MASH) could help quantify markers of various liver disease including liver fat percentage and liver stiffness. 

Cardiometabolic applications of AI include estimation of visceral and subcutaneous fat assessment for obesity trials, which is a growing area amid the breakthroughs with GLP-1 drugs. Automation is also currently improving musculoskeletal evaluations, including scoliosis assessment and fracture risk detection in addition to automated bone age assessments and organ volume assessments.

Future consideration for AI in clinical trials

AI, including automation and machine learning, is quickly being adopted within clinical trials and medical imaging. Some of the AI tools available are approved by regulators and have well validated and published results, while others are at various stages of incomplete approval and more are consistently in development. This proliferation of AI tools is driving many benefits to clinical research, though it should always be incorporated judiciously. 

Successful integration of AI depends on a clear understanding of the benefits of AI, thorough validation to mitigate the inherent biases within it, and clarity on how it fits into existing or improved workflows to support clinical trials in the most efficient ways. Our core team of more than 450 qualified clinical imaging professionals, including radiologists, imaging scientists and certified imaging technologists, offer therapeutic expertise and can consult on the most effective integration approach to optimise time, cost-effectiveness and data quality. By realising the benefits of AI in medical imaging, we can facilitate faster decision-making backed by higher-quality data. Ultimately, we can enable the availability of better biomarkers and endpoints for clinical trials to bring better outcomes to patients more effectively. 

Contact us to discover how to harness the potential of AI in medical imaging and improve your clinical trial outcomes. 

References

1 John C. Gore. Preface Document. Magnetic Resonance Imaging 68 (2020) A1–A4.

2 Currie, et al. ‘Machine Learning and Deep Learning in Medical Imaging: Intelligent Imaging’ Journal of Medical Imaging and Radiation Sciences 50 (2019) 477-487

3 Fazal, et al. 2018. ‘The past, present and future role of arti?cial intelligence in imaging’. European Journal of Radiology 105 (2018) 246–250

4 Jiang, et al. ‘Development and application of artificial intelligence in cardiac imaging’. Br J Radiol. 2020 Sep 1;93(1113):20190812

In this section

Connect with us