The past decade of increasing interest and progress in Artificial Intelligence (AI) based solutions for medical imaging has set the stage for a number of trends that are likely to appear or intensify in the near future.
According to a new eBook, The Complete Guide to Artificial Intelligence in Radiology by Bayer in Radiology, AI has shown promise in positively impacting virtually every facet of a radiology department’s work - from scheduling and protocolling patient scans to interpreting images and reaching diagnoses. AI-based solutions have the potential to provide solutions that may help address some of the challenges facing radiology today, including staffing shortages, burnout, huge volumes of data, and diagnostic errors.
“AI has the potential to transform healthcare, and, particularly in medical imaging, it can turn the growing amounts of data into value-adding insights to support radiologists and their teams in their decision-making,” says Dr. Ryan Lee, Chair, Department of Radiology, Einstein Healthcare Network, Philadelphia. “It is crucial to drive innovation in this area and broaden access to digital tools that can help address the rising demand for solutions which improve the speed and accuracy of diagnoses.”
Looking ahead, here are some of the trending use cases.
Image quality improvement and monitoring
Many AI-based solutions that work in the background of radiology workflows to improve image quality have recently been established. These include solutions for monitoring image quality, reducing image artefacts, improving spatial resolution, and speeding up scans. Such solutions are entering the radiology mainstream, particularly for computed tomography, which for decades used established but artefact-prone methods for reconstructing interpretable images from the raw sensor data1.
These are gradually being replaced by deep-learning based reconstruction methods, which improve image quality while maintaining low radiation doses2. This reconstruction is performed on supercomputers on the CT scanner itself or on the cloud. The balance between radiation dose and image quality can be adjusted on a protocol-specific basis to tailor scans to individual patients and clinical scenarios3. Such approaches have found particular use when scanning children, pregnant women, and obese patients as well as CT scans of the urinary tract and heart4.
Scan reading prioritization
With staff shortages and increasing scan numbers, radiologists face long reading lists. To optimize efficiency and patient care, AI-based solutions have been suggested as a way to prioritize which scans radiologists read and report first, usually by screening acquired images for findings that require urgent intervention5. This has been most extensively studied in neuroradiology, where moving CT scans that were found to have intracranial hemorrhage by an AI-based tool to the top of the reading list reduced the time it took radiologists to view the scans by several minutes6.
Another study found that the time-to diagnosis (which includes the time from image acquisition to viewing by the radiologist and the time to read and report the scans) was reduced from 512 to 19 minutes in an outpatient setting when such a worklist prioritization was used7. A simulation study using AI-based worklist prioritization based on identifying urgent findings on chest radiographs (such as pneumothorax, pleural effusions, and foreign bodies) also found a substantial reduction in the time it took to view and report the scans compared to standard workflow prioritization8.
Image interpretation
The majority of commercially available AI-based solutions in medical imaging focus on some aspect of analyzing and interpreting images9. This includes segmenting parts of the image (for surgical or radiation therapy targeting, for example), bringing suspicious areas to radiologists’ attention, extracting imaging biomarkers (radiomics), comparing images across time, and reaching specific imaging diagnoses.
Looking to the future, AI tools will likely evolve to handle more varied data, become integrated into consolidated workflows, become more transparent, and ultimately more useful for increasing efficiency and improving patient care.
Learn more by downloading The Complete Guide to Artificial Intelligence in Radiology.
References:
1. Deák, Z., Grimm, J. M., Treitl, M., Geyer, L. L., Linsenmaier, U., Körner, M., Reiser, M. F., & Wirt (v0.1) - ASIR and MBIR ability to reduce artefacts (p.6) Artifacts.—ASIR and MBIR showed potential to reduce CT artifacts, such as beam hardening and photon starvation, in critical anatomic locat Deák, Z., Grimm, J. M., Treitl, M., Geyer, L. L., Linsenmaier, U., Körner, M., Reiser, M. F., & Wirt (v0.1) - Spatial resolution is significantly better with ASIR than FBP (p.5) Subjective image quality.—Compared with ASIR, more image noise was found on FBP reformations (median scores for all planes, 21) without impa Singh, S., Kalra, M. K., Hsieh, J., Licato, P. E., Do, S., Pien, H. H., & Blake, M. A. (2010). Abdom (v0.1) - Image quality improvement with iterative reconstruction (p.9) These phantom studies have shown substantial improvement in image quality with iterative reconstruction compared with FBP for CT image r Singh, S., Kalra, M. K., Hsieh, J., Licato, P. E., Do, S., Pien, H. H., & Blake, M. A. (2010). Abdom (v0.1) - Benefits of ASIR over FBP (p.2) This algorithm takes into account precise modeling of the x-ray photon statistics and electronic noise, all of which are less accurate in FB
2. Akagi et al., 2019; H. Chen et al., 2017; Choe et al., 2019; Shan et al., 2019 > Akagi, M., Nakamura, Y., Higaki, T., Narita, K., Honda, Y., Zhou, J., Yu, Z., Akino, N., & Awai, K. (v0.1) - DLR has comparable quality to MBIR whilst being faster (p.2) As DCNN kernel is trained with ideal MBIR images, we expect to see that not only the DLR approach could generate comparable image quality Chen, H., Zhang, Y., Kalra, M. K., Lin, F., Chen, Y., Liao, P., Zhou, J., & Wang, G. (2017). Low-Dos (v0.1) - Algorithms designed to improve image quality (p.1) To address this inherent physical problem, many algorithms were designed to improve the image quality for low-dose CT(LDCT). These algorithm Choe. Image Conversion of CT Kernels Improves Radiomics Reproducibility. Radiology. 2019 (v0.1) - CNN is effective in improving radiomic feature reproducibility (p.5) Our study demonstrates that different reconstruction kernels result in marked reduction of the reproducibility of radiomic features (15.2% [ - 9 (cont) - Shan, H., Padole, A., Homayounieh, F., Kruger, U., Khera, R. D., Nitiwarangkul, C., Kalra, M. K., & (v0.1) - MAP-NN can produce images from raw data at or greater than commercial IR quality images (p.5) Our MAP-NN systematically demonstrates that the DL approach can provide a similar or better image quality in terms of structural fidelity an
3. McLeavy, C. M., Chunara, M. H., Gravell, R. J., Rauf, A., Cushnie, A., Staley Talbot, C., & Hawkins, (v0.1) - DLR triad can be adjusted for tailored protocols (p.2) Given that DLR offers the triad of low-dose, highquality, and fast reconstruction times, we have been able to adjust parameters to provide Willemink, M. J., & Noël, P. B. (2019). The evolution of image reconstruction for CT-from filtered b (v0.1) - Dosing can be reduced to fit the need of the protocol, adjusting parameters will produce different image qualities (p.2) Multiple dose- reduction methods were introduced, including tube current modulation[9], organ-specific care[10], beam-shaping filters[11], a
4. McLeavy, C. M., Chunara, M. H., Gravell, R. J., Rauf, A., Cushnie, A., Staley Talbot, C., & Hawkins, (v0.1) - DLR can benefit obese patients (p.6) Although we currently do not have sufficient data to prove this statistically, we have examples of how DLR offers significantly improved image McLeavy, C. M., Chunara, M. H., Gravell, R. J., Rauf, A., Cushnie, A., Staley Talbot, C., & Hawkins, (v0.1) - Part 1 - Benefit of low dose imaging to high-risk groups (p.2) Low-dose imaging is particularly important in highrisk groups, such as children and pregnant women. Pregnant women are in a hypercoagulable McLeavy, C. M., Chunara, M. H., Gravell, R. J., Rauf, A., Cushnie, A., Staley Talbot, C., & Hawkins, (v0.1) - Part 2 - Benefit of low dose imaging to high-risk groups (p.3) the clinical need outweighs radiation risk; however, a whole-body trauma CT involves significant radiation dose.40 DLR could provide a solu McLeavy, C. M., Chunara, M. H., Gravell, R. J., Rauf, A., Cushnie, A., Staley Talbot, C., & Hawkins, (v0.1) - Part 1 - Benefit of low dose imaging to high-risk groups (p.4) CT of the urinary tract(CT kidneyseuretersebladder, KUB) is becoming the modality of choice for renal calculi follow-up given the low plain
5 and 6. O’Connor, S. D., & Bhalla, M. (2021). Should Artificial Intelligence Tell Radiologists Which Study t (v0.1) - AI can be used to optimise radiology and make it more efficient (p.1) AI tools have the potential to go beyond variables known prior to an examination and incorporate information gleaned from the images thems [PharmaReview Global Global] 13. Several studies have shown excellent accuracy of AI- based methods for the detection and classifica... -> Flanders, A. E., Prevedello, L. M., Shih, G., Halabi, S. S., Kalpathy-Cramer, J., Ball, R., Mongan, (v0.1) - Collection of major dataset for future machine learning applications (p.6) The RSNA Brain Hemorrhage CT Dataset (https://www.kaggle. com/c/rsna-intracranial hemorrhage-detection) is the largest public dataset of i Kuo, W., Hӓne, C., Mukherjee, P., Malik, J., & Yuh, E. L. (2019). Expert-level detection of acute in (v0.1) - Algorithm detecting and classifying CT accurately (p.2) We report a deep learning algorithm with accuracy comparable to that of radiologists for the evaluation of acute intracranial hemorrhage on Ker, J., Singh, S. P., Bai, Y., Rao, J., Lim, T., & Wang, L. (2019). Image Thresholding Improves 3-D (v0.1) - types of brain haemorrhages applied (p.1) In this paper, we apply 3-dimensional convolutional neural networks(3D CNN) to classify computed tomography(CT) brain scans into normal scan
7. Arbabshirani, M. R., Fornwalt, B. K., Mongelluzzo, G. J., Suever, J. D., Geise, B. D., Patel, A. A., (v0.1) - Interpretation time reduced from 512 min to 19 min (p.2) The studies prioritized as “stat” had a median time to clinical interpretation of 19 min(IQR: 22 min), which was significantly lower(p < 0.00
8. Baltruschat, I., Steinmeister, L., Nickisch, H., Saalbach, A., Grass, - 10 (cont) - M., Adam, G., Knopp, T., & Itt (v0.1) - Simulation significantly reduce average RTAT in CXR (p.6) Our clinical workflow simulations demonstrated that a significant reduction of the average RTAT for critical findings in CXRs can be achieved
9. Rezazade Mehrizi, M. H., van Ooijen, P., & Homan, M. (2021). Applications of artificial intelligence (v0.1) - The share of applications focusing on a specific anatomic region (p.5) van Leeuwen, K. G., Schalekamp, S., Rutten, M. J. C. M., van Ginneken, B., & de Rooij, M. (2021). Ar (v0.1) - Fig. 1 Characteristics of 100 CE-marked AI products based on organ- based subspeciality, modality, and main functionality. MSK, musculoskele (p.3)