In our previous blog post, A Nod to Radiology: How Radiology's Digitization Can Inspire the Future Toward a DP Landscape, we visited radiology’s adoption of digital imaging and posited how and what pathology can learn from radiology. We were curious what radiology colleagues were up to these days regarding artificial intelligence (AI). Is radiology further along than pathology in its adoption and acceptance of AI? Afterall, radiology converted to digital images, a necessity for AI training and application, in the mid-1980s. Interestingly, we found radiology and pathology share similar hesitations regarding AI, which include fear of being replaced by AI technology and workflow disruption. Such concerns must be addressed to increase adoption of AI technology and to further experience its benefits of helping physicians provide optimum patient care. A multitude of studies on the applications of AI in radiology are achieving 95% accuracy when compared to gold standard diagnosis. We’ll take a look at those studies, how AI affects the radiology workflow and who is at the heart of this discussion.
AI and the radiologist: why can’t we be friends?
Like pathologists, radiologists fear the inevitable. AI may outperform them, thus denigrating their reputations or putting their jobs at risk. Given our experience and backed up by published studies, we believe this just isn’t so.
One example is AI models that have been trained to classify various maxillofacial cysts and/or tumors. Technically, the procedure of an AI model to classify cysts and/or tumors follows four main steps: lesion detection, segmentation, extraction of texture features and subsequent classification. Currently, the first step of lesion detection in these models is still required to be performed manually by the radiologist so that these models can automatically perform the following steps. It still remains a challenge to develop a fully automated model that can identify cysts and/or tumors.
In addition, although AI may exhibit high performance in one task such as diagnosis of pneumonia, in its current form it cannot replace the radiologist’s role in detecting incidental findings such as asymptomatic tumors. This role for radiologists will continue to be invaluable in the era of worldwide population aging, as large numbers of elderly patients have multimorbidity.
AI could only be an aid to radiologists but will not replace a radiologist. Radiologists who use AI to their benefit, rather than to avoid it out of fear, might supersede those radiologists who do not.
AI will no doubt disrupt the workflow of radiologists, but the disruption brings with it positive effects, most importantly on patient care. It seems more appropriate to describe the workflow change as an enhancement rather than a disruption.
Potential benefits to clinicians and patients start as early as the education of radiologists. Models trained using images labeled by experienced radiologists, specialty radiologists, and/or histopathological reports may in the future provide a training tool to help trainees or general radiologists to gain competence and confidence in difficult diagnoses. AI may also help trained radiologists achieve higher interrater reliability throughout their years in clinical practice. One study demonstrated that the Fleiss’ kappa measure of interrater reliability for detecting anterior cruciate ligament tear, meniscal tear, and abnormality were higher with model assistance than without.
Let’s talk about the increasing workload in radiology. Newer imaging modalities such as CT and MR can provide more detailed information with thinner images and/or multiple series of images, and the time required to collect these images is shorter than before. Therefore, the number of images collected in each examination is increasing, whereas the number of radiologists who interpret these images is not. Radiologist fatigue can be alleviated if AI models can undertake supportive tasks 24 hours a day.
AI can undoubtedly impact every step of a radiologist's workflow. It can simplify every activity like ordering and scheduling, protocoling and acquisition, image interpretation, reporting, communication, and billing. AI could take on time-dependent tasks that can be performed effortlessly, permitting radiologists more time and opportunities to engage in patient care.
AI can also be used to alert radiologists and physicians to patients who require urgent treatment, as in the application in the detection of pneumothorax. The use of AI for proactive screening and detection of urgent imaging findings may prove to be a practical and acceptable workflow enhancement for radiologists.
Some of the clinical applications of AI are quickly improving for breast imaging. It will play an essential role in all steps of mammography and digital breast tomosynthesis (DBT) from image creation and deionizing to risk assessment, cancer detection, and finally, therapy selection and response prediction. AI has a significant role in the interpretation of breast cancer; one study demonstrated the capability of AI to discriminate between benign and malignant lesions using magnetic resonance imaging (MRI) and identify various histological subtypes of breast cancer. In addition, AI has multiple implications in thoracic imaging such as lung nodule assessment, tuberculosis or pneumonia detection, or estimation of diffuse lung diseases.
Liver masses can be classified into five categories (from classical hepatocellular carcinoma as category A to liver cyst as category E) using a combination of dynamic contrast enhanced-computed tomography (CT) images. AI models trained to predict histopathological findings based on noninvasive images, such as AI models that use MR to stage liver fibrosis, may help in reducing the risk of complications from invasive biopsy. The genomic status of gliomas can be estimated by AI models trained on MR images that can predict isocitrate dehydrogenase 1 mutation status and O6-methylguanine-DNA methyltransferase promoter methylation status. A cancer patient’s prognosis may also be estimated with AI. One study reports an AI model was able to stratify patients with non–small cell lung cancer into low- and high-mortality risk groups using standard-of-care CT images. Other AI applications within radiology can assist with image processing at earlier stages. Segmentation of organs or tissues within images is possible with AI. One study reports the development of an AI model that quantifies visceral and subcutaneous fat from MR images of the mouse abdomen.
AI in radiology is not lengths and strides ahead of pathology as one might expect given digital radiology adoption had a decade's worth of a head start. While momentum is indeed gaining to utilize AI in both radiology and pathology settings, it's clear further validation of systems is needed in order for the market to accept AI as an integral tool for clinical decision aid.
Heart of the matter
At the heart of all this is the patient. We were told once by Dr. Donald Karcher, Chair of Pathology at George Washington University, “Put the patient first, and everything will fall into place.” In radiotherapy, from the initial patient encounter to pretreatment disease outcome and toxicity prediction, AI can assist in every step. AI may subsequently aid in treatment planning, dose optimization, and support a greater level of safety, quality, and efficiency of care. Doesn’t every patient deserve this level of care? Instead of being fearful they will be replaced by AI, it may be beneficial for radiologists to view AI as an improved way to provide high-quality, cutting edge care for the benefit of the patient.
Instapath was founded in 2017 by the same engineers and scientists who developed the original prototypes. Our vision is to enable patients to immediately know their cancer diagnosis instead of waiting days or weeks for the results. Instapath builds microscopy platforms to improve patient care in the form of faster turnaround times and prevention of high risk and costly repeat biopsy procedures. Further, our goal is to provide users with a seamless, modernized digital pathology workflow with tools to complete all pathology evaluations needed to provide the most precise and efficient diagnoses for patients.
To learn more about us, visit www.instapathbio.com or email firstname.lastname@example.org.
1. Bansal, G J. “Digital radiography. A comparison with modern conventional imaging.” Postgraduate medical journal vol. 82,969 (2006): 425-8. doi:10.1136/pgmj.2005.038448
2. Coppola, Francesca et al. “Human, All Too Human? An All-Around Appraisal of the "Artificial Intelligence Revolution" in Medical Imaging.” Frontiers in psychology vol. 12 710982. 28 Sep. 2021, doi:10.3389/fpsyg.2021.710982
3. Hung, Kuofeng et al. “The use and performance of artificial intelligence applications in dental and maxillofacial radiology: A systematic review.” Dento maxillo facial radiology vol. 49,1 (2020): 20190107. doi:10.1259/dmfr.20190107
4. Yasaka, Koichiro, and Osamu Abe. “Deep learning and artificial intelligence in radiology: Current applications and future directions.” PLoS medicine vol. 15,11 e1002707. 30 Nov. 2018, doi:10.1371/journal.pmed.1002707
5. Luciano, M et al. “Automated Critical Test Findings Identification and Online Notification System Using Artificial Intelligence in Imaging.” Radiology 2017 285:3, 923-931
6. Bien N, Rajpurkar P, Ball RL, Irvin J, Park AK, Jones E, et al. AI-assisted diagnosis for knee MR: Development and retrospective validation. PLoS Med. 2018;15(11):e1002699 10.1371/journal.pmed.1002699
7. Taylor AG, Mielke C, Mongan J. Automated detection of clinically-significant pneumothorax on frontal chest X-rays using deep convolutional neural networks. PLoS Med. 2018;15(11):e1002697 10.1371/journal.pmed.1002697
8. Geras, Krzysztof J et al. “Artificial Intelligence for Mammography and Digital Breast Tomosynthesis: Current Concepts and Future Perspectives.” Radiology vol. 293,2 (2019): 246-259. doi:10.1148/radiol.2019182627
9. Detection and characterization of MRI breast lesions using deep learning. Herent P, Schmauch B, Jehanno P, et al. Diagn Interv Imaging. 2019;100:219–225.
10. Gampala, Sravani et al. “Is Artificial Intelligence the New Friend for Radiologists? A Review Article.” Cureus vol. 12,10 e11137. 24 Oct. 2020, doi:10.7759/cureus.11137
11. Yasaka K, Akai H, Abe O, Kiryu S. Deep Learning with Convolutional Neural Network for Differentiation of Liver Masses at Dynamic Contrast-enhanced CT: A Preliminary Study. Radiology 2018;286(3):887–896. 10.1148/radiol.2017170706
12. Yasaka K, Akai H, Kunimatsu A, Abe O, Kiryu S. Liver Fibrosis: Deep Convolutional Neural Network for Staging by Using Gadoxetic Acid-enhanced Hepatobiliary Phase MR Images. Radiology 2018;287(1):146–155.
13. Chang P, Grinband J, Weinberg BD, Bardis M, Khy M, Cadena G, et al. Deep-Learning Convolutional Neural Networks Accurately Classify Genetic Mutations in Gliomas. AJNR Am J Neuroradiol 2018;39(7):1201–1207. 10.3174/ajnr.A5667
14. Hosny A, Parmar C, Coroller T, Grossmann P, Zeleznik R, Kumar A, et al. Deep learning for lung cancer prognostication: A retrospective multi-cohort radiomics study. PLoS Med. 2018;15(11):e1002711 10.1371/journal.pmed.1002711
15. Grainger AT, Tustison NJ, Qing K, Roy R, Berr SS, Shi W. Deep learning-based quantification of abdominal fat on magnetic resonance images. PLoS ONE 2018;13(9):e0204071 10.1371/journal.pone.0204071
16. Artificial intelligence in radiation oncology. Deig CR, Kanwar A, Thompson RF. Hematol Oncol Clin North Am. 2019;33:1095–1104.