Innovation in Medical Imaging Never Stops

In the public’s imagination, most medical innovation happens by accident [1]. We imagine the discovery of penicillin by studying moldy petri dishes, accidently infaring a spouse’s hand with X-Rays (link to earlier post on the discovery of radiology) or discoveries in cardiac surgery. However, innovation is the practical implementation of a set of ideas which can result in the optimization of existing services or products, which takes deliberate, painstaking precision from many different sources of information. In the context of medical imaging, notable advances have been made due to increasing how different forms of data are used from imaging exams with the aim of reducing time to diagnosis. This has included introducing artificial intelligence and machine learning (AI/ML) to create extensions or effective changes in the trajectory of processes considered obsolete or redundant. 

Unlike the innovations of the past, medical imaging advances may have less to do with the image acquisition itself, and instead be more about the cutting-edge technologies from other industries propelling it forward [2]. Since the 1990s, medical imaging innovation has experienced huge advances that reduce the time to diagnosis of diseases such as cancer through artificial intelligence techniques. AI has been shown to be a tool with high potential in the screening of mammography exams [3]–[5]. Correlation of clinical findings with imaging findings, as well as the interpretation of both, require a high extensive workload for radiologists and physicians in general. 

Recent developments optimize risk-based breast cancer screening policies with reinforcement learning and to be equally accurate across race. Lead author Adam Yala and a team of scientists from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and Jameel Clinic created robust artificial intelligence tools to predict future cancer using just a patient’s mammogram [6]–[8]. Cardoso et al. created Tempo, a novel reinforcement learning-based framework for personal screening papers on demonstrating its efficacy for breast cancer by testing it against a mammography dataset from Mass General. Their goal was to balance between over-screening and early detection [9]

In this context, carried out the comparison of commercial artificial intelligence systems with the double readings routine in population based-screening [10]. The present study comprised a database which included 752 patients and 205 with interval breast cancer or cancer detected between screening rounds. As a result, the authors show that the artificial intelligence system was able to predict on a scale of 0 (lowest risk) and 10 (highest risk) an 87.6% of the cases and 44.9% of the interval breast cancer  had the highest scores through the system of artificial intelligence. These findings may contribute to reducing the workload through  reducing the volume of reading by radiologists and physicians in general. Despite the encouraging findings about the role, Algen et al. reinforce the limited generalization of the findings, due to the lack variety of the database. 

One of the main gaps in medical imaging research refers to the lack of sample diversity of the studies. In view of this, Gradient presents a business model with the objective of contributing to increasing the data diversity and characteristics of medical image data within research through a platform with instant access to a worldwide database that has more than one hundred centers, including hospitals, clinical research organizations and academic clinical research associations.



[1] M. Parrish, “5 Life-Saving Medical Discoveries That Happened On Accident,” Medical Design and Outsourcing, Jun. 16, 2017. (accessed Oct. 24, 2022).

[2] “Innovations in Medical Imaging – From Images to Informatics,” The Alliance of Advanced BioMedical Engineering. (accessed Oct. 24, 2022).

[3] D. Georgian-Smith et al., “Can Digital Breast Tomosynthesis Replace Full-Field Digital Mammography? A Multireader, Multicase Study of Wide-Angle Tomosynthesis,” Am. J. Roentgenol., vol. 212, no. 6, pp. 1393–1399, Jun. 2019, doi: 10.2214/AJR.18.20294.

[4] E. E. Knippa et al., “Impact of Colorized Display of Mammograms on Lesion Detection,” J. Breast Imaging, vol. 2, no. 1, pp. 22–28, Feb. 2020, doi: 10.1093/jbi/wbz075.

[5] X. Wang, G. Liang, Y. Zhang, H. Blanton, Z. Bessinger, and N. Jacobs, “Inconsistent Performance of Deep Learning Models on Mammogram Classification,” J. Am. Coll. Radiol., vol. 17, no. 6, pp. 796–803, Jun. 2020, doi: 10.1016/j.jacr.2020.01.006.

[6] R. Gordon and M. Csail, “MIT Mirai: Robust Artificial Intelligence Tools To Predict Future Cancer,” SciTechDaily, Feb. 03, 2021. (accessed Jan. 24, 2022).

[7] A. Yala et al., “Multi-Institutional Validation of a Mammography-Based Breast Cancer Risk Model,” J. Clin. Oncol., p. JCO.21.01337, Nov. 2021, doi: 10.1200/JCO.21.01337.

[8] A. Yala et al., “Optimizing risk-based breast cancer screening policies with reinforcement learning,” Nat. Med., pp. 1–8, Jan. 2022, doi: 10.1038/s41591-021-01599-w.

[9] M. J. Cardoso, N. Houssami, G. Pozzi, and B. Séroussi, “Artificial Intelligence (AI) in Breast Cancer Care – Leveraging multidisciplinary skills to improve care,” Artif. Intell. Med., p. 102000, Dec. 2020, doi: 10.1016/j.artmed.2020.102000.

[10] Larsen M, Aglen CF, Lee CI, Hoff SR, Lund-Hanssen H, Lång K, Nygård JF, Ursin G, Hofvind S. Artificial Intelligence Evaluation of 122 969 Mammography Examinations from a Population-based Screening Program. Radiology. 2022 Jun;303(3):502-511. 

Scroll to Top
Scroll to Top