AI in Medical Imaging: Exploring the Concept of “Radiomics” Volume 54- Issue 3

Fatima Ali Afef Ali AlYafei* and Hasina Umra Khan

• Radiography Specialist at Sheikh Shakhbout Medical City (SSMC), UAE

Received: January 02, 2024; Published: January 10, 2024

*Corresponding author: Fatima Ali Afef Ali AlYafei, Radiography Specialist at Sheikh Shakhbout Medical City (SSMC), UAE

DOI: 10.26717/BJSTR.2024.54.008557

Abstract PDF

The workload of diagnostic radiologists has increased due to the continued growth in imaging data. They must read images more quickly while keeping the same level of accuracy. Simultaneously, this has created the perfect foundation for Al to flourish in the field of medical imaging research. Researchers have used artificial intelligence (AI) to identify intricate patterns in imaging data and provide quantitative assessments of radiographic features. With this regard, Radiomics became one of the most well-known clinical imaging research topic. The following essay aims to provide a brief overview of artificial intelligence (AI) in medical imaging research, including its current role and the obstacles that must be overcome before Al can be widely adopted in clinical settings. Also, to dig deeper into the concept of the radiomics. According to Tang’s article, “The Role of artificial intelligence in medical imaging research,” radiomics is the process of characterizing tumor phenotypes by extracting quantitative data from medical images. When AI algorithms are used to analyze radiomics data, they can find biomarkers and subtle patterns that are invisible to the human eye. This supports the development of personalized treatment plans and prognoses (Bottari, et al. [1]). AI has been successfully used in radiation oncology to program tumor growth and organ division. Lambin P. coined the term “radiomics” in 2012 and defined it as “the high-throughput approach of extracting numerous picture highlights from radiation images. Studies have demonstrated the remarkable potential of artificial intelligence in the field of clinical imaging. Nowadays, AI is required due to a lack of high-quality, high-volume, longitudinal results data, this requirement is further complicated by the conflicting need for strict security and privacy insurance (Bashir, et al, [2]).

In the 1980s, researchers began working on computer-aided detection (CAD) systems. Image modalities such as CT, MRI, and mammography were subjected to traditional machine-learning algorithms. Despite a great deal of research effort, the actual clinical applications showed little promise. Ardila et al. proposed a deep-learning algorithm that predicts the risk of lung cancer based on a patient’s past and current CT volumes. Deep-learning algorithms have evolved into a decision-making system for radiological imaging examination. This includes a variety of imaging modalities, such as CT, X-ray, PET, ultrasound, and so on, as well as a range of tasks, such as segmentation, disease prediction, and cancer detection. Deep learning gathers information from vast amounts of image examples, much like human learning does. That being said, it might take a lot less time because it only needs organized data and the associated metadata. Here, we must refer to Venn’s diagram illustrating the various artificial intelligence (AI) concepts. presented in (Dafydd et al. [3]) work and shown in Figure 1. In radiology, Radiomics consists of two phases. Feature extraction is the initial step where it is possible to incorporate images from different modalities. Image segmentation algorithms are utilized to partition the volumes of interest. Elements will be eliminated following the division. Surface, mathematical data, growth volume, shape, thickness, pixel force, and other elements are examples of normal elements.

Figure 1

The next stage is to incorporate the eliminated highlights into numerical models to interpret the total growth for the anticipated treatment outcome. A successful outcome prediction can provide crucial information for a precise treatment strategy. For instance, patients with lung cancer may share factors such as age and histology (Bottari et al, [1]). Despite this, tumor images may appear distinct, and the endurance time may be completely different. Different treatment plans may be chosen if radiomics can take the image data, deconstruct the aggregate, and predict the endurance time before the treatment. We refer to this as personalized medicine. (Cook, et al. [4]). The progress of radiomics might keep away from unwanted complexities brought about by biopsy and accomplish something similar or a better expected result. A radiomic built by Aerts et al. was evaluated using a separate set of lung data. It showed the translational ability of radiomics across various tumors. Researchers additionally showed a huge relationship between the radiomic highlights and quality articulation designs. Utilizing positron emission tomography (PET) images, PET/CT, or PET/MRI, some researchers carried out radiomics modeling. Most applications were on cellular breakdown in the lungs (Cook, et al [4]). Additionally, prostate and head and neck cancers are potential applications. According to Gillies, et al, Radiomics is most well-developed in oncology because of support from the National Cancer Institute (NCI) Quantitative Imaging Network (QIN) and other initiatives from the NCI Cancer Imaging Program. Moreover, concerns about policies are becoming stronger and more widely implemented in healthcare facilities. Al is limited by a lack of high-quality, high-volume longitudinal outcomes data.

The parameters of the imaging setting and conventions may vary amongst clinical settings, even for similar image modalities on similar illness sites. One of the main challenges facing AI-driven medical imaging research is standardizing data organization across different practices. According to (Bellomi, et al. [5]), clinical imaging information association itself may be worthy of being a prominent area of study. Also, strict privacy regulations secured patient health-related data, which restricted the exchange of images between institutions. Clinics have therefore improved security and information sharing procedures and are now more concerned than ever about liability and protections. However, a significant amount of information from numerous organizations is required for the successful application of artificial intelligence. A test is given on how to share photos without jeopardizing security (Bellomi, et al. [5]). Furthermore, AI is essential in imaging extremities because it provides sophisticated tools for accurate analysis and interpretation [6-10]. Artificial intelligence’s capacity to identify minute details in images of the extremities improves the identification of anomalies, fractures, and joint disorders. This is very helpful for orthopedic evaluations and helps with more precise treatment planning. To conclude, radiomics techniques combined with AI integration in medical imaging represent a paradigm shift in diagnostic capabilities. Together with the knowledge gained from radiomics, these technologies work better together to improve patient outcomes and diagnosis accuracy while also creating new opportunities for personalized medicine. However, before artificial intelligence can become more potent and widely used in healthcare facilities, it must overcome a few obstacles.

1. Bottari F, Cousin F, Desir C, Frix A N, Guiot J Hustinx, et al. (2021) Radiomics in Lung Diseases Imaging: State-of-the-Art for Clinicians. Journal of Personalized Medicine 11(7): 602.
2. Bashir M R, Konkel B, Lafata K J, Wang Y, Yin F F, et al. (2021) Radiomics: a primer on high-throughput image phenotyping. Abdominal Radiology.
3. Dafydd D, Doran S J, Downey K, Kumar S, Koh D M, et al. (2021) Radiomics in Oncology: A Practical Guide. RadioGraphics 41(6): 1717-1732.
4. Cook G, Gibbs P, Häggström I, Langs G, Materka A, et al. (2020) Introduction to Radiomics. Journal of Nuclear Medicine 61(4): 488-495.
5. Bellomi M, Botta F, Fanciullo C, Morganti A G, Origgi D, et al. (2018) Radiomics: the facts and the challenges of image

analysis. European Radiology Experimental 2(1).

6. Ates E, Durmaz ES, Kilickesmez O, Kocak B (2019) Radiomics with artificial intelligence: a practical guide for beginners. Diagnostic and Interventional Radiology 25(6): 485-495.
7. Barucci A, Cioni D, Gabelloni M, Saba L, Neri E, et al. (2021) A deep look into radiomics. La radiologia medica 126(10): 1296-1311.
8. Cardona Cardenas D A, De Sá Rebelo M, De F, Krieger J E, Ferreira Junior J R, et al. (2021). Novel Chest Radiographic Biomarkers for COVID-19 Using Radiomic Features Associated with Diagnostics and Outcomes. Journal of Digital Imaging 34(2): 297-307.
9. Gillies R J, Hricak H, Kinahan P E (2016) Radiomics: Images Are More than Pictures, They Are Data. Radiology 278(2): 563-577.
10. Tang X (2020) The role of artificial intelligence in medical imaging research. BJR|Open 2(1): 20190031.