Introduction

Over the past three decades, magnetic resonance imaging (MRI) has become a routine diagnostic procedure. In 2013, the estimated number of operating scanners worldwide was over 30,000, and the number of examinations performed each year reached over 100 million [1]. The constant increase in the number of available locations and the growing number of examinations performed using this method has allowed for its broad use in medicine.
In the 1980s, it was believed that computed tomography (CT) would be replaced by MRI. That did not happen. Over the years, CT has become one of the most commonly used imaging methods. In the context of 3D modelling of CT images, important processes took place in the 1990s, when CT scanners with continuous rotation of the lamp-detector system were introduced, which was enabled by slip ring technology. The introduction of the so-called spiral or helical data acquisition was fundamental in the development and continuous improvement of CT imaging techniques. For the first time, volumetric data of whole organs became available. Images could be reconstructed in any positions. This was a major achievement compared to previous data collection techniques of this type, which provided only a few slices per organ. The option to acquire volumetric data has also paved the way for the application of 3D image processing techniques such as multiplanar transformations, maximum intensity projections, surface shaded displays or volumetric rendering (VRT) techniques in computed tomography [2].

Objective

In this review article, we review articles on the issue of 3D modelling in CT and MRI and its use in diagnostic imaging, with particular emphasis on spinal imaging. The objective of the performed analyses is to determine the possible applications of 3D modelling in imaging diagnostics and its effectiveness. We also determine the direction of development in the field of MRI and CT.

Material and methods

A PubMed search was performed for articles up to 2023 using the following search strategy, yielding a certain number of results:
1. 3D MRI models*[Title/Abstract]) – 10;
2. “3D CT models” *[Title/Abstract] – 41
3. “3D MRI models”[Title/Abstract]) AND “3D CT models” *[Title/Abstract] – 5.
4. “3D CT models” *[Title/Abstract] AND spine*[MeSH Terms] – 3.
Ultimately, 36 articles were selected for further analysis.
Exclusion criteria applied:
• studies published before 1990;
• articles which are not in English;
• the full text of the article is not available;
• the articles discuss only CT or MRI.
After selecting the articles that would be used for the review, they were narrowed down to two categories: development trends in MRI and CT imaging and 3D modelling in diagnostic imaging. In total, 36 articles were used in the review.

Results and discussion

Directions of development of MRI and CT imaging
When analysing the directions of development of CT and MRI, it should be noted that one of the most widespread trends in both methods are actions aimed at creating a large base of installed devices, as well as the improvement of 3D systems. Some of these scanners are now even supplied as hybrid or bimodal imaging systems (combination of PET – positron emission tomography and MRI). There is also growing interest in integrating MRI soft tissue imaging capabilities with interventional procedures such as MRI-guided focused ultrasound surgery [1].
Another direction of work on the development of MRI use involves activities in the field of accelerating image acquisition. Image acquisition in MRI is the process of acquiring data that is used to create images of the patient’s body. Benefits of accelerating MRI image acquisition include reduced motion or data flow, the ability to capture anatomical movement (e.g. a desirable phenomenon in the cardiac imaging process), shorter patient scan duration, and more economical use of high-intensity resources. Improvements in MRI result from the use of techniques such as sparse sampling (or compressed detection). These techniques use repetitive patterns in space or time that are characteristic of MRI. Thanks to them, we can get better images using less data. In other words, we can achieve high-quality images while reducing the number of scans, saving time and increasing scanning efficiency [1].
Around 2015, another trend appeared and it is associated with the use of artificial intelligence in imaging diagnostics. It has been proven that it can surpass humans in some areas. Thus, the registration and segmentation of typical classical image processing tasks has been shown to be better performed by artificial intelligence. Many typical MRI artifacts (e.g. caused by patients’ movements) can be identified and removed from images by artificial intelligence [3]. In addition to these final processing applications, artificial intelligence has already been used to reconstruct the image directly from undersampled data [4].
Attempts to use artificial intelligence to improve image quality are also made in the context of CT imaging. These AI-based image reconstruction techniques share the common goal of improving CT image quality. These methods turned out to be very promising [5].
Over the years, the ability to probe the local tissue environment with multiple MRI contrasts and the development of fast and immobile sequences has developed. The speed of image acquisition has been improved and in the future, advanced model-based reconstruction algorithms supported by artificial intelligence will allow for more reliable and better diagnostic information, increasing diagnostic confidence, offering comfortable and highly patient-oriented images [1].

3D modelling and its use in imaging diagnostics
Finite element modelling plays a key role in the study of musculoskeletal biomechanics. To develop highly accurate and advanced finite element models, diagnostic imaging is essential. The model building process involves several steps. Firstly, it involves segmentation, which is the process of assigning each voxel in an image to specific tissue, which enables three-dimensional reconstruction. Then, in the meshing process, a computational mesh is created. The next step is to assign material properties to individual parts of the model [6].
Computed tomography (CT) is an imaging method most commonly used to create three-dimensional (3D) models for assessing bone and joint morphology in clinical practice. However, 3D models based on MRI data can be just as effective in comprehensive and accurate assessment of the morphology and pathology of bone structures and soft tissues. The quality of 3D MRI models continues to improve with increasing potential to replace 3D CT models in a variety of musculoskeletal system applications. In practice, a single 2D and 3D MRI examination can increase the value of MRI and simplify pre- and postoperative diagnostic imaging. It has been shown that 3D MRI models work very well in the case of shoulder, hip (in the case of femoroacetabular impingement) and knee joint pathologies [7].
The importance of 3D modelling in diagnostic imaging was demonstrated by Rabinov et al. [8]. The researchers showed that thanks to 3D modelling it is possible to generate a detailed picture of anatomical structures. This advantage of 3D MRI modelling was also pointed out by H. Busse et al. (2008), who used 3D MRI modelling to image the shoulder joint [9]. Similar conclusions were reached by Rodrigues et al., Anzia et al., and Alt et al. [10–12]. Inoue et al. showed that the method using 3D MRI models was characterized by a low level of invasiveness for patients and was useful in preoperative planning of the procedure [13]. Suter et al. proved that preoperative planning using 3D CT is characterized by high accuracy [14]. This is also pointed out by the research team of Fahrni et al. and Noguerol et al. [15, 16].
In view of the above, it is recognized that the usefulness of post-processed 3D MRI models will continue to grow and have an increasing number of applications. Computer-aided and AI-assisted final processing techniques have great potential to improve the efficiency of 3D modelling, opening up many opportunities to assess the usefulness of 3D MRI and establish 3D-specific quantification criteria [7].
The main advantage of 3D MRI over standard 2-dimensional MRI is its ability to reduce artifacts, average partial volume, and create multiplanar reconstructions (MPRs) in any plane with any slice thickness from a single high-resolution isotropic acquisition. 3D MRI acquisitions are particularly useful for assessing articular cartilage, which is prone to volume averaging artifacts, and for evaluating longitudinally running structures such as peripheral nerves and tendons, which are better seen with non-perpendicular MPRs. 3D magnetic resonance imaging is also useful during the surface and volume analysis of bones and cartilage before surgery. Current research is aimed at shortening the acquisition time and automating segmentation through machine learning. Thanks to the achievements in this area, it will be possible to overcome the limitations of 3D MRI and create new opportunities for the use of this technique. 3D MRI is now broadly used in musculoskeletal imaging and the popularity of this technique will continue to increase in the coming years. The conducted research focuses primarily on accelerated acquisition techniques and quantitative imaging [17].
Yamaguchi et al. proved usefulness of 3D CT modelling in the evaluation of the degree of deformation of the lumbar spine in the course of osteoarthritis [18]. Kishimoto et al., as part of their research, showed the role of 3D CT modelling in determining the surface and diameter of the vertebral end plate, in surgical procedures, designing and selecting spinal implants [19]. Based on 3D full-spinal computed tomography (CT) models, mesh models of the upper and lower end plates of the bone were developed with a high level of precision.
In their research, Kohyama et al. demonstrated usefulness of 3D modelling in the creation of 3D MRI-CT fusion images allowing for the assessment of the severity of OCD (osteochondrosis dissecans) lesions, due to the exact relationship between the location of the articular cartilage and the subchondral bone [20]. Dolatowski et al. found that 3D-CT models can be used as a reference standard for the assessment of displacements of anatomical structures [21]. In their studies on bone anatomy Chee et al. showed that the differences between 3D-MR and CT models were acceptable (the maximum difference was less than 3.5 mm) [22].
Wagner et al. proved that 3D modelling can also be used in the creation of sacral implants used as an alternative to sacroiliac screws in the treatment of pelvic and sacrum fractures associated with osteoporosis [23]. The possibility of using 3D modelling in the creation of implants implanted in the spine is also pointed out by Haq et al. [24].
The research team of Nagamatsu et al. also discussed the subject of 3D modelling in their research. The authors emphasized that CT/MRI fusion images have recently become available for assessing the anatomy of the spine before performing decompression surgeries or interbody stabilization procedures [25]. Achievements in this area make it possible to assess not only soft tissues such as the nerve root, intervertebral disc, but also bone structures. The Kambin’s Triangle is accessed using a novel CT/MRI fusion imaging technique that extracts magnetic resonance imaging (MRI) images of nerve tissue and intervertebral disc and 3D CT images of bone and combines them into a complex image showing all structures in relation to each other [26].
For many years, the intervertebral foramen was referred to as a “hidden zone” and imaging diagnostics of this anatomical structure was very difficult. Therefore, despite the fact that foramen stenosis in the lumbar spine was identified, there were no available methods to accurately assess its extent. Fusion imaging of the lumbar vertebrae (CT) and the spinal cord (MRI) is not yet commonly used in diagnostics. In their study, Yamanaka et al. undertook to consider the possibility of using 3D MRI/CT fusion imaging. The project was successfully completed. The researchers have shown the possibility of using this method in diagnostics. It was useful in the imaging of nerve root damage in the foramen and extra-foramen areas. The image obtained was realistic and allowed them to better determine the level of pathology than in the case of conventional imaging methods [27].
Another research project discussing the possibility of using 3D modelling in imaging diagnostics was developed by Kamogawe et al. [28]. The authors of this research project analysed the possibility of using 3D models in the diagnostics of degenerative cervical radiculopathy. They noted that there is a lack of imaging methods that would provide detailed three-dimensional (3D) images of cervical nerve roots. MRI is the method of choice for patients with cervical radiculopathy, but there are no clear guidelines for nerve root detection, and root MRI varies from hospital to hospital. CT alone is of limited value in the assessment of cervical radiculopathy, although it is useful in distinguishing the extent of exostosis or foramen infiltration, or the presence of ossification of the posterior longitudinal ligament. Even CT myelography cannot accurately detect the entire nerve root. Most spinal surgeons are able to combine CT and MRI images to perform a thorough evaluation of the pathoanatomical structure. Evaluation of cervical radiculopathy requires both imaging methods because the nerve root is a very small and soft organ while a bone spur is very hard. When planning surgery, the location of the affected nerve root and the extent of the bone spur decompressed should be taken into account [28].
Jung-Ha Kim et al. (2018) reviewed 14 studies on the diagnostic accuracy of lumbar disc herniation imaging [29]. Based on their analysis, they found that the specificity and sensitivity of MRI, CT and myelography were comparable. Similar results were obtained by T. Maus (2010), who stated that the sensitivity and specificity of computed tomography, CT myelography and MRI in the diagnostics of lumbar spinal stenosis and disc herniation are similar [30].
Sometimes MRI is considered the “gold standard” in the lumbar spine diagnostics. It is recommended, e.g. in the diagnostics of radiculopathy when the usual medical treatment fails, or in the examination of lumbar spinal stenosis. Patients with contraindications to MRI should undergo computed tomography [31]. MRI is also a very helpful method in preoperative planning, e.g. MRI is the basis of the 4-point IVF (intervertebral foramen) stenosis assessment system proposed by T.S. Jeong et al. [32].
The diagnostic parameter differentiating surgery – Foraminal Stenotic Ratio – was developed and evaluated by Yamada et al. using 3D magnetic resonance imaging [33]. The ratio was significantly different between the groups subjected to conservative and surgical treatment. On its basis, it was proved that lower IVF stenosis in the lumbar region requires surgical treatment in symptomatic patients with moderate accuracy. Interestingly, when the spinal canal is ≥ 50% occupied by fat obliteration, failure probability of the conservative treatment was 75% [33].
Manabe et al. (2019) used a novel diffusion-weighted magnetic resonance neurography [34]. The applied technique of visualization of nerve tracts in the lumbosacral region was the basis for the statement that 36.6% of patients with lumbar radiculopathy had a high angle of nerve root departure (≥60°) as a result of degenerative changes. Such results show a new potential of MRI in indirect root diagnostics [34].
MRI and CT imaging was discussed in the article by Yamanaki et al. [27]. For their research project, the research team combined 3D computed tomography (CT) bone imaging with 3D magnetic resonance imaging (MRI) of the neural architecture (cauda equina and nerve roots) in two patients. The authors of the study used the VirtualPlace software for this purpose. Although the Yamanaki research team failed to assess the pathology of the nerve roots using MRI, myelography or CT myelography, 3D MRI/CT fusion imaging became the basis for confirming the pathology. It also enabled them to determine the course of the nerve roots, both inside and outside the alveolar foramen, visualized the thickening of the ligamentum lutea and the location, shape and number of dorsal root ganglia. It is also possible to present positional relationships between intervertebral discs or bone spurs and nerve roots [27].
The use of 3D MRI/CT fusion imaging of the lumbar spine successfully revealed the relationship between bone structure (bones, intervertebral joints and discs) and nervous system architecture (cauda equina and nerve roots) in one video, in 3D and in colour. Such images can be useful in explaining complex neurological conditions, such as degenerative lumbar scoliosis (DLS), as well as in the diagnostics and planning of surgery. In their study M. Sammin et al. sought to determine whether 3D-MR imaging of the hip joint could be used to accurately represent the femur and acetabulum morphology in the assessment of patients with femoroacetabular impingement [35].
For this purpose, Samimi’s research team retrospectively reviewed 17 cases (19 hips) with suspected femoroacetabular impingement. Study participants underwent both 3D-CT and 3D-MRI of the same hip joint. Radiologists reviewed imaging for the presence and location of cam deformity, anterior-inferior iliac spine variant, lateral mid-edge angle, and neck angle. The 3D-CT results were taken as the reference standard. The amount of radiation that was avoided after the introduction of 3D-MRI was also assessed [35].
All 17 patients with suspected FAI (femoroacetabular impingement) had a cam deformity on 3D-CT. There was 100% match concerning the diagnosis (19 out of 19) and location (19 out of 19) of the cam deformity when comparing 3D-MRI to 3D-CT. 3D-CT showed three variants of type I and sixteen of type II anteroinferior iliac spines with an 89.5% (17 out of 19) match between 3D-MRI and 3D-CT for characteristics of the anteroinferior iliac spines. The match was 64.7% when comparing the measurements of the neck-body angle (11 out of 17) and the Wiberg angle (11 out of 17). The use of 3D-MRI allowed each patient to avoid an average effective radiation dose of 3.09 mSV, which gives a total reduction of 479 mSV over a period of 4 years [35].
Studies by Samim et al. have demonstrated that 3D-MR imaging can be used to accurately diagnose and quantify the typical bone pathology in the case of femoroacetabular impingement and could potentially eliminate the need for 3D-CT imaging and associated radiation exposure as well as costs for this predominantly young group of patients [35].
Another research project in which the possibility of using 3D modelling in imaging diagnostics was discussed was the study conducted by Jardon et al. [36]. In their research, the authors made a comparison between the 2D MRI image and the 3D MRI image of the cervical spine. In this analysis, they used a reconstruction algorithm based on deep learning. Jardon et al. assumed that the better image quality possible thanks to the use of a 3D model would have a better diagnostic value in the case of spinal stenosis than conventional 2D acquisitions [36].
The obtained material showed that in the case of spinal stenosis, the match was higher in the case of 3D models. The match in diagnosing central stenosis was at a similar level. In addition, this study demonstrated that 3D models had less noticeable movement artifact (p ≤ 0.001-0.036). The average total scanning time was 10.8 minutes for 2D sequences and 7.3 minutes for 3D sequences [36,29]. The cited project showed that 3D MRI modelling provides better quality diagnostic materials than traditional MRI in a shorter time.

Conclusions

The review carried out in this article shows that 3D modelling has a significant potential in diagnostic imaging. The progress that has been made over the last three decades has meant that 3D MRI and 3D CT allow for even more accurate imaging of anatomical structures within the spine in a shorter time. However, further research on the development of this technology is necessary to better understand the potential of using 3D modelling in the diagnostics of complex anatomical structures. There is no doubt that the extensive use of information technology in medicine will facilitate this process.

References

1. Hayden ME, Nacher P-J. History and physical principles of MRI. Luca SABA. Magnetic Resonance Imaging Handbook, 1. CRC press; 2016.
2. Hsieh J, Flohr T. Computed tomography recent history and future perspectives. Journal of Medical Imaging. 2021 8. doi:10.1117/1.JMI.8.5.052109.
3. Börnert P, Norris DG. A half-century of innovation in technology—preparing MRI for the 21st century. Br J Radiol. 2020 93;20200113. doi:10.1259/bjr.20200113.
4. Lundervold AS, Lundervold A. An overview of deep learning in medical imaging focusing on MRI. Z Med Phys. 2019 29;102–27. doi:10.1016/j.zemedi.2018.11.002.
5. Lee T. The Use of Artificial Intelligence in Computed Tomography Image Reconstruction: A Systematic Review. Radiology. 2020.
6. Galbusera F, Cina A, Panico M, et al. Image-based biomechanical models of the musculoskeletal system. Eur Radiol Exp. 2020 4;49. doi:10.1186/s41747-020-00172-3.
7. Samim M. 3D MRI Models of the Musculoskeletal System. Semin Musculoskelet Radiol. 2021 25;388–96. doi:10.1055/s-0041-1730912.
8. Rabinov JD, Barker FG, McKenna MJ, et al.. Virtual Cisternoscopy: 3D MRI Models of the Cerebellopontine Angle for Lesions Related to the Cranial Nerves. Skull Base. 2004 14;93–9. doi: 10.1055/s-2004-828701.
9. Busse H, Thomas M, Seiwerts M, et al. In vivo glenohumeral analysis using 3D MRI models and a flexible software tool: Feasibility and precision. Journal of Magnetic Resonance Imaging. 2008 27;162–70. doi:10.1002/jmri.21228.
10. Cantarelli Rodrigues T, Deniz CM, Alaia EF, et al. Three-dimensional MRI Bone Models of the Glenohumeral Joint Using Deep Learning: Evaluation of Normal Anatomy and Glenoid Bone Loss. Radiol Artif Intell. 2020 2;e190116. doi: 10.1148/ryai.2020190116.
11. Anzia LE, Johnson CJ, Mao L, et al. Comprehensive non-invasive analysis of lower urinary tract anatomy using MRI. Abdominal Radiology. 2021 46;1670–6. doi:10.1007/s00261-020-02808-9.
12. Alt CD, Hampel F, Hallscheidt P, et al. 3 T MRI-based measurements for the integrity of the female pelvic floor in 25 healthy nulliparous women. Neurourol Urodyn. 2016 35;218–23. doi:10.1002/nau.22697.
13. Inoue A, Kohno S, Nishida N, et al. Clinical utility of new three-dimensional model using a zero-echo-time sequence in endoscopic endonasal transsphenoidal surgery. Clin Neurol Neurosurg. 2020 190;105743. doi:10.1016/j.clineuro.2020.105743.
14. Suter T, Gerber Popp A, Kolz CW, et al. Accuracy of free-hand humeral head resection planned on 3D-CT models in shoulder arthroplasty: an in vitro analysis. Arch Orthop Trauma Surg. 2021 142;3141–7. doi:10.1007/s00402-021-03931-y.
15. Fahrni S, Campana L, Dominguez A, et al. CT-scan vs . 3D surface scanning of a skull: first considerations regarding reproducibility issues. Forensic Sci Res. 2017 2;93–9. doi:10.1080/20961790.2017.1334353.
16. Martín-Noguerol T, Paulano-Godino F, Riascos RF, et al. Hybrid computed tomography and magnetic resonance imaging 3D printed models for neurosurgery planning. Ann Transl Med. 2019 7;684–684. doi:10.21037/atm.2019.10.109.
17. Altahawi F, Subhas N. 3D MRI in Musculoskeletal Imaging: Current and Future Applications. Current Radiology Reports. 2018.
18. Yamaguchi H, Nojiri H, Miyagawa K, et al. Segmental coupling effects during correction of three-dimensional lumbar deformity using lateral lumbar interbody fusion. European Spine Journal. 2020 29;879–85. doi:10.1007/s00586-020-06310-0.
19. Kishimoto M, Akeda K, Sudo A, et al. In vivo measurement of vertebral endplate surface area along the whole-spine. Journal of Orthopaedic Research. 2016 34;1418–30. doi:10.1002/jor.23354.
20. Kohyama S, Nishiura Y, Hara Y, et al. A novel three-dimensional MRI-CT image fusion technique for precise preoperative evaluation and treatment of capitellar osteochondritis dissecans. Eur Radiol. 2021 31;5721–33. doi:10.1007/s00330-020-07680-1.
21. Dolatowski FC, Temmesfeld MJ, Pierre-Jerome C, et al. Bilateral symmetrical comparison of the proximal femur using 3D-CT models. Surgical and Radiologic Anatomy. 2018 40;507–13. doi:10.1007/s00276-018-1968-6.
22. Chee CG, Chung HW, Kim W, et al. Differences between 3D isovoxel fat suppression VIBE MRI and CT models of proximal femur osseous anatomy: A preliminary study for bone tumor resection planning. PLoS One. 2021 16;e0250334. doi:10.1371/journal.pone.0250334.
23. Wagner D, Kamer L, Sawaguchi T, et al. Critical dimensions of trans-sacral corridors assessed by 3D CT models: Relevance for implant positioning in fractures of the sacrum. Journal of Orthopaedic Research. 2017 35;2577–84. doi:10.1002/jor.23554.
24. Haq R, Aras R, Besachio DA, et al. 3D lumbar spine intervertebral disc segmentation and compression simulation from MRI using shape-aware models. Int J Comput Assist Radiol Surg. 2015 10;45–54. doi:10.1007/s11548-014-1094-9.
25. Nagamatsu M, Ruparel S, Tanaka M, et al. Assessment of 3D Lumbosacral Vascular Anatomy for OLIF51 by Non-Enhanced MRI and CT Medical Image Fusion Technique. Diagnostics. 2021 11;1744. doi:10.3390/diagnostics11101744.
26. Nagamatsu M, Maste P, Tanaka M, et al. Usefulness of 3D CT/MRI Fusion Imaging for the Evaluation of Lumbar Disc Herniation and Kambin’s Triangle. Diagnostics. 2022 12;956. doi:10.3390/diagnostics12040956.
27. Yamanaka Y, Kamogawa J, Katagi R, et al. 3-D MRI/CT fusion imaging of the lumbar spine. Skeletal Radiol. 2010 39;285–8. doi:10.1007/s00256-009-0788-5.
28. Kamogawa J, Kato O, Morizane T, et al. Virtual pathology of cervical radiculopathy based on 3D MR/CT fusion images: impingement, flattening or twisted condition of the compressed nerve root in three cases. Springerplus. 2015 4;123. doi:10.1186/s40064-015-0898-6.
29. Kim J-H, van Rijn RM, van Tulder MW, et al. Diagnostic accuracy of diagnostic imaging for lumbar disc herniation in adults with low back pain or sciatica is unknown; a systematic review. Chiropr Man Therap. 2018 26;37. doi:10.1186/s12998-018-0207-x.
30. Maus T. Imaging the Back Pain Patient. Phys Med Rehabil Clin N Am. 2010 21;725–66. doi:10.1016/j.pmr.2010.07.004.
31. Adams A, Roche O, Mazumder A, et al. Imaging of degenerative lumbar intervertebral discs; linking anatomy, pathology and imaging. Postgrad Med J. 2014 90;511–9. doi:10.1136/postgradmedj-2013-132193.
32. Jeong TS, Ahn Y, Lee SG, et al. Correlation between MRI Grading System and Surgical Findings for Lumbar Foraminal Stenosis. J Korean Neurosurg Soc. 2017 60;465–70. doi:10.3340/jkns.2016.1010.004.
33. Yamada K, Abe Y, Satoh S, et al. A novel diagnostic parameter, foraminal stenotic ratio using three-dimensional magnetic resonance imaging, as a discriminator for surgery in symptomatic lumbar foraminal stenosis. The Spine Journal. 2017 17;1074–81. doi:10.1016/j.spinee.2017.03.010.
34. Manabe H, Sakai T, Miyagi R, et al. Identification of abnormalities in the lumbar nerve tract using diffusion-weighted magnetic resonance neurography. European Spine Journal. 2019 28;849–54. doi:10.1007/s00586-018-05867-1.
35. Samim M, Eftekhary N, Vigdorchik JM, et al. 3D-MRI versus 3D-CT in the evaluation of osseous anatomy in femoroacetabular impingement using Dixon 3D FLASH sequence. Skeletal Radiol. 2019 48;429–36. doi:10.1007/s00256-018-3049-7.
36. Jardon M, Tan ET, Chazen JL, et al. Deep-learning-reconstructed high-resolution 3D cervical spine MRI for foraminal stenosis evaluation. Skeletal Radiol. 2023 52;725–32. doi:10.1007/s00256-022-04211-5.