Article

Advances in Imaging for Clarity of Oncology Visualizations

Emerging technology in radiation therapy (RT) imaging is addressing barriers to progress in a variety of therapeutic protocols, including combination, individualized and personalized treatment. As a locally administered procedure like surgery, the efficacy of RT is dependent on timely, sensitive, specific, and accurate imaging techniques that assist clinicians in defining key aspects of a disease, such as extent, heterogeneity, proximity, and adjacency to normal tissues.

While many recent technological improvements include software upgrades that advance a variety of processes, such as iterative dosing algorithms that improve efficacy and reduce toxicity, other advancements in oncology visualizations are creating new standards for viewing patient data.

Augmenting visualizations

State of the art medical imaging systems, including computed tomography (CT) and magnetic resonance imaging (MRI), now have the capability to generate high-resolution volumetric datasets containing over 1000 cross-sectional images substantially increasing the amount of available patient data.1,2

Increasing data volume means new methods for viewing are needed for conducting fast and accurate processing that supports critical decision-making.2 Otherwise, clinicians risk missing important details that can result in serious patient consequences, such as an initial presentation in an advanced stage of cancer that impacts survival, patient anxiety, and cost of treatment.Virtual reality (VR) and augmented reality (AR) offer two methods for viewing large volume data sets, though neither has been approved by the FDA for diagnostic radiology.1,2

GEHC_Insights_Banner_700x120_01_050120.jpg

Researchers in France developed an AR system from patient CT and/or MRI medical images that increase a surgeon’s intraoperative vision during liver surgery.3 This minimally invasive surgical procedure requires highly accurate knowledge of liver anatomy.3 It also provides the best survival rate for patients suffering from hepatic cancer.3 Until now, eligibility has been about 50 percent.3 Eligibility was increased by approximately 20 percent with the use of two major processes: 3-dimensional (3D) modeling and visualization of tumors combined with the creation of a virtual transparency of the patient during real-time surgical procedures.3

Between January 2009 and June 2013, researchers modeled 769 patients.3 Software was designed to combine direct volume rendering and surface rendering.3 This was followed by the development of one interactive and one automatic registration technique to produce an intraoperative augmented reality view.3 Ultimately, more than 50 interactive AR-assisted surgical procedures were developed from these initial 3D models illustrating the potential for increasing safety and overcoming other current limits.3

Researchers concluded virtual patient modeling should be mandatory for certain interventions, such as liver surgery, and that AR is the obvious next step in surgical instrumentation.3 They believe intraoperative medical imaging and evolution of Hybrid ORs plus the development of automated AR can overcome current limitations created by the complexity of organ deformations.3

Limitations of AR/VR include the possibility for some structure overlapping, potential user motion sickness during the assessment process, and acceptance of a bulky HMD unit.1,2

Artificial intelligence enhances visualizations

Artificial intelligence (AI) technology has proven it can produce brain MRI images faster with better signal and less noise than conventional techniques.4 Deep learning algorithms are enabling a range of advanced capabilities that improve on a variety of patient and clinician issues, from real-time disease assessment during point-of-care interventions to faster exam times.5 Given the expected impact, current industry reports estimate that AI in the healthcare sector will reach nearly eight billion dollars by 2022, growing at a rate of more than 52 percent.5

One example of AI capability impact on medical imaging was reported by a team of researchers who developed an advanced computing technique specifically designed to rapidly and cost-effectively improve the quality of visualizations.4 Using a set of 50,000 MRI brain scans from the Human Connectome Project for data extraction, they were able to train their AI-based system to apply algorithmic approaches in ways that reconstruct images with reduced noise and artifacts over existing techniques.4

The research team's new AI system uses machine learning and software inspired by the brain’s ability to process information and make choices about the best computational strategies for producing clear, accurate images for a range of medical imaging.4 Since MRI uses magnetic field and radio waves to create detailed images of tissues inside the body, noise from either of these electronic sources or body tissue can reduce the image quality.4 Identifying ways to lessen the noise have always been a priority.4

Researchers concluded their new AI technique could not only become a turning point for the low-dose MRI used in their investigations, but may also benefit other significant MRI applications known for producing low signal-to-noise ratio like multi-nuclear spectroscopy.4 By comparison, mainstream approaches often rely on expensive hardware or prolonged scan times to improve imaging quality.4

3D printing for tumor visualization

Today 3D printed tumor models derived from MRI and CT datasets are being used to plan and guide both surgeries and radiation therapy, teach colleagues, and help explain procedures to patients.6 One team of radiologists is equating the potential impact of 3D printed tumor models to that of CT imaging over plain x- rays.6

Compared to 3D printing of bones and the heart, the soft tissue of tumors is considered a more complicated process because it can become intertwined with the soft tissue parts of organs.6 To ensure a 3D tumor model will be an accurate representation, researchers recommend segmenting the model slice by slice using a thickness of less than one millimeter.6

While the majority of models to-date are created from CT data, MRI is of particular interest because of its excellent tissue characterization and lack of ionizing radiation.Researchers in Texas successfully created MRI-based preoperative 3D printed patient-specific tumor molds using volumetric segmentation of six renal masses.8 Investigators confirmed the molds allowed for increased accuracy in the sectioning of tumors after surgical removal as well as colocalization of in vivo imaging features with tissue-based analysis in radiomics.8

All patients underwent successful partial nephrectomy where the resected mass was bivalved through the preselected MRI plane.8 A slicing guide template was created with notches to correspond with anatomic locations on the magnetic resonance images.8 A thick slab of the tumor was acquired, fixed, and processed as a whole-mount slide and found to correlate with multiparametric MRI findings.8 Distinct in vivo MRI features corresponded with unique pathologic characteristics in the same tumor.8 The average cost of printing each mold ranged from $20.90 to $350.70.8

In a recent prostate cancer study conducted by UCLA researchers, the accuracy of 3D printed molds was successfully characterized by quantifying their usefulness for facilitating MRI-pathology registration.9 Investigators reported that patient-specific molds improved accuracy relative to manual slicing techniques in a phantom model although registration accuracy of surgically resected specimens was limited by their imperfect fit within the molds.9 Ex vivo imaging was recommended to overcome this limitation.9

Tissue-mimicking prostate phantoms created with embedded reference marks were used to measure and compare target registration error (TRE) between phantoms.9 Ten radical prostatectomy specimens were placed inside 3D molds, scanned in MRI and then sliced.9 Phantoms sliced by hand were found to have a 4.1 mm mean TRE versus mold-sliced phantoms which had a 1.9 mm mean TRE.9

Additional findings included a reduced mean angular misalignment around the left-right anatomic axis from 10.7 to 4.5 degrees with mold-assisted slicing.9 A mean 14-degree rotation about the left-right axis was revealed with ex vivo MRI due to misalignment of excised prostates within molds.9 When using molds alone, the mean in-plane TRE was 3.3 mm which was reduced to 2.2 mm after registration was corrected with ex vivo MRI.9

In the data-driven era of precision and personalized medicine, advancements in imaging for greater clarity of oncology visualizations are essential components in meeting the needs of present and future therapeutic outcomes, as well as increasing clinician capability for developing individualized protocols.

REFERENCES:

  1. Augmented Reality Imaging System: 3-dimensional Viewing of a Breast Cancer. Journal of Natural Science. http://www.jnsci.org/files/html/2016/e215.htm Accessed 4/30/2018
  2. Augmented Reality: Advances in Diagnostic Imaging. Multimodal Technologies and Interaction. http://www.mdpi.com/2414-4088/1/4/29/htm Accessed 4/30/2018
  3. Real-time 3-dimensional image reconstruction guidance in liver resection surgery. HSBN Hepatobiliary Surgery & Nutrition. http://hbsn.amegroups.com/article/view/3634/4540 Accessed 6/22/2018
  4. Artificial Intelligence Provides Faster, Clearer MRI Scans. Imaging Technology News. https://www.itnonline.com/content/artificial-intelligence-provides-faster-clearer-mri-scans Accessed 7/18/2018
  5. Artificial Intelligence in Healthcare Market by Offering (Hardware, Software and Services), Technology (Deep Learning, Querying Method, NLP, and Context Aware Processing), Application, End-User Industry, and Geography – Global Forecast to 2022. MarketsAndMarkets. https://www.marketsandmarkets.com/Market-Reports/artificial-intelligence-healthcare-market-54679303.html Accessed 7/18/2018
  6. 3D  Printing Models, Augmented Reality Images Help Surgeons Visualize Tumors. RSNA Daily Bulletin. https://rsna2017.rsna.org/dailybulletin/index.cfm?pg=17thu07 Accessed 7/18/2018
  7. 3D printing from MRI Data: Harnessing strengths and minimizing weaknesses. Journal of Magnetic Resonance Imaging. https://www.ncbi.nlm.nih.gov/pubmed/27875009 Accessed 7/18/2018
  8. Development of a Patient-specific Tumor Mold Using Magnetic Resonance Imaging and 3-Dimensional Printing Technology for Targeted Tissue Procurement and Radiomics Analysis of Renal Masses. Urology. https://www.ncbi.nlm.nih.gov/pubmed/29056576 Accessed 7/18/2018
  9. Registration Accuracy of Patient-Specific, 3D Printed Prostate Molds for Correlating Pathology with Magnetic Resonance Imaging. IEEE Transactions on Biomedical Engineering. https://www.ncbi.nlm.nih.gov/pubmed/29993431 Accessed 7/18/2018