the science behind
RADIOMIC PRECISION
METRICS (RPM)™
FIRST, MEET THE BRAINS BEHIND IT.
This is David Gering, Ph.D.
HealthMyne’s Quantitative Imaging Decision Support (QIDS)® software platform leverages years of knowledge and development by a top-tier team led by Dr. David Gering, a well known leader in the imaging industry whose qualifications are unparalleled. We know that solid algorithmic techniques are critical to ensure that our customers have confidence in the accuracy and consistency of the quantitative imaging data our platform generates, which is why we partnered with someone with David’s impeccable credentials:
- Ph.D. from MIT, Artificial Intelligence Lab
- Developed 3D Slicer as part of his thesis project
- Nearly 20 years of commercial development in the areas of segmentation and deformable registration
- 18 US Patents
- 40 scientific publications with over 2,300 citations
OUR SECRET INGREDIENTS
Shhhhh…don’t tell anyone
It starts with you. Once a radiologist or other multidisciplinary team member uses the Radiomic Precision Metrics (RPM)™ functionality to identify a lesion our algorithm kicks in and these methods are put to work. First of all, our algorithm goes way beyond just looking at Hounsfield Units (HU) values. We can’t tell you the specifics of how they all work, but we built it with a combination of the following disciplines to produce accurate 3D lesion contours that take every voxel into account.
SPECIAL ORGAN
INTELLIGENCE
Special organ intelligence methods are used along with the maths-driven disciplines.
NORMAL VS
ABNORMAL TISSUE
One example is in the lung where we separate normal versus emphysematous lung
VESSEL
MAPPING
These special maps allow our algorithm to subtract vessels from lesions in order to correctly define the boundaries in multiple organs
AN ORGAN MASK THAT
INFORMS AND CONTAINS
The normal organ informs and contains the contours
of lesions near boundaries
AUTOMATIC ORGAN
IDENTIFICATION
All of this is made possible by automatically identifying organs so QIDS knows which intelligence to apply
OUR ALGORITHM IS RIGOROUSLY
TESTED AND VALIDATED
Even with our secret ingredients and special organ intelligence being as amazing as they are, we still make sure everything works through rigorous testing and validation. One way we do that is putting the RPM functionality up against the four expert radiologists in the LIDC dataset. We found that our quantitative volumetric delineation was significantly more consistent (93% vs 85%) and matched each individual radiologist decidedly better than they matched each other (0.87 vs 0.68). Proof that RPM can greatly reduce intra- and inter-reader variability. In a recent ongoing segmentation competition (BraTS 2018), as of Spring 2019, out of 121 entries, our algorithms were in 1st place for Hausdorff Distance, and a very close 2nd place for DICE Coefficient.
CONTINUAL AND
ONGOING VALIDATION
CONTINUAL AND
ONGOING IMPROVEMENTS
At HealthMyne, we release new software every 120 days. Our product releases are loaded with new functionality, content, workflows and iterative updates to our algorithms as we continually incorporate data and feedback from the entire patient care team. Our developers and clinical specialists do not get much time on the beach, which is facilitated by our location in Madison, WI.
AI and Deep Learning are innovative technologies that HealthMyne embraces to strengthen our algorithms, which by the way are novel and based on our imaging scientists’ many years of learning and experience, i.e. “actual intelligence”! We are committed to leveraging all new and existing approaches to ensure we deliver consistent metrics that our clinical customer base needs to facilitate precise patient management.
CONCLUSION
Radiomic Precision Metrics (RPM)™ provides quantitative imaging metrics that are significantly more consistent than a Radiologist measuring on their own. More consistent measurements translate into less uncertainty for clinicians when making important treatment decisions and can lead to better outcomes for patients. Additionally, RPM makes other evidence based metrics such as Volume, Density, Mass, %GGO, Doubling Time and Texture available with a click of the mouse. Each of these can add clinical context when the long and short measurements are stable or inconclusive, further improving the ability of doctors to provide the most timely and applicable treatments. HealthMyne is a leader in the science of quantitative imaging, and we continue to invest significantly to keep our edge. Contact us today to get a demonstration of the HealthMyne platform featuring the Radiomic Precision Metrics (RPM)™ functionality to see how we can make a significant difference in your clinical and research workflow.