“There is no question that the appearance of big data and machine learning offer an exciting chance for revolution, but revolutions demand greater scrutiny, not less.”
A Lancet Digital Health paper that definitely scrutinized the precision medicine revolution.
Imaging Wire Sponsors
- Bayer Radiology – Providing a portfolio of radiology products, solutions, and services that enable radiologists to get the clear answers they need.
- GE Healthcare – Enabling clinicians to make faster, more informed decisions through intelligent devices, data analytics, applications and services.
- Healthcare Administrative Partners – Empowering radiology groups through expert revenue cycle management, clinical analytics, practice support, and specialized coding.
- Hitachi Healthcare Americas – Delivering best in class medical imaging technologies and value-based reporting.
- Nuance – AI and cloud-powered technology solutions to help radiologists stay focused, move quickly, and work smarter.
- Riverain Technologies – Offering artificial intelligence tools dedicated to the early, efficient detection of lung disease.
The Imaging Wire
Precision Medicine Reality Check
A new Lancet Digital Health Viewpoint questioned the belief that machine learning will enable precision medicine breakthroughs, suggesting that some major pivots and compromises might be required in order to revolutionize healthcare. Here’s some details on this argument:
- Reality Check 1 – Although precision medicine optimists expect machine learning to “enable automated diagnoses with unprecedented accuracy,” the authors believe imaging AI diagnostic science is still at the “acknowledging potential” stage and actual scientific evidence remains lacking (few studies, flawed studies, focus on performance vs. clinical value).
- Reality Check 2 – The authors also questioned forecasts that machine learning-powered precision medicine will allow clinicians to identify the best therapies for individual patients. Although this is a common precision medicine goal, the authors suggest that it is “probably also the least likely to ever be fully achieved” because AI can’t identify cause and effect.
- Their Solution – Rather than trying to predict individual patients’ responses to treatment, the authors suggest that precision medicine researchers should instead focus on stratified medicine (identifying and predicting subgroups with a better and worse response). Stratified medicine might be a “more modest” approach than personalized medicine, but they believe it’s a better fit with our current statistical capabilities and it “could still provide a genuine revolution.”
- Counter Critique – The radiology community quickly pointed out that none of the paper’s twelve authors are clinicians, suggesting that they might have omitted some arguments (e.g. questioning why models focus on diagnosis vs. risk) if they had an actual healthcare AI user on board.
German researchers revealed PET/MRI’s advantages over PET/CT, including higher cancer detection rates and reduced radiation exposure, while matching PET/CT’s treatment staging performance. That’s a pretty big deal, and here are some details:
- The Study – The researchers had radiologists and nuclear medicine physicians review scans from 918 cancer patients who underwent both PET/MRI and PET/CT imaging (1,003 exams), comparing lesion detection and radiation dosage.
- The Results – The readers detected additional lesions in 26.3% of the PET/MRI scans, which included 5.3% additional malignant findings, and led to 29 staging changes. Meanwhile, the PET/CT exams revealed additional lesions in 2.9% of the exams, including 1.2% additional malignant findings that led to five staging changes. The PET/MRI scans also had far less radiation exposure (3.6 ± 1.4 mSv vs. 17.6 ± 8.7 mSv).
- The Takeaways – Despite PET/CT’s position as the standard oncology imaging modality, PET/MRI’s higher soft-tissue contrast improves lesion detection and reduces the need for additional scans. This could be especially valuable for younger patients and for ongoing patient monitoring.
- Next Steps – The researchers called for more studies focused on PET/MRI’s technical and operational challenges, as well as studies involving different technology vendors and larger participant cohorts. If those studies are similarly positive, it might “pave the way toward a widespread introduction of PET/MRI into clinical patient care.”
- Context Avoids Opioids: A University of Washington study found that adding context to lumbar spine imaging reports reduces the likelihood that referring physicians will prescribe opioids, but it doesn’t change spine-related health care utilization. Over a three year period, the participating physicians produced 250,401 spinal imaging reports that either followed a standard structure (control group) or included additional prevalence data on similar-age patients without back pain (intervention group). The intervention group patients had a “small but significant decrease” in likelihood of opioid prescriptions (0.95 odds ratio) over the next year.
- Walmart Health Turns One: Walmart just celebrated the first anniversary of its Walmart Health clinics, revealing strong patient feedback and loyalty so far (96% positive reviews, >50% booked return visits), while emphasizing the centers’ comprehensive capabilities (yes, including imaging). Looking into 2021, the retail giant plans to increase its presence to 22 locations (vs. 6 now) with most of the openings taking place in Georgia (+7) and Florida (+7). Twenty-two locations may seem minor compared to Walmart’s >4,750 retail storefronts and the >9k urgent care centers across the U.S., but Walmart’s goal of becoming “America’s neighborhood health destination” is very notable.
- The DBT+SM Advantage: Women with non-dense breasts screened with both DBT and synthetic 2D mammography (DBT+SM) have lower recall rates and higher screening-detected breast cancer rates than those only screened with digital mammography. That’s from a randomized controlled trial in Norway (n = 28,749) that found DBT+SM resulted in lower recall rates than DM among women with Volpara Density Grades of 1 (2.1% vs. 3.3%) and 2 (3.2% vs. 4.3%), while DBT+SM screening had higher detection rates than DM among women from every density grade (2.4 to 2.8 vs. 1.7 to 2.2).
- GE Clariscan Gets Polymer Bottle: GE Healthcare’s polymer +PLUSPAK Pharmacy Bulk Package gained FDA approval for use with GE’s Clariscan gadolinium-based MRI contrast agent. GE Healthcare emphasized the Clariscan +PLUSPAK Pharmacy Bulk Package’s shatterproof design (safety, productivity), ability to transfer to sterile syringes (flexibility), and support for weight-based doses (reducing waste).
- Medicaid Expansion and Thyroid Cancer Diagnosis: University of Michigan researchers found that Medicaid expansion states experienced a notable increase in thyroid cancer diagnoses (post-expansion), which is consistent with increased diagnoses of other cancers in these states. Using data from 246,296 patients diagnosed with thyroid cancer during the 2010-2012 and 2014-2016 periods, the researchers found that expansion states saw thyroid cancer diagnoses increase among Medicaid beneficiaries (3.2 to 5.6 cases per 100k) and people without Medicaid (16.7 to 17.5 cases per 100k). Meanwhile, thyroid cancer diagnoses in non-expansion states increased from just 1.6 to 2 cases per 100k Medicaid beneficiaries. The researchers also suggest that these increases were not associated with higher overdiagnosis rates.
- RANZCR’s AI Risk Standards: The Royal Australian and New Zealand College of Radiologists (RANZCR) launched a new risk-management framework to guide the development, deployment, and monitoring of healthcare AI. The new framework comes almost exactly a year after RANZCR released its AI ethics standards, as the group continues to build out its AI framework.
- FDG-PET/CT for NSCLC Recurrence: New research out of Japan detailed FDG-PET/CT’s ability to detect recurrence among postoperative non-small-cell lung cancer (NSCLC) patients. The researchers studied 496 FDG-PET/CT exams from 187 NSCLC patients who underwent potentially curative surgery, finding that FDG-PET/CT correctly diagnosed recurrence in 46 of 47 (97.9%) patients and identified 68 of 69 (98.6%) recurrent sites. The researchers suggested that annual FDG-PET/CT follow ups could be an effective way to monitor for NSCLC recurrence during the three years after surgery.
- RADLogics & PVmed: US/Israel-based imaging AI company RADLogics and Chinese imaging AI firm Perception Vision Medical Technology Company (PVmed) launched a partnership that will combine their respective strengths / resources to advance the companies’ product development efforts and expand their global reach.
- Automated Body Composition AI: A Germany-based team developed an AI model that performs fully automated body composition analysis using abdominal CT scans, suggesting that an approach like this could allow body tissue composition analysis across the whole abdomen (not just on L3 slices). The researchers built the 3D semantic segmentation CNN using 40 training and 10 testing CTs, all fully annotated on every fifth axial slice with five semantic body regions. The model achieved a 0.9553 Dice score across all semantic regions and >0.99 intra-class correlation coefficients for subclassified tissues.
- Siemens’ New c.cam: Siemens Healthineers unveiled its next-generation c.cam Cardiac SPECT system in the U.S., highlighting its ease-of-installation (2 days), patient/access comfort (reclining chair vs. flat table), cybersecurity features (new Windows platform), and low TCO.
- COVID Imaging Variations: A new European Radiology study of 50 radiology departments in 33 countries found notable COVID imaging variations. For example, some countries used both CXR and CT for severe COVID-19 cases (U.S., Australia, Germany, Iran), while some used a mix of CXR, CT, and US (England, India, Canada), and other countries had inconsistent use between different hospitals (Brazil, Italy, Spain, France, S. Africa). These variations are notable given that 98% of the hospitals claimed to follow standard imaging guidelines.
- US + SWE for CTS: A team of Thomas Jefferson radiologists found that high-frequency ultrasound and shear wave elastography (SWE) could help improve carpal tunnel syndrome (CTS) diagnosis. The researchers examined 25 patients (20 control hands, 27 hands w/ CTS) using SWE and US, finding that by measuring cross-sectional area (w/ US) and stiffness (w/ SWE) they could diagnose CTS with 96% accuracy, 100% specificity, and 93% sensitivity (vs. 94%, 100%, 85% by only measuring cross-sectional area).
- RapidAI’s $25m: RapidAI landed $25m in Series B funding, revealing plans to use the capital to support the global expansion of its cerebrovascular imaging AI solutions.
The Resource Wire
– This is sponsored content.
- Acknowledging the pressures that hospitals and imaging centers are under during the COVID-19 emergency, Hitachi’s latest blog details the equipment financing programs have become most in demand and unveils its new post-COVID programs launched in partnership with Key Equipment Finance.
- This Riverain Technologies case study details how Einstein Medical Center adopted ClearRead CT enterprise-wide (all 13 CT scanners) and how the solution allowed Einstein radiologists to identify small nodules faster and more reliably.
- This GE Healthcare piece details how clinicians adapted their obstetrical exam ultrasound protocols to scan the lungs of expecting mothers for signs of COVID-19, in addition to scanning the fetus.
- Patients have become savvy healthcare shoppers who increasingly rely on price information to make decisions about their care. Join Healthcare Administrative Partners’ CRO, Rebecca Farrington, as she discusses price transparency & consumerism in radiology in this upcoming RBMA webinar.
- In this Bayer Radiology video, East Texas Medical Center Radiology Director, Bill Tobin, details how they used Bayer’s MEDRAD Stellant Smart Injector and contrast dose management to reduce contrast volumes and repeat scans.
- This Nuance case study details how PowerShare Image Sharing helped Navicent Health cut its mammography reporting time by 38%.