Strategies for Implementing Machine Learning Algorithms in the Clinical Practice of Radiology.

Despite recent advancements in machine learning (ML) applications in health care, there have been few benefits and improvements to clinical medicine in the hospital setting. To facilitate clinical adaptation of methods in ML, this review proposes a standardized framework for the step-by-step implementation of artificial intelligence into the clinical practice of radiology that focuses on three key components: problem identification, stakeholder alignment, and pipeline integration. A review of the recent literature and empirical evidence in radiologic imaging applications justifies this approach and offers a discussion on structuring implementation efforts to help other hospital practices leverage ML to improve patient care. Clinical trial registration no. 04242667 © RSNA, 2024 Supplemental material is available for this article.

An investigation into disposal and recycling options for daily disposable and monthly replacement soft contact lens modalities.

PURPOSE: To examine the annualised waste and end-of-life disposal options with two representative soft contact lens (CL) modalities. METHODS: The component parts of two representative soft CL modalities were catalogued, separated, weighed and inspected for material identification: somofilcon A soft CLs (clariti elite, CooperVision Inc.) used with multi-purpose solution (MPS) (All in one Light, CooperVision Inc.) and somofilcon A CLs (clariti 1 day, CooperVision Inc). Using a model that assumed compliant wear and care of CLs, the mass of material solid waste generated by CL use over a year was calculated. Disposal options were explored using household and specialist recycling streams in order to develop recommendations for responsible disposal of CL waste. RESULTS: Full-time daily disposable (DD) CL wear generates 1.06 kg of waste annually compared to 0.83 kg generated by reusable-monthly replacement daily wear ('reusable') CLs. Plastic was the dominant material in both modalities. With full-time use of DD CLs, 64% of waste by mass was plastic blister trays. For full-time use of reusable CLs, where figures from lens and MPS packaging are combined, plastics accounted for 67% of waste by mass. MPS bottles alone made up almost half the waste (45%) associated with full-time reusable CL wear. CONCLUSION: Full-time DD wear generates 27% more waste annually than full-time reusable lens wear. Reusable CL wearers can recycle 78% of waste at home. DD lens wearers have access to recycling options that allow them to recycle 100% of CL related waste. Full-time CL lens wear represents just 0.20-0.26% of the 412 kg of household waste generated per person, per year in the United Kingdom. Worn CLs should never be disposed of down the sink or lavatory. CL wearers should be aware of responsible end-of-life recycling and disposal options for all CL waste.

A 53-Year-Old Man Presents to the ED With Shortness of Breath, Cough, and Fever.

CASE PRESENTATION: A 53-year-old man presented to the ED at a time of low severe acute respiratory syndrome coronavirus 2, also known as coronavirus disease 2019 (COVID-19), prevalence and reported 2 weeks of progressive shortness of breath, dry cough, headache, myalgias, diarrhea, and recurrent low-grade fevers to 39°C for 1 week with several days of recorded peripheral capillary oxygen saturation of 80% to 90% (room air) on home pulse oximeter. Five days earlier, he had visited an urgent care center where a routine respiratory viral panel was reportedly negative. A COVID-19 reverse transcriptase polymerase chain reaction test result was pending at the time of ED visit. He reported a past medical history of gastroesophageal reflux disease that was treated with famotidine. Travel history included an out-of-state trip 3 weeks earlier, but no recent international travel.

The Role of Imaging in the Management of Suspected or Known COVID-19 Pneumonia: A Multidisciplinary Perspective.

COVID-19 is an illness caused by a novel coronavirus that has rapidly escalated into a global pandemic leading to an urgent medical effort to better characterize this disease biologically, clinically and by imaging. In this review, we present the current approach to imaging of COVID-19 pneumonia. We focus on the appropriate utilization of thoracic imaging modalities to guide clinical management. We will also describe radiologic findings that are considered typical, atypical and generally not compatible with of COVID-19 infection. Further, we review imaging examples of COVID-19 imaging mimics, such as organizing pneumonia, eosinophilic pneumonia and other viral infections.

The Role of Imaging in the Management of Suspected or Known COVID-19 Pneumonia. A Multidisciplinary Perspective.

Coronavirus disease (COVID-19) is an illness caused by a novel coronavirus that has rapidly escalated into a global pandemic leading to an urgent medical effort to better characterize this disease biologically, clinically, and by imaging. In this review, we present the current approach to imaging of COVID-19 pneumonia. We focus on the appropriate use of thoracic imaging modalities to guide clinical management. We also describe radiologic findings that are considered typical, atypical, and generally not compatible with COVID-19. Furthermore, we review imaging examples of COVID-19 imaging mimics, such as organizing pneumonia, eosinophilic pneumonia, and other viral infections.

Predictive value of the combination of lesion location and volume of ischemic infarction with rehabilitation outcomes.

PURPOSE: In acute ischemic stroke, infarct location and volume have, separately, been used to determine long-term outcomes after stroke. Little information exists on the combination of these imaging characteristics on rehabilitation outcomes. We evaluated the association between infarct lesion location volume with the Functional Independence Measure (FIM) measure during inpatient rehabilitation facility (IRF) in ischemic stroke patients. METHODS: Between 2012 and 2014, we prospectively enrolled consecutive acute ischemic stroke patients admitted from a Comprehensive Stroke Center and followed to an IRF in Chicago, Illinois. We adjudicated infarct volumes in specific lesion locations using a validated brain atlas. Volumes were calculated using an automated pipeline. FIM measure was extracted from IRF charts. We analyzed the association between acute infarct characteristics and functional measures using adjusted Spearman's correlation. RESULTS: Among 162 stroke patients (mean age 67.6 years, 48.1% male, 58.6% Caucasian), the median FIM at IRF was 52 (IQR 36-62). In an adjusted analysis, infarct volumes in the left basal ganglia (rs = - 0.45, p = 0.02) and left frontal lobe (rs = - 0.38, p = 0.04) were negatively correlated with FIM scores. CONCLUSIONS: There is an association between specific infarct lesion location volume and subsequent FIM scores assessed at IRF. Our findings suggest that imaging during index stroke hospitalization could be used to predict outcomes assessed during IRF.

Infarct location is associated with quality of life after mild ischemic stroke.

Background In patients with mild ischemic stroke, small but eloquent infarcts may have devastating effects, particularly on health-related quality of life. Aim This study investigates the association between acute infarct location and three-month health-related quality of life in patients with mild ischemic stroke. Methods We evaluated consecutively enrolled patients from a single center between August 2012 and July 2013. Our primary outcome at three months was impairment in any health-related quality of life domain (upper extremity, lower extremity, executive function, and general concerns) defined by a T-score <45. We analyzed the association between acute infarct locations and impaired health-related quality of life at three months in univariate and multivariable analysis. Results Among 229 patients (mean age 64.9 years, 55% male, 29.7% black, and median initial NIHSS score 1), impaired health-related quality of life was noted in 84 (36.2%) patients at three months. In univariate analysis, patients with subcortical infarcts (56.0% vs. 39.3%, p = 0.02) and brainstem infarcts (21.4% vs. 10.3%, p = 0.02) were more likely to have impaired health-related quality of life. In multivariable analysis, patients with subcortical and/or brainstem infarcts had increased odds of impaired health-related quality of life (adjusted OR 2.54, 95% CI 1.29-5.01, p = 0.01). Conclusions After mild ischemic stroke, subcortical and brainstem infarct locations predict impairment in health-related quality of life.

Predicting Domain-Specific Health-Related Quality of Life Using Acute Infarct Volume.

BACKGROUND AND PURPOSE: Limited data exist on the relationship between acute infarct volume and health-related quality of life (HRQOL) measures after ischemic stroke. We evaluated whether acute infarct volume predicts domain-specific Neuro-Quality of Life scores at 3 months after stroke. METHODS: Between 2012 and 2014, we prospectively enrolled consecutive patients with ischemic stroke and calculated infarct volume. Outcome scores at 3 months included modified Rankin Scale and Neuro-Quality of Life T scores. We evaluated whether volume organized by quartiles predicted modified Rankin Scale and HRQOL scores at 3 months using logistic and linear regression as appropriate, adjusting for relevant covariates. We calculated variance accounted for (R2) overall and by volume for each domain of HRQOL. RESULTS: Among 490 patients (mean age 64.2±15.86 years; 51.2% male; 63.3% White) included for analysis, 58 (11.8%) were disabled (modified Rankin Scale score of >2) at 3 months. In unadjusted analysis, the highest volume quartile remained a significant predictor of 1 HRQOL domain, applied cognition-general concerns (R2=0.06; P<0.001). Our fully adjusted prediction model explained 32% to 51% of the variance in HRQOL: upper extremity (R2=0.32), lower extremity (R2=0.51), executive function (R2=0.45), and general concerns (R2=0.34). CONCLUSIONS: Acute infarct volume is a poor predictor of HRQOL domains after ischemic stroke, with the exception of the cognitive domain. Overall, clinical and imaging variables explained <50% of the variance in HRQOL outcomes at 3 months. Our data imply that a broad range of factors, some known and others undiscovered, may better predict poststroke HRQOL than what is currently available.

An empirical method for reducing variability and complexity of myocardial perfusion quantification by dual bolus cardiac MRI.

PURPOSE: Myocardial perfusion can be quantified using the "dual bolus" technique, which uses two separate contrast boluses to avoid signal nonlinearity in the blood pool. This technique relies on knowing the precise ratio of contrast concentrations between the two boluses. In this study, we investigated the variability found in these ratios, as well as the error it introduces, and developed a method for correction. METHODS: Five dogs received dual bolus myocardial perfusion MRI scans. Perfusion was calculated separately using assumed contrast dilution ratios and empirically determined contrast ratios. Perfusion was compared with reference standard fluorescent microspheres. The same technique was then applied to a cohort of six patients with no significant coronary artery stenosis by cardiac catheterization. RESULTS: Assumed contrast dilution ratios were 10:1 for all animal and patient scans. Empirically derived contrast ratios were significantly different for animal (8.51:1 ± 1.53:1, P < 0.001) and patient scans (7.32:1 ± 2.27:1, P < 0.01). Incorporating empirically derived ratios for animal scans improved correlation with microspheres from 0.84 to 0.90 (P < 0.05). CONCLUSION: Variability in dual bolus contrast concentration ratios is an important source of experimental error, especially outside of a carefully controlled laboratory setting. Empirically deriving the correct ratio is feasible and improves the accuracy of quantitative perfusion measurements. Magn Reson Med 77:2347-2355, 2017. © 2016 International Society for Magnetic Resonance in Medicine.

Eyelid Margin and Meibomian Gland Characteristics and Symptoms in Lens Wearers.

PURPOSE: To describe the lid margin characteristics of contact lens wearers and relate them to comfort during lens wear. METHODS: Three study sites enrolled habitual contact lens wearers. Subjects completed the Comfort domain of the Contact Lens User Experience (CLUE) questionnaire, and each eye was graded for the presence of mucocutaneous junction (MCJ) displacement, lid margin irregularity, and lid margin vascularity. Examiners counted the number of meibomian gland (MG) orifices in the central centimeter of the lower eyelid and the number of those that showed pouting/plugging and vascular invasion. MG expressibility was graded according to the Shimazaki schema. Subjects were grouped based on presence/absence of each characteristic, total number of orifices (≥5 vs. <5), and expressibility (grade 0 vs. >0). Descriptive statistics are reported. A linear model was used to assess the fixed effect of each characteristic on combined CLUE score and each CLUE statement, if the effect on combined CLUE score showed p < 0.10. RESULTS: The study included 203 subjects (67.5% female) with mean age (±SD) of 30.3 ± 9.6 years. The most commonly observed characteristics were orifice pouting/plugging, compromised MG expressibility, and lid margin vascularity (35.0, 30.3, and 20.4%, respectively). MCJ displacement and MG expressibility had an effect on the combined CLUE score such that individual CLUE statements were analyzed (p = 0.01 and p = 0.06, respectively). MCJ displacement had an effect on comfort upon insertion (p = 0.01), comfort after 5 minutes (p = 0.03), end-of-day comfort (p = 0.01), and ability to maintain ocular moisture (p = 0.030). MG expressibility had a significant effect on general comfort (p = 0.01), comfort throughout the day (p = 0.02), and the ability to maintain ocular moisture (p = 0.02). CONCLUSIONS: MCJ displacement and MG expressibility have an effect on contact lens comfort.

A non-invasive assessment of cardiopulmonary hemodynamics with MRI in pulmonary hypertension.

PURPOSE: We propose a method for non-invasive quantification of hemodynamic changes in the pulmonary arteries resulting from pulmonary hypertension (PH). METHODS: Using a two-element Windkessel model, and input parameters derived from standard MRI evaluation of flow, cardiac function and valvular motion, we derive: pulmonary artery compliance (C), mean pulmonary artery pressure (mPAP), pulmonary vascular resistance (PVR), pulmonary capillary wedge pressure (PCWP), time-averaged intra-pulmonary pressure waveforms and pulmonary artery pressures (systolic (sPAP) and diastolic (dPAP)). MRI results were compared directly to reference standard values from right heart catheterization (RHC) obtained in a series of patients with suspected pulmonary hypertension (PH). RESULTS: In 7 patients with suspected PH undergoing RHC, MRI and echocardiography, there was no statistically significant difference (p<0.05) between parameters measured by MRI and RHC. Using standard clinical cutoffs to define PH (mPAP>25mmHg), MRI was able to correctly identify all patients as having pulmonary hypertension, and to correctly distinguish between pulmonary arterial (mPAP>25mmHg, PCWP<15mmHg) and venous hypertension (mPAP>25mmHg, PCWP>15mmHg) in 5 of 7 cases. CONCLUSIONS: We have developed a mathematical model capable of quantifying physiological parameters that reflect the severity of PH.

Automated analysis of perfusion weighted MRI using asymmetry in vascular territories.

PURPOSE: To determine the feasibility of automatic vascular territory region of interest (ROI) construction as a method for standardized quantification of cerebral blood flow (CBF) images. MATERIALS AND METHODS: An algorithm for automatic construction of vascular territory ROIs was performed on 10 healthy controls and 25 patients with perfusion abnormalities identified by retrospective chart review. The ROIs were used to quantify perfusion asymmetry for each territory, and perfusion asymmetry was compared in the two cohorts and against blinded neuroradiologist interpretation. The algorithm was additionally applied to a separate cohort of 23 prospectively enrolled patients and perfusion asymmetry was correlated against clinical variables. RESULTS: There was significantly greater perfusion asymmetry in territories graded by neuroradiologists as hypoperfused compared to those graded as normally perfused (p<.05) and compared to healthy volunteers (p<.01). An ROC analysis showed that perfusion asymmetry was sensitive and specific for identifying hypoperfusion in vascular territories (84.9% sensitivity and 90.5% specificity for a threshold asymmetry index of .829). In the prospective cohort, perfusion asymmetry was correlated with initial NIH stroke scale (NIHSS) (p<.01) and length of stay (p<.05). CONCLUSIONS: Automatic construction of vascular territory ROIs and calculation of perfusion asymmetry is a feasible method for analyzing CBF images. Because the technique is rapid and minimizes bias, it can facilitate analysis of larger scale research studies.

Leakage and water exchange characterization of gadofosveset in the myocardium.

PURPOSE: To determine the compartmentalization of the blood pool agent gadofosveset and the effect of its transient binding to albumin on the quantification of steady-state fractional myocardial blood volume (fMBV). METHODS: Myocardial vascular fraction measurements were simulated assuming the limiting cases (slow or fast) of two-compartment water exchange for different contrast agent injection concentrations, binding fractions, bound and free relaxivities, and true cardiac vascular fractions. fMBV was measured in five healthy volunteers (4 males, 1 female, average age 33) at 1.5T after administration of five injections of gadofosveset. The measurements in the volunteers were retrospectively compared to measurements of fMBV after three serial injections of the ultra-small, paramagnetic iron oxide (USPIO) blood pool agent ferumoxytol in an experimental animal. The true fMBV and exchange rate of water protons in both human and animal data sets was determined by chi square minimization. RESULTS: Simulations showed an error in the measurement of fMBV due to partial binding of gadofosveset of less than 30%. Measured fMBV values over-estimate simulation predictions, and approach cardiac extracellular volume (22%), which suggests that the intravascular assumption may not be appropriate for the myocardium, although it may apply to more distal perfusion beds. In comparison, fMBV measured with ferumoxytol (5%, with slow water proton exchange across vascular wall) agree with published values of myocardial vascular fraction. Further comparison between myocardium relaxation rates induced by gadofosveset and by other extracellular and intravascular contrast agents showed that gadofosveset behaves like an extracellular contrast agent. CONCLUSIONS: The distribution of the volunteer data indicates that a three-compartment model, with slow water exchange of gadofosveset and water protons between the vascular and interstitial compartments, and fast water exchange between the interstitium and the myocytes, is appropriate. The ferumoxytol measurements indicate that this USPIO is an intravascular contrast agent that can be used to quantify myocardial blood volume, with the appropriate correction for water exchange using a two-compartment water exchange model.

Complex regional pain syndrome is associated with structural abnormalities in pain-related regions of the human brain.

UNLABELLED: Complex regional pain syndrome (CRPS) is a chronic condition that involves significant hyperalgesia of the affected limb, typically accompanied by localized autonomic abnormalities and frequently by motor dysfunction. Although central brain systems are thought to play a role in the development and maintenance of CRPS, these systems have not been well characterized. In this study, we used structural magnetic resonance imaging to characterize differences in gray matter volume between patients with right upper extremity CRPS and matched controls. Analyses were carried out using a whole brain voxel-based morphometry approach. The CRPS group showed decreased gray matter volume in several pain-affect regions, including the dorsal insula, left orbitofrontal cortex, and several aspects of the cingulate cortex. Greater gray matter volume in CRPS patients was seen in the bilateral dorsal putamen and right hypothalamus. Correlation analyses with self-reported pain were then performed on the CRPS group. Pain duration was associated with decreased gray matter in the left dorsolateral prefrontal cortex. Pain intensity was positively correlated with volume in the left posterior hippocampus and left amygdala, and negatively correlated with the bilateral dorsolateral prefrontal cortex. Our findings demonstrate that CRPS is associated with abnormal brain system morphology, particularly pain-related sensory, affect, motor, and autonomic systems. PERSPECTIVE: This paper presents structural changes in the brains of patients with CRPS, helping us differentiate CRPS from other chronic pain syndromes and furthering our understanding of this challenging disease.

Towards a physiology-based measure of pain: patterns of human brain activity distinguish painful from non-painful thermal stimulation.

Pain often exists in the absence of observable injury; therefore, the gold standard for pain assessment has long been self-report. Because the inability to verbally communicate can prevent effective pain management, research efforts have focused on the development of a tool that accurately assesses pain without depending on self-report. Those previous efforts have not proven successful at substituting self-report with a clinically valid, physiology-based measure of pain. Recent neuroimaging data suggest that functional magnetic resonance imaging (fMRI) and support vector machine (SVM) learning can be jointly used to accurately assess cognitive states. Therefore, we hypothesized that an SVM trained on fMRI data can assess pain in the absence of self-report. In fMRI experiments, 24 individuals were presented painful and nonpainful thermal stimuli. Using eight individuals, we trained a linear SVM to distinguish these stimuli using whole-brain patterns of activity. We assessed the performance of this trained SVM model by testing it on 16 individuals whose data were not used for training. The whole-brain SVM was 81% accurate at distinguishing painful from non-painful stimuli (p<0.0000001). Using distance from the SVM hyperplane as a confidence measure, accuracy was further increased to 84%, albeit at the expense of excluding 15% of the stimuli that were the most difficult to classify. Overall performance of the SVM was primarily affected by activity in pain-processing regions of the brain including the primary somatosensory cortex, secondary somatosensory cortex, insular cortex, primary motor cortex, and cingulate cortex. Region of interest (ROI) analyses revealed that whole-brain patterns of activity led to more accurate classification than localized activity from individual brain regions. Our findings demonstrate that fMRI with SVM learning can assess pain without requiring any communication from the person being tested. We outline tasks that should be completed to advance this approach toward use in clinical settings.

Viewing pictures of a romantic partner reduces experimental pain: involvement of neural reward systems.

The early stages of a new romantic relationship are characterized by intense feelings of euphoria, well-being, and preoccupation with the romantic partner. Neuroimaging research has linked those feelings to activation of reward systems in the human brain. The results of those studies may be relevant to pain management in humans, as basic animal research has shown that pharmacologic activation of reward systems can substantially reduce pain. Indeed, viewing pictures of a romantic partner was recently demonstrated to reduce experimental thermal pain. We hypothesized that pain relief evoked by viewing pictures of a romantic partner would be associated with neural activations in reward-processing centers. In this functional magnetic resonance imaging (fMRI) study, we examined fifteen individuals in the first nine months of a new, romantic relationship. Participants completed three tasks under periods of moderate and high thermal pain: 1) viewing pictures of their romantic partner, 2) viewing pictures of an equally attractive and familiar acquaintance, and 3) a word-association distraction task previously demonstrated to reduce pain. The partner and distraction tasks both significantly reduced self-reported pain, although only the partner task was associated with activation of reward systems. Greater analgesia while viewing pictures of a romantic partner was associated with increased activity in several reward-processing regions, including the caudate head, nucleus accumbens, lateral orbitofrontal cortex, amygdala, and dorsolateral prefrontal cortex--regions not associated with distraction-induced analgesia. The results suggest that the activation of neural reward systems via non-pharmacologic means can reduce the experience of pain.