1. Solomon AJ, Naismith RT, Cross AH. Misdiagnosis of multiple sclerosis. Neurology. 2019;92(1):26-33. doi:10.1212/WNL.0000000000006583.
The expert panel that formulated the 2017 McDonald criteria acknowledged that the current criteria “were not developed to differentiate MS from other conditions” but rather to facilitate earlier diagnosis of MS in patients presenting with typical demyelinating syndromes. The use of McDonald criteria in atypical syndromes, or any clinical presentations other than optic neuritis, brainstem/cerebellar syndromes, or transverse myelitis, diminishes accuracy.
In patients with a history of migraine, vascular risk factors, or examination findings suggestive of a functional neurologic disorder who fulfill McDonald criteria, evaluation for CSF OCB or spinal cord lesions should be pursued to support the diagnosis of MS. MRI criteria using a lesion threshold of 6 mm may improve specificity for MS in atypical syndromes and older patients, as even healthy controls can have T2 abnormalities in the 2–4 mm range. The identification of callosal lesions may help to differentiate MRI demyelination from vascular changes. The diagnosis of MS, especially in atypical syndromes, should not rely on questionable spinal cord lesions observed only on sagittal view. Lesions should be confirmed on axial images and on at least 2 different MRI sequences (such as proton density, T2, or short tau inversion recovery). In equivocal cases, repeat imaging may be necessary to confirm a spinal cord lesion. Repeating CSF evaluation may also be prudent in patients with atypical or challenging syndromes.
Recommendations for prevention of multiple sclerosis misdiagnosis when applying 2017 McDonald criteria:
MRI lesions and their characteristics:
1. Juxtacortical lesions must abut the cortex, without intervening white matter
2. Periventricular lesions must abut the ventricles, without intervening white matter
3. Lesions should be 3 mm or larger in diameter
4. Small punctate lesions should not be used to fulfill MRI criteria
4. Use of intracortical and subpial cortical lesions to fulfill criteria should be restricted to experienced imaging centers
Symptomatic MRI lesions for fulfillment of DIS and DIT:
In patients with monophasic syndrome of a single symptomatic brainstem or spinal cord lesion where only 1 additional MRI DIS region is satisfied, you should consider awaiting appearance of an additional MRI lesion or additional clinical event to meet DIS criteria, especially when comorbidities are present.
3 Tables
2. Kaunzner UW, Kang Y, Zhang S, et al. Quantitative susceptibility mapping identifies inflammation in a subset of chronic multiple sclerosis lesions. Brain. 2019;142(1):133-145. doi:10.1093/brain/awy296.
Multiple sclerosis is an inflammatory disease of the CNS, characterized by loss of myelin, leading to axonal damage and subsequent neurodegeneration. Both the adaptive and innate immune system play an important role in the pathophysiology of the disease, and contribute to focal inflammation and demyelination, leading to the formation of multiple sclerosis lesions.
Current imaging techniques can detect accumulation of gadolinium in regions of acute lesion formation, indicating acute inflammation; however, no tool is available to visualize ongoing inflammation in chronic lesions, such as continuous microglial inflammation behind an intact blood–brain barrier.
MRI with a gradient echo (GRE) sequence is sensitive to iron and has been explored by many investigators to detect an iron rim in chronic active multiple sclerosis lesions. In a post-mortem study of predominantly progressive patients, provided histopathological evidence that the presence of a phase rim within chronic lesions, which is derived from GRE data, is representative of iron-laden activated microglia and macrophages. Quantitative susceptibility mapping (QSM) provides an effective means to directly map the distribution of susceptibility sources and allows the most accurate quantification and localization of brain iron as compared to other GRE approaches.
PET in combination with the ligand 11C-PK11195 is used to evaluate activated microglia/macrophages in vivo. PK11195-PET binds to the 18 kDa translocator protein (TSPO), which is expressed on the outer mitochondria membrane of activated microglia/macrophages. Studies with PK11195-PET have confirmed the most prominent uptake within areas of increased inflammation, especially in contrast-enhancing lesions, but has also demonstrated uptake within chronic multiple sclerosis lesions wherein cross-binding with reactive astrocytes has been demonstrated, suggesting that TSPO expression may be reflective of a diffuse glial inflammatory response. While PET provides a higher specificity as compared to MRI, limitations of spatial resolution may impact specific localization of molecular binding within smaller regions of interest.
This study aimed to validate that lesions with a hyperintense rim on quantitative susceptibility mapping from both relapsing and progressive patients demonstrate a higher level of innate immune activation as measured on 11C-PK11195 PET. Thirty patients were enrolled in this study, 24 patients had relapsing remitting multiple sclerosis, six had progressive multiple sclerosis, and all patients had concomitant MRI with a gradient echo sequence and PET with 11C-PK11195. A total of 406 chronic lesions were detected, and 43 chronic lesions with a hyperintense rim on quantitative susceptibility mapping were identified as rim + lesions. Susceptibility was higher in rim + compared to rim – lesions. Among rim + lesions, susceptibility within the rim was significantly higher compared to the core, consistent with the presence of iron. In a mixed-effects model, 11CPK11195 uptake, representing activated microglia/macrophages, was higher in rim + lesions compared to rim – lesions. Validating the in vivo imaging results, multiple sclerosis brain slabs were imaged with quantitative susceptibility mapping and processed for immunohistochemistry. These results showed a positive translocator protein signal throughout the expansive hyperintense border of rim + lesions, which co-localized with iron containing CD68 + microglia and macrophages.
In conclusion, this study provides evidence that suggests that a hyperintense rim on quantitative susceptibility measure within a chronic lesion is a correlate for persistent inflammatory activity and that these lesions can be identified in the relapsing patients. Utilizing quantitative susceptibility measure to differentiate chronic multiple sclerosis lesion subtypes, especially chronic active lesions, would provide a method to assess the impact of these lesions on disease progression.
5 Figures
3. Kim J-T, Cho B-H, Choi K-H, et al. Magnetic Resonance Imaging Versus Computed Tomography Angiography Based Selection for Endovascular Therapy in Patients With Acute Ischemic Stroke. Stroke. 2019;50:1-8. doi:10.1161/STROKEAHA.118.023173.
This study aimed to elucidate whether MRI-based selection for endovascular therapy is a safe and effective tool within and after the 6-hour time window compared with conventional CTA-based selection.
Data from a prospective, nationwide, multicenter stroke registry were analyzed. Workflow timelines were compared between patients selected for endovascular therapy based on MRI (the MRI group) and CTA (the CTA group). Multivariable ordinal and binary logistic regression analyses were performed to explore the relationships between decision imaging for endovascular therapy and clinical outcomes, including good and excellent outcomes (modified Rankin Scale scores of 0–2 and 0–1, respectively) at 3-month, modified Rankin Scale score distributions and safety outcomes (symptomatic intracranial hemorrhage [SICH] and mortality).
In the analysis of over 1200 patients treated with endovascular therapy from a nationwide multicenter registry, MRI-based selection led to substantial delay in image acquisition but comparable workflows after image acquisition compared with CTA-based selection. The MRI-based approach, compared with the CTA based approach, was not associated with the improvement of functional outcomes, but patients who were selected by the MRI-based approach were less likely to develop symptomatic ICH after endovascular therapy than the patients who were selected by the CTA-based approach. However, because significant imbalances were observed between the CTA and MRI groups, the results of our study should be interpreted with caution.
4. Esteva A, Robicquet A, Ramsundar B, et al. A guide to deep learning in healthcare. Nat Med. 2019;25(January). doi:10.1038/s41591-018-0316-z.
ML is distinct from other types of computer programming in that it transforms the inputs of an algorithm into outputs using statistical, data-driven rules that are automatically derived from a large set of examples, rather than being explicitly specified by humans. Constructing a ML system required domain expertise and human engineering to design feature extractors that transformed raw data into suitable representations from which a learning algorithm could detect patterns. In contrast, deep learning is a form of representation learning—in which a machine is fed with raw data and develops its own representations needed for pattern recognition—that is composed of multiple layers of representations. These layers are typically arranged sequentially and composed of a large number of primitive, nonlinear operations, such that the representation of one layer (beginning with the raw data input) is fed into the next layer and transformed into a more abstract representation. As data flows through the layers of the system, the input space becomes iteratively warped until data points become distinguishable. In this manner, highly complex functions can be learned.
Some of the greatest successes of deep learning have been in the field of computer vision (CV). CV focuses on image and video understanding, and deals with tasks such as object classification, detection, and segmentation—which are useful in determining whether a patient’s radiograph contains malignant tumors. Convolutional neural networks (CNNs) a type of deep-learning algorithm designed to process data that exhibits natural spatial invariance (e.g., images, whose meanings do not change under translation), have grown to be central in this field. Medical imaging, for instance, can greatly benefit from recent advances in image classification and object detection. Many studies have demonstrated promising results in complex diagnostics spanning dermatology, radiology, ophthalmology, and pathology. Deep-learning systems could aid physicians by offering second opinions and flagging concerning areas in images.
Natural language processing (NLP) focuses on analyzing text and speech to infer meaning from words. Recurrent neural networks (RNNs)—deep learning algorithms effective at processing sequential inputs such as language, speech, and time-series data—play an important role in this field. Notable successes of Natural language processing include machine translation, text generation, and image captioning. In healthcare, sequential deep learning and language technologies power applications within domains such as electronic health records (EHRs). EHRs are rapidly becoming ubiquitous. The EHR of a large medical organization can capture the medical transactions of over 10 million patients throughout the course of a decade. A single hospitalization alone typically generates ∼ 150,000 pieces of data. The potential benefits derived from this data are significant. In aggregate, an EHR of this scale represents 200,000 years of doctor wisdom and 100 million years of patient outcome data, covering a plethora of rare conditions and maladies. As such, application of deep-learning methods to EHR data is a rapidly expanding area.
Reinforcement learning (RL) refers to a class of techniques designed to train computational agents to successfully interact with their environment, typically to achieve specific goals. This learning can happen through trial and error, through demonstration, or through a hybrid approach. As an agent takes actions within its environment, an iterative feedback loop of reward and consequence trains the agent to better accomplish the goals at hand. Learning from expert demonstration is accomplished either by learning to predict the expert’s actions directly via supervised learning (i.e., imitation learning) or by inferring the expert’s objective (i.e., inverse Reinforcement learning). To successfully train an agent, it is critical to have a model function that can take as input sensory signals from the environment and output the next actions for the agent to take. Deep Reinforcement learning, in which a deep-learning model serves as the model function, shows promise. One healthcare domain that can benefit from deep Reinforcement learning is robotic-assisted surgery. Currently, robotic-assisted surgery largely depends on a surgeon guiding a robot’s instruments in a teleoperated fashion. Deep learning can enhance the robustness and adaptability of robotic-assisted surgery by using computer vision models (e.g., Convolutional neural networks) to perceive surgical environments and reinforcement learning methods to learn from a surgeon’s physical motions.
4 figures
5. Khan S, Mohammad Amin F, Emil Christensen C, et al. Meningeal contribution to migraine pain: A magnetic resonance angiography study. Brain. 2019;142(1):93-102. doi:10.1093/brain/awy300.
Migraine is generally considered a neurovascular headache with intricate pathophysiological connections between deep brain structures and the trigeminal pain pathways. While the brain is largely insensate, stimulation of dura mater near its arteries in humans produces pain that resembles migraine headache as well as migraine-like associated symptoms of nausea and photophobia. In conjunction with this finding, preclinical models have further suggested that activation of trigeminal nerve fibers innervating the dura mater may play a key role in migraine pain. This concept is based on assays that have demonstrated that stimulating the dura mater or trigeminal ganglion with inflammatory, chemical or electrical mediators results in plasma protein extravasation, mast cell degranulation, and dilation of meningeal blood vessels.
In the present study, the authors applied high resolution 3 T magnetic resonance angiography (MRA) to investigate the early circumference change of cranial arteries in unilateral attacks of migraine without aura. The authors used the highly effective phosphodiesterase- 3-inhibitor cilostazol as an experimental migraine trigger. They hypothesized that during unilateral attacks of migraine without aura, intra- and extracranial arteries exhibit ipsilateral dilation, reflecting activation of perivascular nociceptors. They also hypothesized that therapeutic use of the selective migraine abortive drug, sumatriptan, a 5-hydroxytryptamine agonist, would reverse this activation.
Thirty patients underwent magnetic resonance angiography scans, of which 26 patients developed unilateral attacks of migraine without aura and were included in the final analysis. Eleven patients treated their migraine with sumatriptan while the remaining 15 patients did not treat their attacks with analgesics or triptans. At migraine onset, only the middle meningeal artery exhibited greater circumference increase on the pain side compared to the non-pain side. None of the remaining arteries revealed any pain-side specific changes in circumference but exhibited bilateral dilation. Sumatriptan constricted all extracerebral arteries. In the late phase of migraine, they found sustained bilateral dilation of the middle meningeal artery. In conclusion, onset of migraine is associated with increase in middle meningeal artery circumference specific to the head pain side. Their findings suggest that vasodilation of the middle meningeal artery may be a surrogate marker for activation of dural perivascular nociceptors, indicating a meningeal site of migraine headache.
4 figures, 4 tables, no images
6. Zumsteg ZS, Luu M, Kim S, et al. Quantitative lymph node burden as a ‘very-high-risk’ factor identifying head and neck cancer patients benefiting from postoperative chemoradiation. Ann Oncol. 2018;(November 2018):76-84. doi:10.1093/annonc/mdy490.
Postoperative chemoradiation (CRT) is considered a standard of care for head and neck squamous cell carcinoma (HNSCC) patients with positive margins or extranodal extension (ENE) following surgery. This is based on a post hoc combined analysis of two randomized trials that showed improved locoregional control and overall survival (OS) with postoperative CRT versus radiotherapy (RT) alone only in this subgroup of patients. However, CRT also significantly increases toxicity, so further refining which patients are most likely to benefit from multimodality adjuvant therapy is important for optimizing the risks and benefits of treatment.
Recent studies have demonstrated that the number of metastatic cervical lymph nodes (LNs) is the most important factor influencing survival in non-oropharyngeal HNSCC, far exceeding the impact of classic high-risk factors like extranodal extension and positive margins. Thus, it is possible that nodal burden also represents the best indicator of which patients will benefit from intensified adjuvant therapy in non-oropharyngeal HNSCC. In this study, the authors evaluate the interaction between quantitative metastatic LN burden and postoperative CRT with respect to survival in oral cavity, laryngeal, and hypopharyngeal squamous cell carcinoma. They compare the utility of metastatic LN burden, versus other pathologic factors, for determining which patients benefit from postoperative CRT using a large national dataset.
Deidentified patient data were obtained through the National Cancer Database (NCDB), a hospital-based registry maintained by the Commission on Cancer of the American College of Surgeons and the American Cancer Society. The NCDB collects cases from >1500 facilities encompassing approximately 70% of newly diagnosed cancers in the United States. In total, 7144 patients met inclusion criteria, including 2465 undergoing postoperative CRT and 4679 undergoing postoperative RT.
In patients with non-oropharyngeal HNSCC undergoing surgical resection and postoperative radiation, they observed an increasing survival benefit from concomitant CRT as the number of positive LNs increased. Improved survival with postoperative CRT was noted beginning at three positive LN, with most of the benefit seen in the ‘very-high-risk’ group of patients with >6 positive LNs. This cohort, representing 8% of patients in this study, had a 35% decrease in the relative risk of death with CRT versus RT following surgery after adjusting for other covariables, corresponding to a nearly 23% absolute improvement in 3-year OS in propensity score-matched cohorts. An identical pattern, with CRT only benefitting patients with high quantitative nodal burden, was observed even when limiting analysis to the subgroup of patients with positive margins and/or extranodal extension, a cohort of patients who are uniformly recommended to undergo postoperative CRT in national guidelines. The results support quantitative nodal burden as the single most powerful predictor identifying patients that will have improved survival outcomes with intensification of adjuvant therapy.
4 figures, no images
7. Katsanos AH, Tsivgoulis G. Is intravenous thrombolysis still necessary in patients who undergo mechanical thrombectomy? Curr Opin Neurol. 2019;32(1):3-12. doi:10.1097/WCO.0000000000000633.
During the last 3 years, emerging evidence from independent randomized-controlled clinical trials highlighted mechanical thrombectomy as a well-tolerated and highly effective treatment for AIS patients with LVOs in the anterior circulation, which is associated with higher rates of successful reperfusion and more favorable functional outcomes compared with standard therapy with or without IVT. All rt-PA eligible patients randomized to endovascular reperfusion therapies in these RCTs were pretreated with IVT prior to the initiation of mechanical thrombectomy, while mechanical thrombectomy also improved neurological outcomes in patients with contraindications to IVT when compared with standard therapy.
The substantial improvement in 3-month functional outcomes (NNT ranging between 2.5 and 7) in LVO patients treated with mechanical thrombectomy compared with standard therapy combined with the poor response of proximal intracranial occlusions (especially terminal internal carotid and proximal middle cerebral artery) to IVT monotherapy, resulted in questioning the utility of IVT pretreatment in LVO patients undergoing mechanical thrombectomy. In view of the former considerations, the authors sought to review and summarize all available evidence comparing the safety and efficacy of bridging therapy (bridging therapy: IVT and mechanical thrombectomy) with direct mechanical thrombectomy (dMT). Despite the theoretical concern of a higher bleeding risk with IVT pretreatment, no published study or meta-analysis to date has reported evidence of increased symptomatic intracerebral hemorrhage (sICH risk) with bridging therapy compared with direct mechanical thrombectomy (dMT). Taking into account existing literature evidence, the authors consider that IVT and mechanical thrombectomy should be regarded as two highly effective and complementary reperfusion therapies.
The findings of the present review corroborate the recommendations of current American Heart Association/ American Stroke Association guidelines advocating that IVT and mechanical thrombectomy should be regarded as two highly effective and complementary reperfusion therapies. Both IVT and mechanical thrombectomy should be used synergistically for the optimal treatment of AIS patients with LVO. All AIS patients with LVO and no contraindications to IVT should receive promptly rt-PA bolus followed by immediate transfer to the angio-suite for prompt initiation of mechanical thrombectomy, unless future randomized controlled trials directly comparing bridging therapy with dMT provide evidence to proceed differently.
8. Ganesh A, Wong JH, Menon BK. Practice Current: How do you manage patients with a “hot carotid”? Bartolini L, ed. Neurol Clin Pract. 2018;8(6):527-536. doi:10.1212/CPJ.0000000000000562.
Carotid atherosclerosis causes about 15%–20% of ischemic stroke and TIAs, primarily via thromboembolism from ruptured plaque and sometimes from low-flow states. Carotid endarterectomy (CEA), or the alternative of carotid angioplasty and stenting (CAS), are used for revascularization of the symptomatic extracranial carotid artery. A meta-analysis of data from the European Carotid Surgery Trial (ECST) and North American Symptomatic Carotid Endarterectomy Trial (NASCET) demonstrated that the benefit of CEA is time-sensitive: greatest when performed within 2 weeks of the last ischemic event and falling rapidly with increasing delay. This urgency of treatment underscores the high upfront risk of recurrent strokes in patients presenting with acutely symptomatic carotid stenosis (a “hot carotid,” defined pragmatically for the purpose of this report as a recent stroke/TIA [within hours–days of symptom onset] and thought to be etiologically related to the carotid artery stenosis).
Questions were posed to experts from 3 different continents, representing differing medical systems and patient populations. The authors summarized their responses which address their preferred imaging and revascularization approaches for patients with a hot carotid, their preferred antithrombotic regimens in these patients, and clinical or imaging characteristics that influence their management decisions. Similar questions were posed to the rest of the readership in an online survey.
The majority of survey-takers practiced in a hospital-based setting (82%) outside of the United States (73%). Consistent with previous surveys, the most represented countries were India (10%), Brazil (10%), Spain (9%), Germany (7%), and Italy (7%).
The clinical case was a 65-year-old man presenting with acute onset of right-sided hemiparesis lasting several hours who is found to have 80% left-sided extracranial internal carotid artery stenosis. For vascular imaging, responders chose CTA (71%) and Doppler ultrasound (57%), while a minority opted for MRA (14%) or digital subtraction angiography (11%). While CEA was the procedure of choice for 69% of survey takers, almost one-third chose CAS, primarily outside of the United States.
Uncertainty prevailed, both within and outside the United States, regarding the timing for carotid revascularization: 38% chose 3–7 days, 36% more than 7 days, and 27% 2–3 days. While waiting for the procedure, most respondents chose aspirin (44% low-dose and 42% high-dose) and clopidogrel (44%), either alone or in combination. The majority of survey takers opted for heparin (36%), clopidogrel (25%), or low-molecular weight heparin (23%) in the presence of an intravascular thrombus.
The 3 experts interviewed all preferred CTA for primary vascular imaging in hot carotids and CEA for carotid revascularization. The 3 experts also noted the benefit of DAPT in preventing recurrent events in this population. They recognized concerns expressed by surgeons regarding increased bleeding risk in patients with DAPT undergoing CEA but were also aware of varying preferences of surgeons in this regard (some surgeons were comfortable operating on DAPT while others were not).
9. Chaturvedi S. Treatment of a hot carotid. Neurol Clin Pract. 2018;8(6):466-467. doi:10.1212/CPJ.0000000000000561.
Are the survey results rational and evidence-based? The report card is mixed. The migration to CTA is not surprising since CTA is frequently performed in the emergency department as part of an acute stroke evaluation. Many centers will utilize 2 concordant imaging studies before proceeding to carotid revascularization.
It is surprising that CAS was selected as the revascularization option by about one third of the survey respondents. In the United States, close to 80% of carotid revascularization procedures are CEA operations. In addition, a combined analysis of several trials found that among symptomatic patients who had a procedure performed within 7 days of the last symptomatic event, CAS had a notably higher stroke/death rate compared to CEA (8.3% vs 1.3%). This translates into a number needed to harm of 14. All 3 expert physicians preferred CEA.
An important topic not mentioned by Ganesh et al. is the extremely outdated evidence base on which patients with symptomatic carotid stenosis are treated. Initial results from NASCET were published in 1991. At that time, clopidogrel was not in use, high-potency statins were not available, CAS was still being refined, and management of blood pressure was suboptimal. Dr. Chaturvedi asks if there is any other area of modern neurology where patients are treated on the basis of data that are 28 years old? Some trainees were not even born in 1991, yet the ghost of NASCET marches forward in clinical practice guidelines.