Accuracy of an artificial intelligence as a medical device as part of a UK-based skin cancer teledermatology service.

Introduction: An artificial intelligence as a medical device (AIaMD), built on convolutional neural networks, has demonstrated high sensitivity for melanoma. To be of clinical value, it needs to safely reduce referral rates. The primary objective of this study was to demonstrate that the AIaMD had a higher rate of correctly classifying lesions that did not need to be referred for biopsy or urgent face-to-face dermatologist review, compared to teledermatology standard of care (SoC), while achieving the same sensitivity to detect malignancy. Secondary endpoints included the sensitivity, specificity, positive and negative predictive values, and number needed to biopsy to identify one case of melanoma or squamous cell carcinoma (SCC) by both the AIaMD and SoC. Methods: This prospective, single-centre, single-arm, masked, non-inferiority, adaptive, group sequential design trial recruited patients referred to a teledermatology cancer pathway (clinicaltrials.gov NCT04123678). Additional dermoscopic images of each suspicious lesion were taken using a smartphone with a dermoscopic lens attachment. The images were assessed independently by a consultant dermatologist and the AIaMD. The outputs were compared with the final histological or clinical diagnosis. Results: A total of 700 patients with 867 lesions were recruited, of which 622 participants with 789 lesions were included in the per-protocol (PP) population. In total, 63.3% of PP participants were female; 89.0% identified as white, and the median age was 51 (range 18-95); and all Fitzpatrick skin types were represented including 25/622 (4.0%) type IV-VI skin. A total of 67 malignant lesions were identified, including 8 diagnosed as melanoma. The AIaMD sensitivity was set at 91 and 92.5%, to match the literature-defined clinician sensitivity (91.46%) as closely as possible. In both settings, the AIaMD identified had a significantly higher rate of identifying lesions that did not need a biopsy or urgent referral compared to SoC (p-value = 0.001) with comparable sensitivity for skin cancer. Discussion: The AIaMD identified significantly more lesions that did not need to be referred for biopsy or urgent face-to-face dermatologist review, compared to teledermatologists. This has the potential to reduce the burden of unnecessary referrals when used as part of a teledermatology service.

Effectiveness of an image analyzing AI-based Digital Health Technology to identify Non-Melanoma Skin Cancer and other skin lesions: results of the DERM-003 study.

Introduction: Identification of skin cancer by an Artificial Intelligence (AI)-based Digital Health Technology could help improve the triage and management of suspicious skin lesions. Methods: The DERM-003 study (NCT04116983) was a prospective, multi-center, single-arm, masked study that aimed to demonstrate the effectiveness of an AI as a Medical Device (AIaMD) to identify Squamous Cell Carcinoma (SCC), Basal Cell Carcinoma (BCC), pre-malignant and benign lesions from dermoscopic images of suspicious skin lesions. Suspicious skin lesions that were suitable for photography were photographed with 3 smartphone cameras (iPhone 6S, iPhone 11, Samsung 10) with a DL1 dermoscopic lens attachment. Dermatologists provided clinical diagnoses and histopathology results were obtained for biopsied lesions. Each image was assessed by the AIaMD and the output compared to the ground truth diagnosis. Results: 572 patients (49.5% female, mean age 68.5 years, 96.9% Fitzpatrick skin types I-III) were recruited from 4 UK NHS Trusts, providing images of 611 suspicious lesions. 395 (64.6%) lesions were biopsied; 47 (11%) were diagnosed as SCC and 184 (44%) as BCC. The AIaMD AUROC on images taken by iPhone 6S was 0.88 (95% CI: 0.83-0.93) for SCC and 0.87 (95% CI: 0.84-0.91) for BCC. For Samsung 10 the AUROCs were 0.85 (95% CI: 0.79-0.90) and 0.87 (95% CI, 0.83-0.90), and for the iPhone 11 they were 0.88 (95% CI, 0.84-0.93) and 0.89 (95% CI, 0.86-0.92) for SCC and BCC, respectively. Using pre-determined diagnostic thresholds on images taken on the iPhone 6S the AIaMD achieved a sensitivity and specificity of 98% (95% CI, 88-100%) and 38% (95% CI, 33-44%) for SCC; and 94% (95% CI, 90-97%) and 28% (95 CI, 21-35%) for BCC. All 16 lesions diagnosed as melanoma in the study were correctly classified by the AIaMD. Discussion: The AIaMD has the potential to support the timely diagnosis of malignant and premalignant skin lesions.

The Spatiotemporal Dynamics of Cerebral Autoregulation in Functional Magnetic Resonance Imaging.

The thigh-cuff release (TCR) maneuver is a physiological challenge that is widely used to assess dynamic cerebral autoregulation (dCA). It is often applied in conjunction with Transcranial Doppler ultrasound (TCD), which provides temporal information of the global flow response in the brain. This established method can only yield very limited insights into the regional variability of dCA, whereas functional MRI (fMRI) has the ability to reveal the spatial distribution of flow responses in the brain with high spatial resolution. The aim of this study was to use whole-brain blood-oxygenation-level-dependent (BOLD) fMRI to characterize the spatiotemporal dynamics of the flow response to the TCR challenge, and thus pave the way toward mapping dCA in the brain. We used a data driven approach to derive a novel basis set that was then used to provide a voxel-wise estimate of the TCR associated haemodynamic response function (HRF TCR ). We found that the HRF TCR evolves with a specific spatiotemporal pattern, with gray and white matter showing an asynchronous response, which likely reflects the anatomical structure of cerebral blood supply. Thus, we propose that TCR challenge fMRI is a promising method for mapping spatial variability in dCA, which will likely prove to be clinically advantageous.

Measuring Arterial Pulsatility With Dynamic Inflow Magnitude Contrast.

The pulsatility of blood flow through cerebral arteries is clinically important, as it is intrinsically associated with cerebrovascular health. In this study we outline a new MRI approach to measuring the real-time pulsatile flow in cerebral arteries, which is based on the inflow phenomenon associated with fast gradient-recalled-echo acquisitions. Unlike traditional phase-contrast techniques, this new method, which we dub dynamic inflow magnitude contrast (DIMAC), does not require velocity-encoding gradients as sensitivity to flow velocity is derived purely from the inflow effect. We achieved this using a highly accelerated single slice EPI acquisition with a very short TR (15 ms) and a 90° flip angle, thus maximizing inflow contrast. We simulate the spoiled GRE signal in the presence of large arteries and perform a sensitivity analysis. The sensitivity analysis demonstrates that in the regime of high inflow contrast, DIMAC shows much greater sensitivity to flow velocity over blood volume changes. We support this theoretical prediction with in-vivo data collected in two separate experiments designed to demonstrate the utility of the DIMAC signal contrast. We perform a hypercapnia challenge experiment in order to experimentally modulate arterial tone within subjects, and thus modulate the arterial pulsatile flow waveform. We also perform a thigh-cuff release challenge, designed to induce a transient drop in blood pressure, and demonstrate that the continuous DIMAC signal captures the complex transient change in the pulsatile and non-pulsatile components of flow. In summary, this study proposes a new role for a well-established source of MR image contrast and demonstrates its potential for measuring both steady-state and dynamic changes in arterial tone.

Corrigendum: Cerebral Autoregulation Evidenced by Synchronized Low Frequency Oscillations in Blood Pressure and Resting-State fMRI.

[This corrects the article DOI: 10.3389/fnins.2019.00433.].

Cerebral Autoregulation Evidenced by Synchronized Low Frequency Oscillations in Blood Pressure and Resting-State fMRI.

Resting-state functional magnetic resonance imaging (rs-fMRI) is a widely used technique for mapping the brain's functional architecture, so delineating the main sources of variance comprising the signal is crucial. Low frequency oscillations (LFO) that are not of neural origin, but which are driven by mechanisms related to cerebral autoregulation (CA), are present in the blood-oxygenation-level-dependent (BOLD) signal within the rs-fMRI frequency band. In this study we use a MR compatible device (Caretaker, Biopac) to obtain a non-invasive estimate of beat-to-beat mean arterial pressure (MAP) fluctuations concurrently with rs-fMRI at 3T. Healthy adult subjects (n = 9; 5 male) completed two 20-min rs-fMRI scans. MAP fluctuations were decomposed into different frequency scales using a discrete wavelet transform, and oscillations at approximately 0.1 Hz show a high degree of spatially structured correlations with matched frequency fMRI fluctuations. On average across subjects, MAP fluctuations at this scale of the wavelet decomposition explain ∼2.2% of matched frequency fMRI signal variance. Additionally, a simultaneous multi-slice multi-echo acquisition was used to collect 10-min rs-fMRI at three echo times at 7T in a separate group of healthy adults (n = 5; 5 male). Multiple echo times were used to estimate the R2 ∗ decay at every time point, and MAP was shown to strongly correlate with this signal, which suggests a purely BOLD (i.e., blood flow related) origin. This study demonstrates that there is a significant component of the BOLD signal that has a systemic physiological origin, and highlights the fact that not all localized BOLD signal changes necessarily reflect blood flow supporting local neural activity. Instead, these data show that a proportion of BOLD signal fluctuations in rs-fMRI are due to localized control of blood flow that is independent of local neural activity, most likely reflecting more general systemic autoregulatory processes. Thus, fMRI is a promising tool for studying flow changes associated with cerebral autoregulation with high spatial resolution.

Developmental changes of GABA immunoreactivity in cortico-thalamic networks of an absence seizure model.

Absence seizures (ASs) are associated with abnormalities in gamma-aminobutyric acid (GABA) neurotransmission in the thalamus and the cortex. In the present study, we used light microscopy GABA immunocytochemistry to quantify the GABA-immunoreactive (GABA-IR) neurons and neuropil in the thalamic ventral basal (VB) nucleus, the nucleus reticularis thalami (NRT), the dorsal lateral geniculate (dLGN), the primary motor cortex (M1) and perioral region of the somatosensory cortex (S1po) of genetic absence epilepsy rats from Strasbourg (GAERS). We used both the relative non-epileptic control (NEC) and normal Wistar rats as control strains, and investigated GABA immunostaining at postnatal day 15 (P15), P25, and P90. The main findings were i) an increase in GABA-IR neuropil in the VB at P25 and P90 in GAERS but not in NEC and Wistar rats; ii) an increase in NRT GABA-IR neurons in GAERS and NEC, but not Wistar, rats at both P25 and P90; and iii) an increase in GABA-IR neuron density in S1po of GAERS at P25 and P90 and in Wistar at P90. These results indicate that the increased GABAergic innervation in the VB at P25 most likely contributes to the enhanced tonic inhibition observed in GAERS prior to AS onset, whereas the lack of any anatomo-morphological GABAergic differences in GAERS S1po suggests that functional more than structural abnormalities underlie the origin of cortical paroxysms in S1po of this AS model. This article is part of the "Special Issue Dedicated to Norman G. Bowery".

Cortical drive and thalamic feed-forward inhibition control thalamic output synchrony during absence seizures.

Behaviorally and pathologically relevant cortico-thalamo-cortical oscillations are driven by diverse interacting cell-intrinsic and synaptic processes. However, the mechanism that gives rise to the paroxysmal oscillations of absence seizures (ASs) remains unknown. Here we report that, during ASs in behaving animals, cortico-thalamic excitation drives thalamic firing by preferentially eliciting tonic rather than T-type Ca 2+ channel (T-channel)-dependent burst firing in thalamocortical (TC) neurons and by temporally framing thalamic output via feedforward reticular thalamic (NRT)-to-TC neuron inhibition. In TC neurons, overall ictal firing was markedly reduced and bursts rarely occurred. Moreover, blockade of T-channels in cortical and NRT neurons suppressed ASs, but such blockade in TC neurons had no effect on seizures or on ictal thalamic output synchrony. These results demonstrate ictal bidirectional cortico-thalamic communications and provide the first mechanistic understanding of cortico-thalamo-cortical network firing dynamics during ASs in behaving animals.

Differential Effect of APOE Alleles on Brain Glucose Metabolism in Targeted Replacement Mice: An [18F]FDG-µPET Study.

BACKGROUND: The Apolipoprotein E (ApoE) alleles ɛ2, ɛ3, and ɛ4 are known to differentially modulate cerebral glucose metabolism and the risk for Alzheimer's disease (AD) via both amyloid-ß (Aß)-dependent and independent mechanisms. OBJECTIVE: We investigated the influence of ApoE on cerebral glucose metabolism in humanized APOE Targeted Replacement (TR) mice at ages that precede the comparison of Aß parenchymal deposits in APOE4-TR mice. METHODS: Fludeoxyglucose ([18F]FDG) positron emission tomography (PET) measures were performed longitudinally in homozygous APOE-TR mice (APOE2, APOE3, APOE4; n = 10 for each group) at 3, 5, 11, and 15 months. Results were quantified using standard uptake values and analyzed statistically using a linear mixed effects model. Levels of the Aß40 and Aß42 peptides were quantified ex vivo using enzyme-linked immunosorbent assay (ELISA) at 15 months in the same animals. RESULTS: APOE2 mice (versus APOE3) showed a significant increase in glucose metabolism starting at 6 months, peaking at 9 months. No evidence of hypometabolism was apparent in any region or time point for APOE4 mice, which instead displayed a hypermetabolism at 15 months. Whole brain soluble Aß40 and Aß42 levels were not significantly different between genotypes at 15 months. CONCLUSIONS: Introduction of human APOE alleles ɛ2 and ɛ4 is sufficient to produce alterations in brain glucose metabolism in comparison to the control allele ɛ3, without a concomitant alteration in Aß40 and Aß42 levels. These results suggest novel Aß-independent metabolic phenotypes conferred by ɛ2 and ɛ4 alleles and have important implications for preclinical studies using TR-mice.

Role for serotonin2A (5-HT2A) and 2C (5-HT2C) receptors in experimental absence seizures.

Absence seizures (ASs) are the hallmark of childhood/juvenile absence epilepsy. Monotherapy with first-line anti-absence drugs only controls ASs in 50% of patients, indicating the need for novel therapeutic targets. Since serotonin family-2 receptors (5-HT2Rs) are known to modulate neuronal activity in the cortico-thalamo-cortical loop, the main network involved in AS generation, we investigated the effect of selective 5-HT2AR and 5-HT2CR ligands on ASs in the Genetic Absence Epilepsy Rats from Strasbourg (GAERS), a well established polygenic rat model of these non-convulsive seizures. GAERS rats were implanted with fronto-parietal EEG electrodes under general anesthesia, and their ASs were later recorded under freely moving conditions before and after intraperitoneal administration of various 5-HT2AR and 5-HT2CR ligands. The 5-HT2A agonist TCB-2 dose-dependently decreased the total time spent in ASs, an effect that was blocked by the selective 5-HT2A antagonist MDL11,939. Both MDL11,939 and another selective 5-HT2A antagonist (M100,907) increased the length of individual seizures when injected alone. The 5-HT2C agonists lorcaserin and CP-809,101 dose-dependently suppressed ASs, an effect blocked by the selective 5-HT2C antagonist SB 242984. In summary, 5-HT2ARs and 5-HT2CRs negatively control the expression of experimental ASs, indicating that selective agonists at these 5-HT2R subtypes might be potential novel anti-absence drugs.

A critical evaluation of the gamma-hydroxybutyrate (GHB) model of absence seizures.

Typical absence seizures (ASs) are nonconvulsive epileptic events which are commonly observed in pediatric and juvenile epilepsies and may be present in adults suffering from other idiopathic generalized epilepsies. Our understanding of the pathophysiological mechanisms of ASs has been greatly advanced by the availability of genetic and pharmacological models, in particular the γ-hydroxybutyrate (GHB) model which, in recent years, has been extensively used in studies in transgenic mice. GHB is an endogenous brain molecule that upon administration to various species, including humans, induces not only ASs but also a state of sedation/hypnosis. Analysis of the available data clearly indicates that only in the rat does there exist a set of GHB-elicited behavioral and EEG events that can be confidently classified as ASs. Other GHB activities, particularly in mice, appear to be mostly of a sedative/hypnotic nature: thus, their relevance to ASs requires further investigation. At the molecular level, GHB acts as a weak GABA-B agonist, while the existence of a GHB receptor remains elusive. The pre- and postsynaptic actions underlying GHB-elicited ASs have been thoroughly elucidated in thalamus, but little is known about the cellular/network effects of GHB in neocortex, the other brain region involved in the generation of ASs.