Applying Machine Learning to the Visual Inspection of Filled Injectable Drug Products.

With machine learning (ML), we see the potential to better harness the intelligence and decision-making abilities of human inspectors performing manual visual inspection (MVI) and apply this to automated visual inspection (AVI) with the inherent improvements in throughput and consistency. This article is intended to capture current experience with this new technology and provides points to consider for successful application to AVI of injectable drug products. The technology is available today for such AVI applications. Machine vision companies have integrated ML as an additional visual inspection tool with minimal upgrades to existing hardware. Studies have demonstrated superior results in defect detection and reduction in false rejects, when compared with conventional inspection tools. ML implementation does not require modifications to current AVI qualification strategies. The utilization of this technology for AVI will accelerate recipe development by use of faster computers rather than by direct human configuration and coding of vision tools. By freezing the model developed with artificial intelligence tools and subjecting it to current validation strategies, assurance of reliable performance in the production environment can be achieved.

Mechanical sensitization, increased axonal excitability, and spontaneous activity in C-nociceptors after ultraviolet B irradiation in pig skin.

ABSTRACT: Ultraviolet B (UVB) irradiation induces hyperalgesia in human and animal pain models. We investigated mechanical sensitization, increase in axonal excitability, and spontaneous activity in different C-nociceptor classes after UVB in pig skin. We focused on units with receptive fields covering both irradiated and nonirradiated skin allowing intraindividual comparisons. Thirty-five pigs were irradiated in a chessboard pattern, and extracellular single-fibre recordings were obtained 10 to 28 hours later (152 fibers). Units from the contralateral hind limb served as a control (n = 112). Irradiated and nonirradiated parts of the same innervation territory were compared in 36 neurons; low threshold C-touch fibers (n = 10) and sympathetic efferents (n = 2) were unchanged, but lower mechanical thresholds and higher discharge frequency at threshold were found in mechanosensitive nociceptors (n = 12). Half of them could be activated with nonnoxious brush stimuli in the sunburn. Four of 12 mechanoinsensitive nociceptors were found sensitized to mechanical stimulation in the irradiated part of the receptive field. Activity-dependent slowing of conduction was reduced in the irradiated and in the nonirradiated skin as compared with the control leg, whereas increased ability to follow high stimulation frequencies was restricted to the sunburn (108.5 ± 37 Hz UVB vs 6.3 ± 1 Hz control). Spontaneous activity was more frequent in the sunburn (72/152 vs 31/112). Mechanical sensitization of primary nociceptors and higher maximum after frequency are suggested to contribute to primary hyperalgesia, whereas the spontaneous activity of silent nociceptors might offer a mechanistic link contributing to ongoing pain and facilitated induction of spinal sensitization.

Maximum axonal following frequency separates classes of cutaneous unmyelinated nociceptors in the pig.

KEY POINTS: C-nociceptors are generally assumed to have a low maximum discharge frequency of 10-30 Hz. However, only mechano-insensitive 'silent' C-nociceptors cannot follow electrical stimulation at 5 Hz (75 pulses) whereas polymodal C-nociceptors in the pig follow stimulation at up to 100 Hz without conduction failure. Sensitization by nerve growth factor increases the maximum following frequency of 'silent' nociceptors in pig skin and might thereby contribute in particular to intense pain sensations in chronic inflammation. A distinct class of C-nociceptors with mechanical thresholds >150 mN resembles 'silent' nociceptors at low stimulation frequencies in pigs and humans, but is capable of 100 Hz discharge and thus is suited to encode painfulness of noxious mechanical stimuli. ABSTRACT: Using extracellular single-fibre recordings from the saphenous nerve in pig in vivo, we investigated peak following frequencies (5-100 Hz) in different classes of C-nociceptors and their modulation by nerve growth factor. Classes were defined by sensory (mechano-sensitivity) and axonal characteristics (activity dependent slowing of conduction, ADS). Mechano-insensitive C-nociceptors (CMi) showed the highest ADS (34% ± 8%), followed only 66% ± 27% of 75 pulses at 5 Hz and increasingly blocked conduction at higher frequencies. Three weeks following intradermal injections of nerve growth factor, peak following frequency increased specifically in the sensitized mechano-insensitive nociceptors (20% ± 16% to 38% ± 23% response rate after 72 pulses at 100 Hz). In contrast, untreated polymodal nociceptors with moderate ADS (15.2% ± 10.2%) followed stimulation frequencies of 100 Hz without conduction failure (98.5% ± 6%). A distinct class of C-nociceptors was exclusively sensitive to strong forces above 150 mN. This class had a high ADS (27.2% ± 7.6%), but displayed almost no propagation failure even at 100 Hz stimulation (84.7% ± 17%). Also, among human mechanosensitive nociceptors (n = 153) those with thresholds above 150 mN (n = 5) showed ADS typical of silent nociceptors. C-fibres with particularly high mechanical thresholds and high following frequency form a distinct nociceptor class ideally suited to encode noxious mechanical stimulation under normal conditions when regular silent nociceptors are inactive. Sensitization by nerve growth factor increases maximum discharge frequency of silent nociceptors, thereby increasing the frequency range beyond their physiological limit, which possibly contributes to excruciating pain under inflammatory conditions.

Heat-resistant action potentials require TTX-resistant sodium channels NaV1.8 and NaV1.9.

Damage-sensing nociceptors in the skin provide an indispensable protective function thanks to their specialized ability to detect and transmit hot temperatures that would block or inflict irreversible damage in other mammalian neurons. Here we show that the exceptional capacity of skin C-fiber nociceptors to encode noxiously hot temperatures depends on two tetrodotoxin (TTX)-resistant sodium channel α-subunits: NaV1.8 and NaV1.9. We demonstrate that NaV1.9, which is commonly considered an amplifier of subthreshold depolarizations at 20°C, undergoes a large gain of function when temperatures rise to the pain threshold. We also show that this gain of function renders NaV1.9 capable of generating action potentials with a clear inflection point and positive overshoot. In the skin, heat-resistant nociceptors appear as two distinct types with unique and possibly specialized features: one is blocked by TTX and relies on NaV1.9, and the second type is insensitive to TTX and composed of both NaV1.8 and NaV1.9. Independent of rapidly gated TTX-sensitive NaV channels that form the action potential at pain threshold, NaV1.8 is required in all heat-resistant nociceptors to encode temperatures higher than ∼46°C, whereas NaV1.9 is crucial for shaping the action potential upstroke and keeping the NaV1.8 voltage threshold within reach.

Sodium Channel Nav1.8 Underlies TTX-Resistant Axonal Action Potential Conduction in Somatosensory C-Fibers of Distal Cutaneous Nerves.

Voltage-gated sodium (NaV) channels are responsible for the initiation and conduction of action potentials within primary afferents. The nine NaV channel isoforms recognized in mammals are often functionally divided into tetrodotoxin (TTX)-sensitive (TTX-s) channels (NaV1.1-NaV1.4, NaV1.6-NaV1.7) that are blocked by nanomolar concentrations and TTX-resistant (TTX-r) channels (NaV1.8 and NaV1.9) inhibited by millimolar concentrations, with NaV1.5 having an intermediate toxin sensitivity. For small-diameter primary afferent neurons, it is unclear to what extent different NaV channel isoforms are distributed along the peripheral and central branches of their bifurcated axons. To determine the relative contribution of TTX-s and TTX-r channels to action potential conduction in different axonal compartments, we investigated the effects of TTX on C-fiber-mediated compound action potentials (C-CAPs) of proximal and distal peripheral nerve segments and dorsal roots from mice and pigtail monkeys (Macaca nemestrina). In the dorsal roots and proximal peripheral nerves of mice and nonhuman primates, TTX reduced the C-CAP amplitude to 16% of the baseline. In contrast, >30% of the C-CAP was resistant to TTX in distal peripheral branches of monkeys and WT and NaV1.9-/- mice. In nerves from NaV1.8-/- mice, TTX-r C-CAPs could not be detected. These data indicate that NaV1.8 is the primary isoform underlying TTX-r conduction in distal axons of somatosensory C-fibers. Furthermore, there is a differential spatial distribution of NaV1.8 within C-fiber axons, being functionally more prominent in the most distal axons and terminal regions. The enrichment of NaV1.8 in distal axons may provide a useful target in the treatment of pain of peripheral origin.SIGNIFICANCE STATEMENT It is unclear whether individual sodium channel isoforms exert differential roles in action potential conduction along the axonal membrane of nociceptive, unmyelinated peripheral nerve fibers, but clarifying the role of sodium channel subtypes in different axonal segments may be useful for the development of novel analgesic strategies. Here, we provide evidence from mice and nonhuman primates that a substantial portion of the C-fiber compound action potential in distal peripheral nerves, but not proximal nerves or dorsal roots, is resistant to tetrodotoxin and that, in mice, this effect is mediated solely by voltage-gated sodium channel 1.8 (NaV1.8). The functional prominence of NaV1.8 within the axonal compartment immediately proximal to its termination may affect strategies targeting pain of peripheral origin.

Involvement of Spinal IL-6 Trans-Signaling in the Induction of Hyperexcitability of Deep Dorsal Horn Neurons by Spinal Tumor Necrosis Factor-Alpha.

UNLABELLED: During peripheral inflammation, both spinal TNF-α and IL-6 are released within the spinal cord and support the generation of inflammation-evoked spinal hyperexcitability. However, whether spinal TNF-α and IL-6 act independently in parallel or in a functionally dependent manner has not been investigated. In extracellular recordings from mechanonociceptive deep dorsal horn neurons of normal rats in vivo, we found that spinal application of TNF-α increased spinal neuronal responses to mechanical stimulation of knee and ankle joints. This effect was significantly attenuated by either sgp130, which blocks IL-6 trans-signaling mediated by IL-6 and its soluble receptor IL-6R (sIL-6R); by an antibody to the IL-6 receptor; or by minocycline, which inhibits the microglia. IL-6 was localized in neurons of the spinal cord and, upon peripheral noxious stimulation in the presence of spinal TNF-α, IL-6 was released spinally. Furthermore, TNF-α recruited microglial cells to provide sIL-6R, which can form complexes with IL-6. Spinal application of IL-6 plus sIL-6R, but not of IL-6 alone, enhanced spinal hyperexcitability similar to TNF-α and the inhibition of TNF-α-induced hyperexcitability by minocycline was overcome by coadministration of sIL-6R, showing that sIL-6R is required. Neither minocycline nor the TNF-α-neutralizing compound etanercept inhibited the induction of hyperexcitability by IL-6 plus sIL-6R. Together, these data show that the induction of hyperexcitability of nociceptive deep dorsal horn neurons by TNF-α largely depends on the formation of IL-6/sIL-6R complexes that are downstream of TNF-α and requires the interactions of neurons and microglia orchestrated by TNF-α. SIGNIFICANCE STATEMENT: Both spinal TNF-α and IL-6 induce a state of spinal hyperexcitability. We present the novel finding that the full effect of TNF-α on the development of spinal hyperexcitability depends on IL-6 trans-signaling acting downstream of TNF-α. IL-6 trans-signaling requires the formation of complexes of IL-6 and soluble IL-6 receptor. Spinal TNF-α furthers the release of IL-6 from neurons in the spinal cord during peripheral noxious stimulation and recruits microglial cells to provide soluble IL-6 receptor, which can form complexes with IL-6. Therefore, a specific interaction between neurons and microglia is required for the full development of TNF-α-induced hyperexcitability of nociceptive deep horsal horn neurons.

Automated detection of latency tracks in microneurography recordings using track correlation.

BACKGROUND: The marking technique in microneurography uses stimulus-induced changes in neural conduction velocity to characterize human C-fibers. Changes in conduction velocity are manifested as variations in the temporal latency between periodic electrical stimuli and the resulting APs. When successive recorded sweeps are displayed vertically in a "waterfall" format, APs correlated with the stimulus form visible vertical tracks. Automated detection of these latency tracks is made difficult by sometimes poor signal-to-noise ratio in recordings, spontaneous neural firings uncorrelated with the stimuli, and multi-unit recordings with crossing or closely parallel tracks. NEW METHOD: We developed an automated track-detection technique based on a local linearization of the latency tracks of stimulus-correlated APs. This technique enhances latency tracks, eliminates transient noise spikes and spontaneous neural activity not correlated with the stimulus, and automatically detects latency tracks across successive sweeps in a recording. RESULTS: We evaluated our method on microneurography recordings showing varying signal quality, spontaneous firing, and multiple tracks that run closely parallel and cross. The method showed excellent detection of latency tracks in all of our recordings. COMPARISON WITH EXISTING METHOD(S): We compare our method to the commonly used track detection method of Hammarberg as implemented in the Drever program. CONCLUSIONS: Our method is a robust means of automatically detecting latency tracks in typical microneurography recordings.

Three functionally distinct classes of C-fibre nociceptors in primates.

In primates, C-fibre polymodal nociceptors are broadly classified into two groups based on mechanosensitivity. Here we demonstrate that mechanically sensitive polymodal nociceptors that respond either quickly (QC) or slowly (SC) to a heat stimulus differ in responses to a mild burn, heat sensitization, conductive properties and chemosensitivity. Superficially applied capsaicin and intradermal injection of ß-alanine, an MrgprD agonist, excite vigorously all QCs. Only 40% of SCs respond to ß-alanine, and their response is only half that of QCs. Mechanically insensitive C-fibres (C-MIAs) are ß-alanine insensitive but vigorously respond to capsaicin and histamine with distinct discharge patterns. Calcium imaging reveals that ß-alanine and histamine activate distinct populations of capsaicin-responsive neurons in primate dorsal root ganglion. We suggest that histamine itch and capsaicin pain are peripherally encoded in C-MIAs, and that primate polymodal nociceptive afferents form three functionally distinct subpopulations with ß-alanine responsive QC fibres likely corresponding to murine MrgprD-expressing, non-peptidergic nociceptive afferents.

Nerve growth factor induces sensitization of nociceptors without evidence for increased intraepidermal nerve fiber density.

Nerve growth factor (NGF) is involved in the long-term sensitization of nociceptive processing linked to chronic pain. Functional and structural ("sprouting") changes can contribute. Thus, humans report long-lasting hyperalgesia to mechanical and electrical stimulation after intradermal NGF injection and NGF-induced sprouting has been reported to underlie cancer bone pain and visceral pain. Using a human-like animal model we investigated the relationship between the structure and function of unmyelinated porcine nociceptors 3 weeks after intradermal NGF treatment. Axonal and sensory characteristics were studied by in vivo single-fiber electrophysiology and immunohistochemistry. C fibers recorded extracellularly were classified based on mechanical response and activity-dependent slowing (ADS) of conduction velocity. Intraepidermal nerve fiber (IENF) densities were assessed by immunohistochemistry in pigs and in human volunteers using the same NGF model. NGF increased conduction velocity and reduced ADS and propagation failure in mechano-insensitive nociceptors. The proportion of mechano-sensitive C nociceptors within NGF-treated skin areas increased from 45.1% (control) to 71% and their median mechanical thresholds decreased from 40 to 20 mN. After NGF application, the mechanical receptive fields of nociceptors increased from 25 to 43 mm(2). At the structural level, however, IENF density was not increased by NGF. In conclusion, intradermal NGF induces long-lasting axonal and mechanical sensitization in porcine C nociceptors that corresponds to hyperalgesia observed in humans. Sensitization is not accompanied by increased IENF density, suggesting that NGF-induced hyperalgesia might not depend on changes in nerve fiber density but could be linked to the recruitment of previously silent nociceptors.

Local loperamide injection reduces mechanosensitivity of rat cutaneous, nociceptive C-fibers.

Loperamide reverses signs of mechanical hypersensitivity in an animal model of neuropathic pain suggesting that peripheral opioid receptors may be suitable targets for the treatment of neuropathic pain. Since little is known about loperamide effects on the responsiveness of primary afferent nerve fibers, in vivo electrophysiological recordings from unmyelinated afferents innervating the glabrous skin of the hind paw were performed in rats with an L5 spinal nerve lesion or sham surgery. Mechanical threshold and responsiveness to suprathreshold stimulation were tested before and after loperamide (1.25, 2.5 and 5 µg in 10 µl) or vehicle injection into the cutaneous receptive field. Loperamide dose-dependently decreased mechanosensitivity in unmyelinated afferents of nerve-injured and sham animals, and this effect was not blocked by naloxone pretreatment. We then investigated loperamide effects on nerve conduction by recording compound action potentials in vitro during incubation of the sciatic nerve with increasing loperamide concentrations. Loperamide dose-dependently decreased compound action potentials of myelinated and unmyelinated fibers (ED50 = 8 and 4 µg/10 µl, respectively). This blockade was not prevented by pre-incubation with naloxone. These results suggest that loperamide reversal of behavioral signs of neuropathic pain may be mediated, at least in part, by mechanisms independent of opioid receptors, most probably by local anesthetic actions.

The differential effects of two sodium channel modulators on the conductive properties of C-fibers in pig skin in vivo.

BACKGROUND: Axonal sodium channels are attractive targets for chronic pain treatment, and recent evidence suggests that specific targeting of the slow inactivation of sodium channels (NaV) might exert analgesic effects. Using a human-like animal model, the pig, we compared changes in the conductive properties of different C-fiber classes on acute administration of lidocaine (nonselective NaV blocker) and lacosamide (selective enhancer of NaV slow inactivation). METHODS: Single-fiber extracellular recordings from saphenous nerves were performed. We classified C-fibers according to mechanical responsiveness and amount of activity-dependent slowing (ADS) of conduction velocity. Lidocaine (4 mM; 100 µL), lacosamide (4 mM; 100 µL), or saline was injected intradermally at the stimulation site, and changes of fibers' conductive properties were assessed. RESULTS: Conduction latencies evoked by lidocaine were more prominent in mechanosensitive (5.5%± 2.1%) than in mechano-insensitive nociceptors (2.5% ± 1%), whereas lacosamide increased conduction latencies to a greater extent in the mechano-insensitive (3% ± 1%) than in mechanosensitive C-nociceptors (2% ± 0.9%). Lidocaine, but not lacosamide, increased electrical thresholds in all mechanosensitive, but not in the mechano-insensitive, C-fibers. Lacosamide blocked conduction and, in addition, reduced ADS in mechano-insensitive nociceptors significantly more than in mechanosensitive nociceptors (ΔADS: 2.4% ± 0.5% vs 1.6% ± 0.5%), whereas lidocaine had opposite effects. Saline had no significant effect on the conductive properties of C-fibers. CONCLUSION: Local application of test compounds in pig skin allows for functional assessment of steady-state and use-dependent modulation of sodium channels in nociceptive and nonnociceptive C-fibers. Increased analgesic specificity might derive from selective enhancement of slow inactivation of sodium channels.

Nerve growth factor selectively decreases activity-dependent conduction slowing in mechano-insensitive C-nociceptors.

Nerve growth factor (NGF) induces acute sensitization of nociceptive sensory endings and long-lasting hyperalgesia. NGF modulation of sodium channel expression might contribute to neurotrophin-induced hyperalgesia. Here, we investigated NGF-evoked changes of the activity-dependent slowing of conduction in porcine C-fibers. Animals received intradermal injections of NGF (2 µg or 8 µg) or saline in both hind limbs. Extracellular recordings from the saphenous nerves were performed 1 week later. Based on sensory thresholds and electrically induced activity-dependent slowing (ADS) of axonal conduction, C-fibers were classified as mechano-sensitive afferents, mechano-insensitive afferents, cold nociceptors, and sympathetic efferents. NGF (2 µg) increased conduction velocity in C-fibers from 1.0±0.05 m/s to 1.2±0.07 m/s. In mechano-insensitive afferents, NGF (8 µg) reduced activity-dependent slowing of conduction, from 5.3±0.2% to 3.2±0.5% (0.125-0.5 Hz stimulation) and from 28.5±1.3% to 20.9±1.9% (2 Hz stimulation), such that ADS no longer differentiated between mechano-sensitive and mechano-insensitive fibers. Accordingly, the number of fibers with pronounced ADS decreased but more units with pronounced ADS were mechano-sensitive. Spontaneously active C-fibers were increased above the control level (1%) by NGF 8 µg (8%). The results demonstrate that NGF changes the functional axonal characteristics of mechano-insensitive C-fibers and enhances spontaneous activity thereby possibly contributing to hyperalgesia.

Phenotyping sensory nerve endings in vitro in the mouse.

This protocol details methods to identify and record from cutaneous primary afferent axons in an isolated mammalian skin-saphenous nerve preparation. The method is based on extracellular recordings of propagated action potentials from single-fiber receptive fields. Cutaneous nerve endings show graded sensitivities to various stimulus modalities that are quantified by adequate and controlled stimulation of the superfused skin with heat, cold, touch, constant punctate pressure or chemicals. Responses recorded from single-fibers are comparable with those obtained in previous in vivo experiments on the same species. We describe the components and the setting-up of the basic equipment of a skin-nerve recording station (few days), the preparation of the skin and the adherent saphenous nerve in the mouse (15-45 min) and the isolation and recording of neurons (approximately 1-3 h per recording). In addition, stimulation techniques, protocols to achieve single-fiber recordings, issues of data acquisition and action potential discrimination are discussed in detail.

Neural spike classification using parallel selection of all algorithm parameters.

The Forster-Handwerker template-matching algorithm (J. Neurosci. Methods 31 (1990) 109) classifies neuronal spikes according to three parameters selected by the experimenter prior to running the algorithm. Thousands of different combinations of these parameter values are possible producing hundreds of different classifications for each input file. Using a 40-processor Linux-based parallel computing cluster, we ran their algorithm with an effective sampling of all combinations of parameter values in order to generate a list of the classifications that can be generated by the algorithm. A distance measure was used to quantify the similarity between classifications and then to create a distance table containing entries for the distances between all pairs of classifications. Using a self-organizing neural network (SON) and the distance table we group the classifications by similarity and select the best representative classifications that the Forster-Handwerker algorithm can produce.