Human selection bias drives the linear nature of the more ground truth effect in explainable deep learning optical coherence tomography image segmentation

Supervised deep learning (DL) algorithms are highly dependent on training data for which human graders are assigned, for example, for optical coherence tomography (OCT) image annotation. Despite the tremendous success of DL, due to human judgment, these ground truth labels can be inaccurate and/or ambiguous and cause a human selection bias. We therefore investigated the impact of the size of the ground truth and variable numbers of graders on the predictive performance of the same DL architecture and repeated each experiment three times. The largest training dataset delivered a prediction performance close to that of human experts. All DL systems utilized were highly consistent. Nevertheless, the DL under-performers could not achieve any further autonomous improvement even after repeated training. Furthermore, a quantifiable linear relationship between ground truth ambiguity and the beneficial effect of having a larger amount of ground truth data was detected and marked as the more-ground-truth effect.

Cynomolgus monkey's retina volume reference database based on hybrid deep learning optical coherence tomography segmentation

Cynomolgus monkeys (Macaca fascicularis) are commonly used in pre-clinical ocular studies. However, studies that report the morphological features of the macaque retina are based only on minimal sample sizes; therefore, little is known about the normal distribution and background variation. This study was conducted using optical coherence tomography (OCT) imaging to investigate the variations in retinal volumes of healthy cynomolgus monkeys and the effects of sex, origin, and eye side on the retinal volumes to establish a comprehensive reference database. A machine-learning algorithm was employed to segment the retina within the OCT data (i.e., generated pixel-wise labels). Furthermore, a classical computer vision algorithm has identified the deepest point in a foveolar depression. The retinal volumes were determined and analyzed based on this reference point and segmented retinal compartments. Notably, the overall foveolar mean volume in zone 1, which is the region of the sharpest vision, was 0.205 mm³ (range 0.154-0.268 mm³), with a relatively low coefficient of variation of 7.9%. Generally, retinal volumes exhibit a relatively low degree of variation. However, significant differences in the retinal volumes due to the monkey's origin were identified. Additionally, sex had a significant impact on the paracentral retinal volume. Therefore, the origin and sex of cynomolgus monkeys should be considered when evaluating the macaque retinal volumes based on this dataset.

Cynomolgus monkey's choroid reference database derived from hybrid deep learning optical coherence tomography segmentation

Cynomolgus monkeys exhibit human-like features, such as a fovea, so they are often used in non-clinical research. Nevertheless, little is known about the natural variation of the choroidal thickness in relation to origin and sex. A combination of deep learning and a deterministic computer vision algorithm was applied for automatic segmentation of foveolar optical coherence tomography images in cynomolgus monkeys. The main evaluation parameters were choroidal thickness and surface area directed from the deepest point on OCT images within the fovea, marked as the nulla with regard to sex and origin. Reference choroid landmarks were set underneath the nulla and at 500 µm intervals laterally up to a distance of 2000 µm nasally and temporally, complemented by a sub-analysis of the central bouquet of cones. 203 animals contributed 374 eyes for a reference choroid database. The overall average central choroidal thickness was 193 µm with a coefficient of variation of 7.8%, and the overall mean surface area of the central bouquet temporally was 19,335 µm² and nasally was 19,283 µm². The choroidal thickness of the fovea appears relatively homogeneous between the sexes and the studied origins. However, considerable natural variation has been observed, which needs to be appreciated.

Retinal Features in Cynomolgus Macaques (Macaca fascicularis) Assessed by Using Scanning Laser Ophthalmoscopy and Spectral Domain Optical Coherence Tomography

Cynomolgus macaques are an important and commonly used species in preclinical toxicology studies, but structural reports of in vivo retinal findings are rare in this species. The purpose of this study was to diminish this gap and document optical coherence tomography and scanning laser ophthalmoscopy imaging data in the healthy posterior pole of cynomolgus monkeys' eyes at predose examinations. The current study is a retrospective assessment of baseline spectral domain OCT data obtained from the 768 eyes of 384 cynomolgus monkeys (192 males and 192 females) of Mauritian origin. The data set was obtained from studies conducted over a 4-y period in the context of ocular safety evaluations of various compounds under preclinical development. The most prevalent findings were the presence of Bergmeister papilla and intravitreal hyperreflective spots. Less common findings included disorganization of retinal zones, abnormalities of the retinal vasculature, partial posterior vitreous detachment, and abnormally shaped foveal pits. Thoughtful consideration of these physiologic findings will aid in distinguishing normal features from toxic outcomes in future preclinical ophthalmic studies.

Macular thickness measurements of healthy, naïve cynomolgus monkeys assessed with spectral-domain optical coherence tomography (SD-OCT)

The purpose of this study was to measure central macular thickness in an unprecedented number of cynomolgus monkeys. Macular thickness was measured with Heidelberg spectral-domain OCT in 320 eyes of healthy and treatment-naïve cynomolgus monkeys (80 males and 80 females). The macula was successfully measured in all 320 eyes. Macular thickness was not significantly different between the sexes. The mean central macular thickness was 244 μm (+/- 21 μm). Macular thicknesses in the quadrants were 327 +/-17 μm (temporal inner), 339 +/- 17 μm (inferior inner), 341 +/- 14 μm (superior inner), 341 +/-18 μm (nasal inner), and 299 +/- 20 μm (temporal outer), 320 +/- 16 μm (superior outer), 332 +/-23 μm (inferior outer), and 337 +/-18 μm (nasal outer). Highly significant differences between the nasal and temporal quadrants were detected. This study successfully demonstrated the feasibility of retinal thickness measurements in healthy cynomolgus monkeys. The present findings indicate that the macula is thicker in cynomolgus monkeys than in humans and provide important normative data for future studies.

Drug-induced Obstructive and Retrograde Nephropathy Associated with α2u-globulin in Male Rats

The chemically induced accumulation of α2u-globulin protein in male rats causes specific renal lesions and subsequent nephropathy. Herein, we report additional parallel findings in the kidney of male rats consistent with obstructive and retrograde nephropathy. Kidney and urinary bladder samples were evaluated from Wistar rats treated with RG7129 for 2 week and 8 week and from an 8-week mechanistic study using females, intact and castrated males. Histopathological findings were present in intact males in all studies, including hyaline droplet accumulation and granular casts consistent with α2u-globulin nephropathy. In addition, tubular degeneration and regeneration, tubular changes extending from papilla to cortex, tubular dilation, and interstitial and luminal inflammation were observed consistent with retrograde and obstructive nephropathy. Renal and urinary lesions and their severity increased in a time- and dose-dependent manner. Urinalysis findings, including increases in leukocytes, protein, and in kidney biomarkers, kidney injury molecule 1 and clusterin, were present only in intact males. No treatment-related changes were observed in female rats or in castrated males. These results indicate that RG7129 induces α2u-globulin nephropathy, associated with retrograde and obstructive nephropathy secondary to precipitation in intact male rats only, constituting a species- and sex-specific syndrome that is not expected to occur in humans or other species.

Absence of central nervous system and hypothermic effects after single oral administration of high doses of oseltamivir in the rat

Oseltamivir is widely used for the treatment and prophylaxis of influenza. Renewed interest in the central nervous system (CNS) tolerability profile of oseltamivir has been triggered by the reports of neuropsychiatric adverse events in patients with influenza. In addition, a recent pre-clinical study in rodents suggested a hypothermic effect of oseltamivir. The current studies investigated the CNS effects, body temperature effect and toxicokinetic profile of oseltamivir in rats. The CNS/temperature study included three groups receiving oseltamivir (500, 763 and 1000 mg/kg free base by oral gavage), one vehicle/control group and one reference group (D-amphetamine, 10 mg/kg). CNS parameters (behaviour, motor activity and co-ordination and sensory/motor reflex responses) and rectal temperature were measured at baseline and at five intervals until 8 hr after dosing. In the toxicokinetic study, rats received oseltamivir by oral gavage at 763 or 1000 mg/kg free base. Plasma, cerebrospinal fluid (CSF) and perfused brain concentrations of oseltamivir and its active metabolite, oseltamivir carboxylate (OC), were measured until 8 hr after dosing. Median scores for CNS parameters were similar in controls and animals receiving oseltamivir at all time points. Oseltamivir had no physiologically relevant effect on body temperature, but induced a short-lived and small dose-independent decrease in temperature in all active treatment groups at 1 hr after dosing only. Plasma concentrations of OC were higher than of oseltamivir, but the reverse was true in CSF and brain. CNS penetration was low for both moieties. In rats, oseltamivir at supratherapeutic doses up to 1000 mg/kg free base did not exert any effects on CNS function or hypothermic effects and led to limited CNS exposure, resulting in large safety margins.

Age-dependent cognitive decline and amygdala pathology in alpha-synuclein transgenic mice

Intraneuronal alpha-synuclein (alphaSYN) inclusions constitute the hallmark lesions of a number of neurodegenerative diseases, including Parkinson's disease and dementia with Lewy bodies. In a transgenic mouse model expressing mutant [A30P]alphaSYN under control of the pan-neuronal Thy1 promoter, motor impairment became significant beyond 17 months of age. Cognitive performance was measured in the Morris water maze and upon fear conditioning. At 4 months of age, transgenic mice performed like controls. However, performance in these tasks was significantly impaired in (Thy1)-h[A30P]alphaSYN mice at 12 months of age. After completion of the cognition tests, mice were sacrificed and the regional distribution of neuropathology was examined. In contrast to 4 months old animals, 12 months old transgenic mice showed alpha-synucleinopathy in several brain regions, including the central nucleus of the amygdala, which is involved in cognitive behavior of mice, and is susceptible to alphaSYN pathology in human patients. Thus, age-dependent fibrillization of alphaSYN in specific cortical regions concomitant with cognitive decline may reflect dementia with Lewy bodies in a transgenic mouse model.

Acute changes in the neuronal expression of GABA and glutamate decarboxylase isoforms in the rat piriform cortex following status epilepticus

The piriform cortex (PC) is the largest region of the mammalian olfactory cortex with strong connections to other limbic structures, including the amygdala, hippocampus, and entorhinal cortex. In addition to its functional importance in the classification of olfactory stimuli, the PC has been implicated in the study of memory processing, spread of excitatory information, and the facilitation and propagation of seizures within the limbic system. Previous data from the kindling model of epilepsy indicated that alterations in GABAergic inhibition in the transition zone between the anterior and posterior PC, termed here central PC, are particularly involved in the processes underlying seizure propagation. In the present study we studied alterations in GABAergic neurons in different parts of the PC following seizures induced by kainate or pilocarpine in rats. GABA neurons were labeled either immunohistochemically for GABA or its synthesizing enzyme glutamate decarboxylase (GAD) or by in situ hybridization using antisense probes for GAD65 and GAD67 mRNAs. For comparison with the PC, labeled neurons were examined in the basolateral amygdala, substantia nigra pars reticulata, and the hippocampal formation. In the PC of controls, immunohistochemical labeling for GABA and GAD yielded consistently higher neuronal densities in most cell layers than labeling for GAD65 or GAD67 mRNAs, indicating a low basal activity of these neurons. Eight hours following kainate- or pilocarpine-induced seizures, severe neuronal damage was observed in the PC. Counting of GABA neurons in the PC demonstrated significant decreases in densities of neurons labeled for GABA or GAD proteins. However, a significantly increased density of neurons labeled for GAD65 and GAD67 mRNAs was determined in layer II of the central PC, indicating that a subpopulation of remaining neurons up-regulated the mRNAs for the GAD isoenzymes. One likely explanation for this finding is that remaining GABA neurons in layer II of the central PC maintain high levels of activity to control the increased excitability of the region. In line with previous studies, an up-regulation of GAD67 mRNA, but not GAD65 mRNA, was observed in dentate granule cells following seizures, whereas no indication of such up-regulation was determined for the other brain regions examined. The data substantiate the particular susceptibility of the central PC to seizure-induced plasticity and indicate that this brain region provides an interesting tool to study the regulation of GAD isoenzymes.

Distribution of GABAergic neurons in the striatum of amygdala-kindled rats: An immunohistochemical and in situ hybridization study

A large body of experimental evidence suggests that the basal ganglia circuitry may be part of a remote control system modulating the spread of epileptic seizures. In the kindling model of temporal lobe epilepsy, this endogenous inhibitory control mechanism seems to be impaired. Neurochemical and neurophysiological studies have indicated that the activity of the GABAergic projection from the striatum to the substantia nigra pars reticulata is reduced in kindled rats, but the exact mechanisms involved in this observation are not known. Possible explanations include a kindling-induced loss of striatal GABAergic projection neurons to the substantia nigra or enhanced inhibition of these neurons by GABAergic interneurons. In the present experiments, the GABAergic system of the striatum (caudate-putamen) of amygdala-kindled rats and controls was studied immunohistochemically with a monoclonal antibody to GABA and with nonisotopic in situ hybridization with cRNA probes selective for glutamic acid decarboxylase 65 (GAD65) and GAD67, respectively. Compared to sham controls, an increased density of neurons heavily labeled for GAD67 mRNA was observed in the anterior striatum of kindled rats when cells were counted 6 weeks after the last kindled seizure. This subgroup of striatal GABAergic neurons has been suggested previously to correspond to the medium-sized aspiny interneurons in the striatum, indicating that kindling is associated with an increased activity of these neurons. Our previous finding of reduced GAD and GABA levels in synaptosomes isolated from the substantia nigra of kindled rats together with the present observation of increased density of GABAergic striatal interneurons in such rats suggest that kindling affects the regulation of the GABAergic projections from the striatum to the substantia nigra rather than directly damaging GABAergic neurons in the striatum.

Amygdala-kindling does not induce a persistent loss of GABA neurons in the substantia nigra pars reticulata of rats

GABAergic inhibition of the substantia nigra pars reticulata (SNR) has been shown to suppress seizures in most models of epilepsy, including the amygdala-kindling model of temporal lobe epilepsy (TLE). A dysfunction of this seizure gating mechanism of the SNR may lead to facilitation of seizure propagation in such models. In post-status epilepticus models of TLE, GABAergic neurons in the SNR are damaged, but it is not known whether such damage also occurs in kindling. By using stereological techniques for cell counting in amygdala-kindled rats, we determined the density of SNR neurons that were labeled for GABA by immunohistochemistry or for the two isoforms of the GABA-synthesizing enzyme glutamate decarboxylase (GAD), GAD65 and GAD67, by in situ hybridization (ISH). In addition, GABA neurons in the basolateral amygdala (BLA) were counted. While there was a significant reduction of GAD65 mRNA expressing neurons in the BLA of kindled rats, no alteration in the density of neurons was observed in the anterior or posterior SNR when cells were counted 6 weeks after the last kindled seizure. Our previous finding of reduced GAD and GABA levels in synaptosomes isolated from the SN of kindled rats together with the present observation of unchanged density of SNR neurons in such rats suggest that kindling affects the GABAergic projections from the striatum or globus pallidus to the SNR rather than directly affecting GABA neurons in the SNR.

Effect of phenytoin on sodium and calcium currents in hippocampal CA1 neurons of phenytoin-resistant kindled rats

About 20-30% of patients with epilepsy continue to have seizures despite carefully monitored treatment with antiepileptic drugs. The mechanisms explaining why some patients' respond and others prove resistant to antiepileptic drugs are poorly understood. It has been proposed that pharmacoresistance is related to reduced sensitivity of sodium channels in hippocampal neurons to antiepileptic drugs such as carbamazepine or phenytoin. In line with this proposal, a reduced effect of carbamazepine on sodium currents in hippocampal CA1 neurons was found in the rat kindling model of temporal lobe epilepsy (TLE), i.e. a form of epilepsy with the poorest prognosis of all epilepsy types in adult patients. To address directly the possibility that neuronal sodium currents in the hippocampus play a crucial role in the pharmacoresistance of TLE, we selected amygdala-kindled rats with respect to their in vivo anticonvulsant response to phenytoin into responders and nonresponders and then compared phenytoin's effect on voltage-activated sodium currents in CA1 neurons. Furthermore, in view of the potential role of calcium current modulation in the anticonvulsant action of phenytoin, the effect of phenytoin on high-voltage-activated calcium currents was studied in CA1 neurons. Electrode-implanted but not kindled rats were used as sham controls for comparison with the kindled rats. In all experiments, the interval between last kindled seizure and ion channel measurements was at least 5 weeks. In kindled rats with in vivo resistance to the anticonvulsant effect of phenytoin (phenytoin nonresponders), in vitro modulation of sodium and calcium currents by phenytoin in hippocampal CA1 neurons did not significantly differ from respective data obtained in phenytoin responders, i.e. phenytoin resistance was not associated with a changed modulation of the sodium or calcium currents by this drug. Compared to sham controls, phenytoin's inhibitory effect on sodium currents was significantly reduced by kindling without difference between the responder and nonresponder subgroups. Further studies in phenytoin-resistant kindled rats may help to elucidate the mechanisms that can explain therapy resistance.