[18F]DPA-714: Effect of co-medications, age, sex, BMI and TSPO polymorphism on the human plasma input function.

PURPOSE: We aimed to assess the effect of concomitant medication, age, sex, body mass index and 18-kDa translocator protein (TSPO) binding affinity status on the metabolism and plasma pharmacokinetics of [18F]DPA-714 and their influence on the plasma input function in a large cohort of 201 subjects who underwent brain and whole-body PET imaging to investigate the role of neuroinflammation in neurological diseases. METHODS: The non-metabolized fraction of [18F]DPA-714 was estimated in venous plasma of 138 patients and 63 healthy controls (HCs; including additional arterial sampling in 16 subjects) during the 90 min brain PET acquisition using a direct solid-phase extraction method. The mean fraction between 70 and 90 min post-injection ([18F]DPA-71470-90) and corresponding normalized plasma concentration (SUV70-90) were correlated with all factors using a multiple linear regression model. Differences between groups (arterial vs venous measurements; HCs vs patients; high- (HAB), mixed- (MAB) and low-affinity binders (LAB); subjects with vs without co-medications, females vs males were also assessed using the non-parametric Mann-Whitney or Kruskal-Wallis ANOVA tests. Finally, the impact of co-medications on the brain uptake of [18F]DPA-714 at equilibrium was investigated. RESULTS: As no significant differences were observed between arterial and venous [18F]DPA-71470-90 and SUV70-90, venous plasma was used for correlations. [18F]DPA-71470-90 was not significantly different between patients and HCS (59.7 ± 12.3% vs 60.2 ± 12.9%) despite high interindividual variability. However, 47 subjects exhibiting a huge increase or decrease of [18F]DPA-71470-90 (up to 88% or down to 23%) and SUV70-90 values (2-threefold) were found to receive co-medications identified as inhibitors or inducers of CYP3A4, known to catalyse [18F]DPA-714 metabolism. Comparison between cortex-to-plasma ratios using individual input function (VTIND) or population-based input function derived from untreated HCs (VTPBIF) indicated that non-considering the individual metabolism rate led to a bias of about 30% in VT values. Multiple linear regression model analysis of subjects free of these co-medications suggested significant correlations between [18F]DPA-71470-90 and age, BMI and sex while TSPO polymorphism did not influence the metabolism of the radiotracer. [18F]DPA-714 metabolism fell with age and BMI and was significantly faster in females than in males. Whole-body PET/CT exhibited a high uptake of the tracer in TSPO-rich organs (heart wall, spleen, kidneys…) and those involved in metabolism and excretion pathways (liver, gallbladder) in HAB and MAB with a strong decrease in LAB (-89% and -85%) resulting in tracer accumulation in plasma (4.5 and 3.3-fold increase). CONCLUSION: Any co-medication that inhibits or induces CYP3A4 as well as TSPO genetic status, age, BMI and sex mostly contribute to interindividual variations of the radiotracer metabolism and/or concentration that may affect the input function of [18F]DPA-714 and consequently its human brain and peripheral uptake. TRIAL REGISTRATION: INFLAPARK, NCT02319382, registered December 18, 2014, retrospectively registered; IMABIO 3, NCT01775696, registered January 25, 2013, retrospectively registered; INFLASEP, NCT02305264, registered December 2, 2014, retrospectively registered; EPI-TEP, EudraCT 2017-003381-27, registered September 24, 2018.

[Reconciliating neurology and psychiatry: The prototypical case of frontotemporal dementia]. / Rapprochement entre neurologie et psychiatrie : le cas prototypique de la dégénérescence frontotemporale.

Frontotemporal degeneration (FTD) in its behavioral variant (bvFTD) is probably one of the conditions that best illustrates the links between psychiatry and neurology. It is indeed admitted that between a third and half of patients with this condition, especially in early-onset forms, receive an initial diagnosis of psychiatric disorder (depression, schizophrenia, bipolar disorder) and are then referred to a psychiatric ward. BvFTD can thus be considered a neurological disorder with a psychiatric presentation. Among psychiatric symptoms reported in this disease, psychotic symptoms (hallucinations, delusions, especially of persecution), which have long been underestimated in bvFTD and are not part of the current diagnostic criteria, are present in about 20% of cases and may be inaugural. They are particularly common in the genetic forms related to a mutation in the C9orf72 gene (up to 50%), and to a lesser extent in the GRN gene (up to 25%). C9orf72 gene mutation is often associated with a family history of dementia or motor neuron disease but also of psychiatric disorders. It has also been described in sporadic presentation forms. Sometimes, the moderate degree of brain atrophy on MRI described in patients carrying this mutation may complicate the differential diagnosis with late-onset psychiatric diseases. In the present article, we underline the importance of considering that psychiatric - especially psychotic - symptoms are not rare in bvFTD, which should lead to a revision of the diagnostic criteria of this disease by taking greater account of this fact. We also propose a diagnostic chart, based on concerted evaluation by neurologists and psychiatrists for cases of atypical psychiatric symptoms (late-onset or pharmacoresistant troubles) leading to consider the possibility of a neurological disorder, in order to shed a new light on these difficult clinical situations. In the field of research, bvFTD may constitute a model to explore the neural basis of certain psychiatric disorders, and a possible molecular link between bvFTD and psychoses, which could eventually lead to new therapeutic approaches, has been recently suggested. Thus, bvFTD illustrates how the links between neurology and psychiatry are close and tend to evolve with the progress of scientific knowledge. It is necessary to strengthen collaboration between the two disciplines both to improve the care - diagnosis and management of these patients - and to promote the emergence of innovative clinical research.

The economic burden of urinary tract infections in women visiting general practices in France: a cross-sectional survey.

BACKGROUND: Urinary tract infections (UTIs) are among the most common bacterial infections. Despite this burden, there are few studies of the costs of UTIs. The objective of this study was to determine the costs of UTIs in women over 18 years of age who visit general practitioners in France. METHODS: The direct and indirect costs of clinical UTIs were estimated from societal, French National Health Insurance and patient perspectives. The study population was derived from a national cross-sectional survey entitled the Drug-Resistant Urinary Tract Infection (Druti). The Druti included every woman over 18 years of age who presented with symptoms of UTI and was conducted in France in 2012 and 2013 to estimate the annual incidence of UTIs due to antibiotic-resistant Enterobacteriaceae in women visiting general practitioners (GPs) for suspected UTIs. RESULTS: Of the 538 women included in Druti, 460 were followed over 8 weeks and included in the cost analysis. The mean age of the women was 46 years old. The median cost of care for one episode of a suspected UTI was €38, and the mean cost was €70. The annual societal cost was €58 million, and €29 million of this was reimbursed by the French National Health Insurance system. In 25 % of the cases, the suspected UTIs were associated with negative urine cultures. The societal cost of these suspected UTIs with negative urine cultures was €13.5 million. No significant difference was found between the costs of the UTIs due to antibiotic-resistant E. coli and those due to wild E. coli (p = 0.63). CONCLUSION: In the current context in which the care costs are continually increasing, the results of this study suggests that it is possible to decrease the cost of UTIs by reducing the costs of suspected UTIs and unnecessary treatments, as well as limiting the use of non-recommended tests.

Magnetic resonance imaging for diagnosis of early Alzheimer's disease.

A major challenge for neuroimaging is to contribute to the early diagnosis of Alzheimer's disease (AD). In particular, magnetic resonance imaging (MRI) allows detecting different types of structural and functional abnormalities at an early stage of the disease. Anatomical MRI is the most widely used technique and provides local and global measures of atrophy. The recent diagnostic criteria of "mild cognitive impairment due to AD" include hippocampal atrophy, which is considered a marker of neuronal injury. Advanced image analysis techniques generate automatic and reproducible measures both in the hippocampus and throughout the whole brain. Recent modalities such as diffusion-tensor imaging and resting-state functional MRI provide additional measures that could contribute to the early diagnosis but require further validation.

Disease-specific adaptive immune biomarkers in Alzheimer's disease and related pathologies.

Identification of disease-specific diagnostic and prognostic biomarkers allowing for an early characterization and accurate clinical follow-up of Alzheimer's disease (AD) patients is a major clinical objective. Increasing evidences implicate both humoral and cellular adaptive immune responses in the pathophysiology of AD. Such disease-related B- and T-cell responses constitute a promising source of potential specific early biomarkers. Among them, levels of anti-Aß antibodies in the serum and/or cerebrospinal fluid of patients may correlate with AD progression, clinical presentation of the disease, and occurrence of associated pathologies related to cerebral amyloid angiopathy. In the same line, Aß-specific T cell responses and immune regulatory populations implicated in their modulation appear to play a role in the pathophysiology of AD and cerebral amyloid angiopathy. Further characterization of both autoantibodies and T cell responses specific for disease-related proteins, i.e. Aß and hyperphosphorylated Tau, will allow better deciphering their interest as early diagnostic and prognostic markers in AD. Biomarkers of adaptive immune responses specific for other pathological proteins may also apply to other neurological disorders associated with abnormal protein deposition.

Alzheimer disease: from biomarkers to diagnosis.

The discovery of biomarkers, considered as surrogate markers of the underlying pathological changes, led an international work group (IWG) to propose a new conceptual framework for AD in 2007 Dubois et al. (2007). According to the IWG, AD is now defined as a dual clinico-biological entity that can be recognized in vivo, prior to the onset of the dementia syndrome, on the basis of: i) a specific core clinical phenotype comprised of an amnestic syndrome of the hippocampal type and ii) supportive evidence from biomarkers reflecting the location or the nature of Alzheimer-type changes. Therefore, AD is diagnosed with the same criteria throughout all symptomatic phases of the disease based on the biologically-based approach to diagnosis independent of clinical expression of disease severity. The definitions were further clarified in 2010 (Dubois et al., 2010). Although the new criteria are proposed for research purposes, we encourage expert centres with adequate resources to begin to use the proposed algorithm in order to move the field forward and facilitate translation into clinical practice.

[Markers of prodromal Alzheimer's disease]. / Les marqueurs de la maladie d'Alzheimer prodromale.

The diagnosis of Alzheimer's disease has long been considered a diagnosis of probability, as the definitive diagnosis can only be established by histopathological examination. However, the development of in-vivo biomarkers, considered a reflection of physiopathological processes, has changed our view of the disease. New criteria have recently been proposed that integrate such biomarkers as found in the cerebrospinal fluid (CSF) using new diagnostic tools such as magnetic resonance imaging (MRI), brain scintigraphy, FDG-positron emission tomography (PET) and PET amyloid ligand uptake studies. The value of these new criteria for the diagnosis of prodromal Alzheimer's disease and the prospect of disease-modifying drugs are also discussed.

[Beneficial effect of a cognitive behavioral and multidisciplinary program in Alzheimer Disease on spouse caregiver anxiety: French study ELMMA]. / Bénéfice d'un programme cognitivo-comportemental et pluridisciplinaire de prise en charge de la maladie d'Alzheimer sur l'anxiété du conjoint : étude française ELMMA.

INTRODUCTION: Over the last decade, several programs have been developed for caregivers of Alzheimer disease patients. In France however, studies exploring their effects are still scarce. We conducted a study to compare two different interventions: a structured multidisciplinary program versus a classical intervention designed for Alzheimer disease patients and their spouses. METHODS: Sixteen couples (Alzheimer's disease patient and spouse) residing in our administrative district participated in this monocentric study. For at least two years, these couples participated in a multidisciplinary program (n=8 couples) or received usual care (n=8 couples). The multidisciplinary program involved biannual consultations with a neurologist, a neuropsychologist and a psychologist, in addition to an annual meeting, stratified on the patient's MMSE score, for spouses). Usual care involved biannual consultations with the neurologist. The multidisciplinary program included a psychological intervention based on cognitive behavioral theories and centered on psycho-education, problem solving, adaptation strategies and on prevention of depression and anxiety. The spouses and the patients evaluated the 2-year follow-up during clinical interviews, completed by questionnaires. Sociodemographic data were noted for the patients and their spouses. Levels of depression and anxiety (Mini International Neuropsychiatric Inventory, Montgomery and Asberg Depression Scale, State-Trait Anxiety Inventory), perceived stress (Perceived Stress Scale) and care burden (Zarit Burden Inventory) were evaluated in spouses. Levels of cognitive impairment (Mini Mental State Examination), autonomy (Instrumental Activities of Daily Living), psychological state (Montgomery and Asberg Depression Scale, Covi Anxiety Scale), and behavioral symptoms frequency (Neuropsychiatric Inventory) were assessed in patients. RESULTS: The main significant result showed that the spouses' state of anxiety was lower among participants in the multidisciplinary program, compared with the classical neurological intervention. It also was found that the spouses and the patients who participated in this multidisciplinary program were less depressed. CONCLUSION: This study shows that a multidisciplinary structured intervention, with only two annual consultations and one annual meeting for spouses, can contribute to decrease significantly the spouses' state of anxiety. Further studies including a larger number of subjects should be conducted to confirm these findings.

[Semantic dementia associated with amyotrophic lateral sclerosis]. / Démence sémantique associée à une sclérose latérale amyotrophique.

The association semantic dementia-amyotrophic lateral sclerosis has been rarely reported in the literature. We report a case study of a patient with a severe prosopagnosia in relation with semantic dementia associated, 12 months later, with a motor neuron disease.

[Semantic dementia: reflexions of a French working group for diagnostic criteria and constitution of a patient cohort]. / Démence sémantique: réflexions d'un groupe de travail pour des critères de diagnostic en français et la constitution d'une cohorte de patients.

Semantic dementia (SD) is a syndrome of progressive loss of semantic knowledge for objects and people. International criteria propose that SD be included in the frontotemporal lobar degeneration syndromes, with progressive non-fluent aphasia and frontotemporal dementia (FTD). However, several related syndromes have been defined that clinically and conceptually share both similarities and differences with SD: fluent progressive aphasia, progressive prosopagnosia, temporal variant of FTD. In order to establish a French consensus for the diagnosis and modalities of evaluation and follow-up of SD, a working group, composed of neurologists, neuropsychologists and speech-therapists, was established by the Groupe de réflexion sur les évaluations cognitives (GRECO). New criteria were elaborated, based on clinical, neuropsychological, and imaging data. They define typical and atypical forms of SD. A diagnosis of typical SD relies on an isolated and progressive loss of semantic knowledge, attested by a deficit of word comprehension and a deficit of objects and/or people identification, with imaging showing temporal atrophy and/or hypometabolism. SD is atypical if the deficit of semantic knowledge is present only within a single modality (verbal versus visual), or if non-semantic deficits (mild and not present at onset) and/or neurological signs, are associated with the semantic loss.

Combined assessment of DYRK1A, BDNF and homocysteine levels as diagnostic marker for Alzheimer's disease.

Early identification of Alzheimer's disease (AD) risk factors would aid development of interventions to delay the onset of dementia, but current biomarkers are invasive and/or costly to assess. Validated plasma biomarkers would circumvent these challenges. We previously identified the kinase DYRK1A in plasma. To validate DYRK1A as a biomarker for AD diagnosis, we assessed the levels of DYRK1A and the related markers brain-derived neurotrophic factor (BDNF) and homocysteine in two unrelated AD patient cohorts with age-matched controls. Receiver-operating characteristic curves and logistic regression analyses showed that combined assessment of DYRK1A, BDNF and homocysteine has a sensitivity of 0.952, a specificity of 0.889 and an accuracy of 0.933 in testing for AD. The blood levels of these markers provide a diagnosis assessment profile. Combined assessment of these three markers outperforms most of the previous markers and could become a useful substitute to the current panel of AD biomarkers. These results associate a decreased level of DYRK1A with AD and challenge the use of DYRK1A inhibitors in peripheral tissues as treatment. These measures will be useful for diagnosis purposes.

Apathy in patients with mild cognitive impairment and the risk of developing dementia of Alzheimer's disease: a one-year follow-up study.

OBJECTIVE: To evaluate the relation between apathy and development of dementia in patients with amnestic mild cognitive impairment (MCI). METHODS: Two hundred and fifty-one French-speaking outpatients fulfilling the criteria of amnestic MCI were enrolled. Apathy was assessed with the Apathy Inventory (IA). Neuropsychiatric evaluation also included the Goldberg anxiety scale and the Montgomery and Asberg Depressive Rating Scale (MADRS). The main end point considered after a 1-year follow-up was the development of dementia of Alzheimer type (DAT). RESULTS: At baseline there were 86 (39.8%) subjects presenting at least one symptom of apathy among the 216 included in analysis. After a 1-year follow-up, 22 patients developed DAT. Of the patients with apathy at baseline 13 (15.1%) developed DAT in comparison with 9 (6.9%) of the non-apathetic patients. At the 1-year follow-up, patients developing DAT had a significantly higher frequency of apathetic symptoms (91.7%) than patients without DAT (26.9%). CONCLUSION: Taking into account that apathy is one of the most frequently observed neuropsychiatric symptoms in MCI and in DAT the present study suggests that patients with MCI and apathy should be more closely observed.

Modifications of the endosomal compartment in peripheral blood mononuclear cells and fibroblasts from Alzheimer's disease patients.

Identification of blood-based biomarkers of Alzheimer's disease (AD) remains a challenge. Neuropathological studies have identified enlarged endosomes in post-mortem brains as the earliest cellular change associated to AD. Here the presence of enlarged endosomes was investigated in peripheral blood mononuclear cells from 48 biologically defined AD patients (25 with mild cognitive impairment and 23 with dementia (AD-D)), and 23 age-matched healthy controls using immunocytochemistry and confocal microscopy. The volume and number of endosomes were not significantly different between AD and controls. However, the percentage of cells containing enlarged endosomes was significantly higher in the AD-D group as compared with controls. Furthermore, endosomal volumes significantly correlated to [C(11)]PiB cortical index measured by positron emission tomography in the AD group, independently of the APOE genotype, but not to the levels of amyloid-beta, tau and phosphorylated tau measured in the cerebrospinal fluid. Importantly, we confirmed the presence of enlarged endosomes in fibroblasts from six unrelated AD-D patients as compared with five cognitively normal controls. This study is the first, to our knowledge, to report morphological alterations of the endosomal compartment in peripheral cells from AD patients correlated to amyloid load that will now be evaluated as a possible biomarker.

Clinicometabolic dissociation of cognitive functions and social behavior in frontal lobe lesions.

OBJECTIVE/BACKGROUND: Case studies suggest a dissociation between cognitive functions that have been impaired after damage to the dorsolateral prefrontal cortex and social skills disturbed when the ventromedial prefrontal areas are affected. Because this dissociation had not been confirmed in a clinical setting, clinicometabolic correlations were sought in 13 patients with various lesions of the prefrontal cortex. DESIGN/METHODS: The clinical assessment included extensive testing of executive functions and evaluation of behavioral abnormalities based on an informant questionnaire. Regional cerebral glucose metabolism (rCMRGlu) was measured with [l8F] fluorodeoxyglucose ([18F] FDG) and 31-slide high-resolution PET. RESULTS: Executive-function test performance was significantly correlated with rCMRGlu in the dorsolateral prefrontal cortex (Brodmann's areas 8, 9, 45, 46, and 47) and anterior cingulate cortex (Brodmann's areas 24 and 32). Behavioral scores were significantly correlated with rCMRGlu in the frontopolar (Brodmann's area 10) and orbitofrontal cortex (Brodmann's areas 11, 12, 13, and 14). CONCLUSION: These results show that impaired executive functions and serial skill deficits are associated with distinct metabolic patterns in patients with frontal lobe pathology. In agreement with activation studies in normal subjects, our data suggest the existence of a modular organization of the frontal cortex in humans, as previously reported in nonhuman primates.

Primary central nervous system lymphoma: treatment with chemotherapy and radiotherapy.

Between 1989 and 1993, 22 HIV negative patients with primary central nervous system lymphoma (PCNLS) were treated with three different regimens. In group A, 13 patients received preradiotherapy systemic and intrathecal methotrexate (MTX), radiotherapy (RT) and three courses of post-RT chemotherapy (CT) with thiotepa and procarbazine. In group B, 4 patients received a similar CT only after RT and without intrathecal MTX in 3/4 cases. In group C, 5 elderly patients received CT alone. In group A, 9/13 patients achieved response after pre-RT CT and 12/13 were in complete response (CR) after RT. After a median follow-up of 27 months, 8/13 (62%) patients are alive but 4 have leucoencephalopathy and cognitive dysfunction. In group B, all 4 patients were in CR after RT but the 3 patients who did not receive intrathecal MTX died within 10 months with meningeal recurrence. In group C, 4/5 patients had a response to CT. 2 patients died of recurrent tumour at 5 and 10 months, and 2 are living in CR 11+ and 21+ months after diagnosis, 1 after salvage CT. Combined treatment with RT and CT is useful in PCNSL but adequate treatment of the meninges is required. CT alone is sometimes of value in elderly patients in whom RT is not indicated.

Memory for spatial location is affected in Parkinson's disease.

Are the striato-frontal neuronal circuits implicated in learning of item-specific spatial coordinates? To answer this question, we compared the performance of 20 patients with Parkinson's disease to that of 14 controls matched for age, global cognitive efficiency and mood, on a visuo-spatial learning task with little involvement of motor and constructive functions, allowing control of encoding and comparison of free recall, cued recall and recognition. Compared to controls, patients showed a severe memory impairment for visuo-spatial location of pictures, contrasting with relative preservation of verbal memory, and mild difficulties in perceptive visuo-spatial and executive functions. These results implicate striato-frontal neuronal circuits in memory for spatial location.

Influence on bone metabolism of dietary trace elements, protein, fat, carbohydrates, and vitamins.

Osteoporosis is a multifactorial disease driven primarily by the genetic factors that control bone metabolism. Among environmental factors, diet may play a key role, affording a target for low-cost intervention. Calcium and vitamin D are well known to affect bone metabolism. Other nutrients may influence bone mass changes; for instance, a number of trace elements and vitamins other than vitamin D are essential to many of the steps of bone metabolism. A wide variety of foods provide these nutrients, and in industrialized countries deficiencies are more often due to idiosyncratic eating habits than to cultural influences. Both culture and vogue influence the amount of carbohydrate, fat, and protein in the typical diet. In children, the current trend is to reduce protein and to increase carbohydrate and fat. Data from epidemiological and animal studies suggest that this may adversely affect bone mass and the fracture risk.

[Mild Cognitive Impairment or pre-demential Alzheimer's disease?]. / Trouble cognitif léger ou maladie d'Alzheimer au stade prédémentiel?

The concept of Mild Cognitive Impairment (MCI) is proposed to define a state of cognition where the deficiency is greater than expected for a subject's age and socio-cultural background, but not sufficiently severe to satisfy the criteria of nosographic classifications of dementia. The concept of MCI cannot provide an etiological diagnosis but can be useful for recognizing and following patients with mild impairment. It raises the question of early diagnosis of Alzheimer's disease. The risk of progression to dementia is greater in patients with MCI, particularly those whose memory is the only domain involved - "amnesic MCI". Patients with amnesic MCI can be distinguished from those with MCI with multiple domains slightly impaired or non-amnesic MCI where only one domain (language, visual-spatial.) other than memory is involved. Depending on the type of deficiency, amnesic MCI may progress to different types of disease states. At the early stage of pure memory impairment, a distinction should be made between Alzheimer-related memory disorders and those related to another neurological conditions or the brain aging process. The question is whether the diagnosis of Alzheimer's disease can be made at this stage. Biological markers and brain imaging provide useful information, but these diagnostic tools remain imprecise and have not been validated for routine use. The process of diagnosis must therefore focus on an analysis of the cognitive impairment. The predementia phase of Alzheimer's disease is characterized by a progressive "hippocampal" decline in memory, sometimes associated with impaired execution. A careful analysis of the cognitive impairment helps identify "hippocampal amnesia syndrome" suggestive of the diagnosis of Alzheimer's disease.

[Procedural learning and Parkinson disease: implication of striato-frontal loops]. / Apprentissage procédural et maladie de Parkinson: implication des boucles striato-frontales.

OBJECTIVE: To investigate procedural learning in non demented patients with idiopathic Parkinson's disease (PD). BACKGROUND: Experimental evidence implicate the basal ganglia in procedural learning. Selective impairment has more recently been described in patients with frontal lesions. METHODS: The performance of 20 demented patients and 15 matched normal controls was studied in the serial reaction time task (SRTT). Performance on procedural task was further compared with that of 9 normal controls and with patients' performance on tests assessing explicit memory, executive functions and global efficiency. RESULTS: The group of patients with PD showed impaired procedural learning. The difference of response time between the repeated and the non-repeated blocks was smaller in PD when compared to controls. Subsequent analyses separated PD patients into two subgroups according to their performance on SRTT, measured by the rebound effect. PD patients whose learning was normal differed from PD patients whose learning was impaired on performance in tests sensitive to frontal lobe dysfunction. CONCLUSION: These results confirm the implication of the striatum in procedural learning and suggest that performance on cognitive procedural learning depends on the striato-frontal circuits.

Dynamical recurrent neural networks--towards environmental time series prediction.

Dynamical Recurrent Neural Networks (DRNN) (Aussem 1995a) are a class of fully recurrent networks obtained by modeling synapses as autoregressive filters. By virtue of their internal dynamic, these networks approximate the underlying law governing the time series by a system of nonlinear difference equations of internal variables. They therefore provide history-sensitive forecasts without having to be explicitly fed with external memory. The model is trained by a local and recursive error propagation algorithm called temporal-recurrent-backpropagation. The efficiency of the procedure benefits from the exponential decay of the gradient terms backpropagated through the adjoint network. We assess the predictive ability of the DRNN model with meterological and astronomical time series recorded around the candidate observation sites for the future VLT telescope. The hope is that reliable environmental forecasts provided with the model will allow the modern telescopes to be preset, a few hours in advance, in the most suited instrumental mode. In this perspective, the model is first appraised on precipitation measurements with traditional nonlinear AR and ARMA techniques using feedforward networks. Then we tackle a complex problem, namely the prediction of astronomical seeing, known to be a very erratic time series. A fuzzy coding approach is used to reduce the complexity of the underlying laws governing the seeing. Then, a fuzzy correspondence analysis is carried out to explore the internal relationships in the data. Based on a carefully selected set of meteorological variables at the same time-point, a nonlinear multiple regression, termed nowcasting (Murtagh et al. 1993, 1995), is carried out on the fuzzily coded seeing records. The DRNN is shown to outperform the fuzzy k-nearest neighbors method.