This critical synopsis of prior work by Luca Parisi is submitted in support of a PhD by Published Work. The work focuses on deriving accurate, reliable and explainable clinical decision support systems as minimum clinically viable solutions leveraging Machine Learning (ML) and evolutionary algorithms, for the first time, to facilitate early diagnostic predictions of Parkinson's Disease and hypothermia in hospitals, as well as prognostic predictions of optimal postoperative recovery area and of chronic hepatitis. Despite the various pathological aetiologies, the underlying capability of ML-based algorithms to serve as a minimum clinically viable solution for predicting early diagnosis and prognosis has been thoroughly demonstrated. Feature selection (FS) is a proven method for increasing the performance of ML-based classifiers for several applications. Although advances in ML, such as Deep Learning (DL), have denied the usefulness of any extrinsic FS by incorporating it in their architectures, e.g., convolutional filters in convolutional neural networks, DL algorithms often lack the required explainability to be understood and interpreted by clinicians within the context of the diagnostic and prognostic tasks of interest. Their relatively complicated architectures, the hardware required for running them and the limited explainability or interpretability of their architectures, the decision-making process – although as assistive tools - driven by the algorithms’ training and predictive outcomes have hindered their application in a clinical setting. Luca Parisi’s work fills this translational research gap by harnessing the explainability of using traditional ML- and evolutionary algorithms-based FS methods for improving the performance of ML-based algorithms and devise minimum viable solutions for diagnostic and prognostic purposes. The work submitted here involves independent research work, including collaborative studies with Marianne Lyne Manaog (MedIntellego®) and Narrendar RaviChandran (University of Auckland). In particular, conciliating his work as a Senior Artificial Intelligence Engineer and volunteering commitment as the President and Research Committee Leader of a student-led association named the “University of Auckland Rehabilitative Technologies Association”, Luca Parisi decided to embark on most research works included in this synopsis to add value to society via accurate, reliable and explainable, hence clinically viable applications of AI. The key findings of these studies are: (i) ML-based FS algorithms are sufficient for devising accurate, reliable and explainable ML-based classifiers for aiding prediction of early diagnosis for Parkinson’s Disease and chronic hepatitis; (ii) evolutionary algorithms-based optimisation is a preferred method for improving the accuracy and reliability of decision support systems aimed at aiding early diagnosis of hypothermia; (iii) evolutionary algorithms-based optimisation methods enable to devise optimised ML-based classifiers for improving postoperative discharge; (iv) whilst ML-based algorithms coupled with ML based FS methods are the minimum clinically viable solution for binary classification problems, ML-based classifiers leveraging evolutionary algorithms for FS yield more accurate and reliable predictions, as reducing the search space and overlapping regions for tackling multi-class classification problems more effectively, which involve a higher number of degrees of freedom. Collectively, these findings suggest that, despite advances in ML, state-of-the-art ML algorithms, coupled with ML-based or evolutionary algorithms for FS, are enough to devise accurate, reliable and explainable decision support systems for performing both an early diagnosis and a prediction of prognosis of various pathologies.

京都大学 / 新制・課程博士 / 博士(医学) / 甲第24086号 / 医博第4862号 / 新制||医||1059(附属図書館) / 京都大学大学院医学研究科医学専攻 / (主査)教授 井上 治久, 教授 渡邉 大, 教授 高橋 淳 / 学位規則第4条第1項該当 / Doctor of Medical Science / Kyoto University / DFAM

Parkinson’s Disease (PD) is a neurodegenerative disorder, within this categoryof diseases it is among the most prevalent worldwide. The etiology of PD isbased in progressive deterioration of neural tissue in the basal ganglia (neuronalnuclei located at the base of the cerebrum) and their related structures. Current research is focusing on treatment approaches to either enhance or replaceexisting pharmaceutical treatment approaches, such as dopamine replacementtherapy. In this project, the focus was on finding correlates between movementdata and neurological signals to provide insight into potential biomarkers forcomplex motor symptoms of PD. This will in turn provide a starting point forspecifically targeted closed-loop neural stimulation that alleviates these symptoms. Although the data available at the time of this thesis did not providesufficient insight to derive a conclusion on the neural correlates, a pipeline wasdeveloped, which analyzes and synchronizes kinematic and neural data and willenable further exploration as additional data is obtained.

Yes / Background: Riluzole is a neuroprotective drug used in the treatment of motor neurone disease. Recent evidence suggests that riluzole can up-regulate the expression and activity of the astrocyte glutamate transporter, GLT-1. Given that regulation of glutamate transport is predicted to be neuroprotective in Parkinson’s disease, we tested the effect of riluzole in parkinsonian rats which had received a unilateral 6-hydroxydopamine injection into the median forebrain bundle. Results: Rats were treated with intraperitoneal riluzole (4 mg/kg or 8 mg/kg), 1 hour before the lesion then once daily for seven days. Riluzole produced a modest but significant attenuation of dopamine neurone degeneration, assessed by suppression of amphetamine-induced rotations, preservation of tyrosine hydroxylase positive neuronal cell bodies in the substantia nigra pars compacta and attenuation of striatal tyrosine hydroxylase protein loss. Seven days after 6-hydroxydopamine lesion, reactive astrocytosis was observed in the striatum, as determined by increases in expression of glial fibrillary acidic protein, however the glutamate transporter, GLT-1, which is also expressed in astrocytes was not regulated by the lesion. Conclusions: The results confirm that riluzole is a neuroprotective agent in a rodent model of parkinson’s disease. Riluzole administration did not regulate GLT-1 levels but significantly reduced GFAP levels, in the lesioned striatum. Riluzole suppression of reactive astrocytosis is an intriguing finding which might contribute to the neuroprotective effects of this drug.

Over the past few years, several studies have been published to propose algorithms for the automated diagnosis of Parkinson’s Disease using simple exams such as drawing and voice exams. However, at the same time as all classifiers appear to have been outperformed by another classifier in at least one study, there appear to lack a study on how well different classifiers work with a certain feature selection algorithm and sampling method. More importantly, there appear to lack a study that compares the proposed feature selection algorithm and/or sampling method with a baseline that does not involve any feature selection or oversampling. This leaves us with the question of which combination of feature selection algorithm, sampling method and classifier is the best as well as what impact feature selection and oversampling may have on the performance. Given the importance of providing a quick and accurate diagnosis of Parkinson’s disease, a comparison is made between different systems of classifier, feature selection and sampling method with a focus on the predictive performance. A system was chosen as the best system for the diagnosis of Parkinson’s disease based on its comparative predictive performance on two sets of data - one from drawing exams and one from voice exams. / Som en av världens mest vanligaste sjukdom med en tendens att leda till funktionshinder har Parkinsons sjukdom länge varit i centrum av forskning. För att se till att så många som möjligt får en behandling innan det blir för sent har flera studier publicerats för att föreslå algoritmer för automatisk diagnos av Parkinsons sjukdom. Samtidigt som alla klassificerare verkar ha överträffats av en annan klassificerare i minst en studie, verkar det saknas en studie om hur väl olika klassificerare fungerar med en viss kombination av urvalsalgoritm (feature selection algorithm på engelska) och provtagningsmetod. Därutöver verkar det saknas en studie där resultatet från den föreslagna urvalsalgoritmen och/eller samplingsmetoden jämförs med resultatet av att applicera klassificeraren direkt på datan utan någon urvalsalgoritm eller resampling. Detta lämnar oss en fråga om vilket system av klassificerare, urvalsalgoritm och samplingsmetod man bör välja och ifall det är värt att använda en urvalsalgoritm och överprovtagningsmetod. Med tanke på vikten av att snabbt och noggrant upptäcka Parkinsons sjukdom har en jämförelse gjorts för att hitta den bästa kombinationen av klassificerare, urvalsalgoritm och provtagningsalgoritm för den automatiska diagnosen av Parkinsons sjukdom.

Les stratégies thérapeutiques mises en place dans la maladie de Parkinson sont symptomatiques et ne permettent pas de ralentir la progression de la maladie, nécessitant le développement de nouvelles approches neuroprotectrices. La neuroinflammation joue un rôle majeur dans le processus neurodégénératif où elle se manifeste précocement par l’activation de cellules gliales (microglie et astrocytes). En s’appuyant sur l’utilisation de modèles animaux mimant les stades précoces de la maladie, l’élaboration de stratégies thérapeutiques à visée anti-inflammatoire constitue donc une approche thérapeutique prometteuse. Ce travail de thèse a consisté à mettre au point et à caractériser un modèle de lésion partielle à la 6-hydroxydopamine chez le rat afin d’évaluer les effets d’une stratégie thérapeutique originale basée sur l’utilisation en combinaison d’un agoniste des récepteurs nicotiniques α7 et d’un agoniste des récepteurs σ1. En utilisant différentes approches expérimentales, nous avons tout d’abord évalué le processus neurodégénératif et la neuroinflammation dans le modèle que nous avons mis en place. Nos résultats ont montré une dégénérescence partielle et reproductible des neurones dopaminergiques nigro-striataux associée à une importante neuroinflammation. Nos analyses métabolomiques ont également révélé plusieurs altérations spécifiques, apportant ainsi de nouvelles informations sur les mécanismes intervenant dans le processus neurodégénératif. En s’appuyant sur l’utilisation de la tomographie par émission de positrons, nous avons ensuite évalué longitudinalement le profil d’expression des récepteurs nicotiniques α7 dans les structures clés de la voie nigrostriée. Nos résultats ont montré des modifications transitoires de la densité de ces récepteurs pouvant être liées à des réponses microgliales biphasiques en association avec la cinétique de la dégénérescence neuronale. Ainsi, ces résultats renforcent l’idée de cibler spécifiquement les récepteurs nicotiniques α7 dans l’atténuation des processus neuroinflammatoires. Nous avons enfin évalué les effets de notre stratégie thérapeutique dans le modèle et nos résultats ont permis de montrer que ce type de combinaison préserve partiellement l’intégrité des neurones dopaminergiques nigro-striataux et réduit les réactions gliales chez les animaux lésés. Bien qu’il sera nécessaire de confirmer et de compléter ces résultats avec d’autres analyses, ce type de combinaison pourrait constituer une nouvelle entité biochimique prometteuse dans le traitement de la maladie de Parkinson. / Currently, therapeutic strategies in Parkinson’s disease are symptomatic and the progression of the disease is uncontrolled, requiring the development of new neuroprotective approaches. Neuroinflammation plays a major role in the neurodegenerative process where it occurs early through the activation of glial cells (microglia and astrocytes). Based on the use of animal models mimicking the early stages of the disease, the development of anti-inflammatory strategies is therefore a promising therapeutic approach. This thesis work consisted in the development and the characterisation of a partial 6-hydroxydopamine lesion model in rats in order to evaluate the effects of an original therapeutic strategy based on the combined use of a α7 nicotinic receptors agonist and a σ1 receptors agonist. Using different experimental approaches, we first evaluated the neurodegenerative and neuroinflammation processes in the model that we developped. Our results showed a partial and reproductible degeneration of nigro-striatal dopaminergic neurons associated with a marked neuroinflammation. Our metabolic analyses have also revealed several specific alterations, providing new insight on the mechanisms involved in the neurodegenerative process. Using positron emission tomography imaging, we then evaluated longitudinally the expression profile of α7 nicotinic receptors in the key structures of the nigro-striatal pathway. Our results showed transient changes in the density of these receptors that may be linked to biphasic microglial responses in association with the kinetics of neuronal degeneration. Thus, these results reinforce the hypothesis of specifically targeting α7 nicotinic receptors in order to reduce the neuroinflammatory processes. Finally, we evaluated the effects of our therapeutic strategy in the model and our results showed that this type of combination partially preserves the integrity of nigro-striatal dopaminergic neurons and reduces glial reactions in lesioned animals. Although it is necessary to confirm and extend these results, this type of combination could represent a promising new pharmacological approach in the treatment of Parkinson’s disease.

Bakgrund: 20 000 personer i Sverige lever med Parkinsons sjukdom. Parkinsons sjukdom är en kronisk, neurologisk sjukdom som karaktäriseras av motoriska och icke-motoriska symtom. De motoriska symtomen försvårar utförandet av aktiviteter i det dagliga livet. Syfte: Syftet var att beskriva hur personer med Parkinsons sjukdom upplever att de motoriska symtomen påverkar det dagliga livet. Metod: En allmänlitteraturöversikt utifrån tio kvalitativa vetenskapliga artiklar. Resultat: En oförutsägbar kropp, förlorad självständighet och begränsning av det sociala livet var tre huvudkategorier som framkom i resultatet med respektive subkategorier. Resultatet visade att personer med Parkinsons sjukdom uttryckte att kroppen var olika från dag till dag vilket stoppade dem i det dagliga livet. Det ledde till en förändrad identitet och en känsla av förlorad självständighet. Konklusion: En ökad kunskap och förståelse för hur det är för personer att leva med Parkinsons sjukdom, kan ge sjuksköterskor möjlighet att bedriva en bättre personcentrerad vård kring personer med Parkinsons sjukdom. / Background: 20 000 people in Sweden are living with Parkinson’s disease. Parkinson´s disease is a chronic neurologic disease characterised by motor and non-motor symptoms. The motor symptoms complicate the performance of activities in everyday life. Aim: The aim of this study was to describe how people with Parkinson’s disease experience how the motor symptoms affect their everyday life. Method: A general literature study based on ten qualitative scientific articles. Results: An unpredictable body, lost independence and limitations of social life was three main categories that emerged in the result with respective subcategories. The result showed that people with Parkinson’s disease expressed that the body was different from day to day, which stopped them in everyday life. This led to a changed identity and a sense of lost independence. Conclusion: An increased understanding in how people with Parkinson`s disease experience life, can give nurses an opportunity to give a better person-centered care to people with Parkinson´s disease.

Le gène GALIG produit deux protéines, la Mitogaligine et la Cytogaligine. Son expression conduit à la mort cellulaire par un processus encore mal défini. Dans ce cadre, nous nous sommes intéressés principalement à la Cytogaligine et avons précisé qu’elle est localisée dans le cytoplasme et le noyau mais également dans la mitochondrie au niveau de la membrane interne. Un test de complémentation montre que la Cytogaligine interagit avec la Mitogaligine. Cette interaction pourrait être un facteur important dans la mise en place de la voie d’apoptose médiée par l’expression de GALIG. D‘autres protéines interagissant avec la Cytogaligine ont été identifiées, notamment l’α-Synucléine, protéine centrale dans la maladie de Parkinson (MP), qui est connue pour s’agréger dans les cellules et induire la fragmentation des mitochondries. Dans la mesure où la surexpression de l’α-Synucléine conduit à des défauts de l’autophagie et du système Ubiquitine-protéasome dans la MP, nous avons recherché d’éventuels partenaires de la Cytogaligine associés à ces fonctions. De fait, la Cytogaligine interagit, en autre, avec les protéines de l’autophagie LC3B, GABARAP, p62/SQSTM1, la protéine chaperon Hsc70 ainsi que les protéines du système UPS, HUWE1 et UBQLN4. Ces résultats ouvrent de nouvelles pistes sur les conséquences fonctionnelles de l’expression de la Cytogaligine. Dans une deuxième partie, nous avons réalisé une étude clinique visant à évaluer le profil d’expression des gènes précédemment étudiés dans les cellules mononuclées du sang périphérique de patients atteints de la MP. Si l’expression du gène GALIG ne présente pas de variations entre les patients et les contrôles, une dérégulation de l’expression de différents gènes associés au processus de l’autophagie est mise en évidence. Parmi ces données, celles combinant l’expression du couple de gènes LC3B et GAPDH pourraient représenter un marqueur potentiel de la maladie dans le cadre d’un test diagnostic non invasif. / The GALIG gene produces two proteins, Mitogaligin and Cytogaligin. GALIG expression induces cell death by a still poorly defined process. In this context, we focused mainly on Cytogaligin and specified that it is localized in cytoplasm and nucleus but also in mitochondria close to the inner membrane. A functional complementation test showed that Cytogaligin interacted with Mitogaligin. This interaction could be an important factor in the establishment of the apoptosis pathway mediated by GALIG expression. Other proteins interacting with Cytogaligin have been identified, including α-Synuclein, a central protein in Parkinson's disease (PD), which is known to aggregate in cells and induce fragmentation of mitochondria. Since overexpression of α-synuclein leads to autophagy and Ubiquitin-proteasome system disruptions, we have looked for potential Cytogaligin partners associated with these functions. Cytogaligin interacted with the autophagy proteins LC3B, GABARAP, p62/SQSTM1, the chaperone protein Hsc70 as well as the UPS system proteins HUWE1 and UBQLN4. These results open perspectives regarding the functional consequences of the expression of Cytogaligin. In a second part, we carried out a clinical study aimed at evaluating the expression profile of the previously studied genes in the peripheral blood mononuclear cells of PD patients. If the expression of the GALIG gene does not show variations between PD patients and controls, deregulation of the expression of genes associated with autophagy was highlighted. Among these data, those combining the expression of the two genes LC3B and GAPDH could represent a potential marker of the disease as a non-invasive diagnostic test.

Base teórica: A doença de Parkinson (DP) é a segunda enfermidade neurodegenerativa mais frequente, crescendo proporcionalmente com o aumento da idade. É uma doença de comprometimento motor e não motor. Levodopa e agonistas dopaminérgicos (AD) são usados no tratamento da DP, permitindo um controle ótimo dos sintomas nos primeiros anos. Entretanto, em 5 anos, metade dos pacientes terão complicações motoras e nãomotoras induzidas pelo uso de antiparkinsonianos. Manifestações neuropsiquiátricas são frequentes, entre elas depressão, ansiedade, prejuízos cognitivos, sintomas psicóticos e transtorno de descontrole dos impulsos(DI). O DI é uma condição caracterizada pela falência em resistir a impulsos ou tentação de executar atos. O DI está associado ao uso de antiparkinsonianos, em especial, os agonistas dopaminérgicos A identificação desta condição é primordial para seu tratamento e estudo adequados. Na literatura atual, o questionário padrão-ouro foi validado na língua inglesa (QUIP), não existindo uma validação para língua portuguesa. Objetivo: Traduzir e validar o questionário QUIP-CS, em sua versão curta e aplicável no momento presente da DP, para lingua Portuguesa do Brasil Métodos: A versão curta da QUIP (QUIP-CS) foi traduzida para o Português por tradutor juramentado. Após, esta foi avaliada por 5 especialistas em DP no Brasil, sendo sugeridas pequenas correções. A versão corrigida em português foi retrotraduzida para o inglês por 2 tradutores juramentados nativos na língua original da escala (inglês), que compararam suas versões posteriormente, chegando-se a uma nova versão final neste idioma. Esta foi enviada ao autor da escala original, que concordou com esta versão, ou seja, foram mantidas as propriedades semânticas do instrumento. Após, a versão final em Português foi auto-aplicada em 65 indivíduos com diagnóstico de DP em tratamento no ambulatório especializado no HCPA, sendo que, de forma aleatória, para 30 foi aplicado um questionário de avaliação sobre o grau de dificuldade de compreensão de suas perguntas. Resultados: Em uma escala de 1 a 5 pontos, onde 1 era nenhuma compreensão das perguntas e 5, clara compreensão, a média de entendimento pelos pacientes foi de 4,06 +/- 0,69 DP. Conclusão: A avaliação desta versão foi considerada de fácil compreensão pelos próprios pacientes. O artigo para validação da tradução do conteúdo da versão em Português da QUIP-CS está em fase de revisão para publicação. / Background: Parkinson’s disease (PD) is the second most often neurodegenerative disease and proportionally growing with people aging. PD is a disease with motor and nonmotor clinical features. Levodopa and dopaminergic agonists (DA) are used for PD treatment, allowing an exquisite control of the motor symptoms during the first years. However, in five years, half patients will present motor or non-motor complications induced by cronic use of these medications. Neuropsychiatric symptoms are often, for example, depression, anxiety, cognitive impairment, psychotic symptoms and impulse control disorders (ICD). The ICD is characterized by failure on resisting an impulse or on performing an specific act. Identification of the PD affected patients is crucial for proper management and study of this condition. There is an already validated self-reported questionnaire for this purpose, the Parkinson's Disease Impulsive-Compulsive Disorders Questionnaire (QUIP), without equivalent in portuguese language. Objective: To translate and to validate the Portuguese short version of the gold-standard questionnaire for identifying ICDs PD affected patients, applicable at the current moment of PD Methods: QUIP-CS was first translated to Portuguese by a professional translator. This translated version was shown to 5 PD neurologist specialists. in Brazil, being suggested minor modifications on it. This new Portuguese revised version was back translated to English by two independent native English speakers. They were both asked to compare the version one another and checked for differences. Then, they contacted each other and got a final back translated version. This one was sent for the original author, that approved its new version comparing to his original and validated one, with no loss of it’s original properties. The Portuguese corrected version was applied to 65 patients in a random way at PD’s ambulatory at HCPA. From these, 30 were asked to answer a number that would represent their level of QUIP-CS questions’ comprehension. Results: In a 1 to 5 point scale, being 1 no comprehension and 5, total comprehension, the average was 4,06 +/- 0,69 DP. Conclusion: Our results on Portuguese version of QUIP-CS show that QUIP-CS translated and corrected version was easily understood and easily self-applied. The article is under revision to be submitted for publication.

With 9.9 million new dementia cases each year, Alzheimer’s and Parkinson’s disease (AD and PD) are the most prevalent form of neurodegenerative disorder (NDG) affecting the aging population. Despite years of pharmaceutical research, no cure is yet available. Most neuropathological aspects of these diseases are extremely complex but the study of the rare genetic cases allowed to model these diseases in animals and uncover key pathophysiological processes. Transgenic Drosophila NDG models have been used for in vivo studies for many years with a range of relevant phenotypes. The cellular and molecular biology of the Central Nervous System, as well as the mechanisms underlying neurodegeneration, are well conserved between Drosophila and Humans (with a 75% of human disease-related genes having homologs in flies). Most NDG studies are performed in the aging flies. However, there are reports of measurable phenotypes for a variety of AD and PD models in juvenile Drosophila melanogaster (larval stage) with an unexploited considerable potential for drug discovery and screening for this outstanding model. Here I sought to develop a new assay for research into NDGs that focus on the earliest phenotypes. During this Ph.D. project a customized crawling assay apparatus was developed, for the assessment of locomotor ability in humanised larval Drosophila (overexpressing human proteins/peptides linked to AD and PD). A locomotor phenotype was identified in larvae overexpressing different variation of Amyloid-β42, tau and α-Synuclein pan neutrally: these animals crawl on agarose surface at a reduced mean speed when compared to controls. The defect was proven partially rescuable by administration of Tacrine and Methylene Blue, renewing the importance of such models for future applications in drug discovery and screening. The motor impairment supports the hypothesis of a neurotoxic effect of the protein/peptide. Thus, to test this further, the overexpression of the human transgenes was restricted to neurons involved in larval olfaction (olfactory impairment is often the earliest symptom in PD and AD) and odour associated learning tasks (both PD and AD are characterized by severe cognitive dysfunction). Interestingly, larvae overexpressing the Amyloid-β42 ARC peptide in the Olfactory Sensory Neurons showed a subtle navigation defect during chemotaxis (in 1-Hexanol odour gradient) that could possibly be addressed to premature neural habituation to the olfactory stimulus. Furthermore, the overexpression of the peptide in the larval Mushroom Bodies influenced the performances of the animals in associative learning tasks. Lastly, using immunohistochemistry and confocal imaging techniques I showed that the gross morphology of neurons is not altered by the targeted overexpression of the Amyloid-β42 ARC. Even though physiological studies are required to characterize the chemosensory/learning defect shown by the Amyloid-β42 ARC larvae, this Ph.D. work further confirms that the effects of the overexpression of the human transgenes are robust and measurable already at larval stage. These findings may also be relevant to the development of new, fast, and cost-effective compound screening procedures, for applications in early stages of the drug discovery process.