Brage NMBU
Not a member yet
19188 research outputs found
Sort by
Seasonal and Demographic Patterns of Red Deer Habitat Selection: Influence of Extrinsic and Intrinsic Factors
Full text linkUnderstanding how animals select habitats is fundamental for effective wildlife management, especially in predator-free environments shaped by human activity. Habitat selection is influenced by both spatial and temporal factors, which can shift in response to ecological pressures and anthropogenic disturbance. In this study, I explored seasonal and sex-specific patterns of habitat selection in red deer (Cervus elaphus) using a year-long camera trap dataset from Svanøy, a 10 km² predator-free island off the coast of western Norway, inhabited by approximately 50 residents.
The results revealed that both males and females exhibited similar patterns of habitat use across seasons, with no distinct sex-based differences. Both sexes showed strong use of open areas, such as infield pastures and arable land, particularly during the fall. No sex-specific preference for forested habitats was observed, especially in the winter months. However, spatial use was higher in remote areas with greater forest cover, particularly near forest edges, transitional zones that provide both cover and access to open foraging areas. A marked decline in red deer activity was observed following the onset of the hunting season, likely indicating behavioral avoidance of high-risk, easily accessible areas. This pattern was further supported by a significant spatial use of areas with greater cover and greater distances from trails, roads, and forest edges during the 10-day period both before and during the start of the hunting season. This response was evident in both sexes, suggesting a general risk- avoidance strategy rather than distinct sex-specific differences in seasonal habitat selection.
Contrary to expectations, there is little evidence of strong seasonal habitat selection in red deer on Svanøy. Instead, they show flexible spatial use of forest features (cover and edges), proximity to roads and trails, and overall use open habitats. This pattern appears generalized and consistent across sexes, with limited variation in response to seasonal changes.
Although these patterns are shaped by the low-disturbance conditions of Svanøy, they offer valuable insights for managing red deer in more complex, heavily used landscapes. The findings underscore the need for adaptive management strategies that incorporate spatial refuges and seasonal behavioral changes. Strategies such as rotating hunting pressure, protecting critical forest habitats during sensitive periods, and adjusting monitoring protocols to capture temporal variation in detection could enhance sustainable red deer management. Future research should focus on diel activity patterns and movement using GPS and behavioral data to assess the costs associated with human-induced risk avoidance
Stormwater management of road runoff : purification effect of a delaying drainage pond at rv. 4
Full text linkEn nyetablert fordrøyningsdam ved riksvei 4 er et pilotprosjekt knyttet til utbyggingen av den nye riksveien i Roa og skal håndtere overflateavrenning fra både trafikkert vei og jordbruksarealer. Formålet er å undersøke hvordan dammen påvirker vannkvaliteten gjennom en vekstperiode fra 28. mai til 3. september 2024. Hovedhypotesen er at fordrøyningsdammen holder tilbake partikler og næringsstoffer som endrer vannkvaliteten mellom oppstrøms og nedstrøms stasjoner. Det ble gjennomført jevnlig prøvetaking av vann annenhver uke fra seks prøvetakningsstasjoner. Analysene omfatter total suspendert stoff (TSS), pH og konduktivitet, fosfor, nitrogen, løst organisk karbon (DOC) og utvalgte metaller blant annet jern (Fe), mangan (Mn), aluminium (Al), sink (Zn), kobber (Cu), kadmium (Cd) og krom (Cr). I tillegg ble hydrologiske forhold som nedbør, vannstand og vannføring dokumentert.
Resultatene viser at fordrøyningsdammen har en viss tilbakeholdende effekt på total nitrogen (TN), med lavere gjennomsnittskonsentrasjoner i utløpet enn i innløpet. Når det gjelder total fosfor (TP) og metaller er tilbakeholdelsen mer ukjent, da den i større grad påvirkes av eksterne faktorer som tilførsel fra omkringliggende arealer, sedimentdynamikk og sesongmessige variasjoner. Tilsvarende variasjon er observert for TSS. Dette indikerer at dammens funksjon er styrt av faktorer som vannføring, oppholdstid og tilførsel fra sedimenter og at punktobservasjoner av innløp og utløp alene ikke gir tilstrekkelig grunnlag for å beregne totale mengder med partikler som går inn og ut av dammen.
Basert på resultatene kan hovedhypotesen anses som delvis bekreftet. Fordrøyningsdammen viser en tydelig evne til å tilbakeholde nitrogen, noe som understreker dens rolle i å redusere næringsstoffbelastningen i vannet. De observerte forskjellene i vannkvalitet mellom oppstrøms og nedstrøms bekrefter dammens funksjon som både en buffer og en aktiv del av i transport- og tilbakeholdelsesprosesser. Samlet sett bidrar fordrøyningsdammen til å regulere transport av partikler og stoffer, men effekten varierer med sesong, hydrologiske og geokjemiske forhold over tid.A newly established sedimentation pond along highway. 4 is part of a pilot project connected to the development of the new highway in Roa and is designed to manage surface runoff from both heavily trafficked roads and adjacent agricultural land. The purpose is to investigate how the pond affects water quality over a growing season, from May 28 to September 3, 2024. The main hypothesis is that the sedimentation pond retains particles and nutrients, thereby altering water quality between upstream and downstream stations. Water sampling was conducted every other week from six designated sampling stations. The analyses included total suspended solids (TSS), pH and conductivity, phosphorus, nitrogen, dissolved organic carbon (DOC), and selected metals such as iron (Fe), manganese (Mn), aluminum (Al), zinc (Zn), copper (Cu), cadmium (Cd), and chromium (Cr). In addition, hydrological parameters such as precipitation, water level, and discharge were documented.
The results indicate that the sedimentation pond has a certain retention effect on total nitrogen (TN), with lower average concentrations observed at the outlet compared to the inlet. However, the retention of total phosphorus (TP) and metals is more uncertain, as it appears to be more strongly influenced by external factors such as inputs from surrounding land areas, sediment dynamics and seasonal variation. Similar patterns were observed for TSS. This suggests that the pond's function is governed by factors like flow conditions, hydraulic retention time, and inputs from sediments. Consequently, point observations between input and output alone are insufficient to determine the total particle load entering and leaving the sedimentation pond.
Based on the results, the main hypothesis can be considered partially confirmed. The sedimentation pond demonstrates a clear ability to retain nitrogen, highlighting its role in reducing nutrient loads in the water. The observed differences in water quality between upstream and downstream stations confirm the pond's function as both a buffer and an active component in transport and retention processes. Overall, the sedimentation pond contributes to regulating the transport of particles and substances, but the effect varies with seasonal, hydrological, and geochemical conditions over time.M-MILJ
Classification Confidence Visualization of Artificial Neural Networks with Adversarial Robustness
Full text linkThis thesis addresses the limitations of Deep Neural Networks, focusing on their vulnerability to adversarial attacks and lack of interpretability. Standard training methods, such as Backpropagation, often lead to overconfident and fragile models. To improve robustness and understanding, the study explores Bidirectional Learning and a hybrid approach, which combines discriminative and generative training. The objective is to compare these strategies across clean, noisy, and adversarial settings using both quantitative metrics and visual tools, such as principal component analysis. Additionally, the study examines whether Gaussian-based input expansion enhances robustness, especially in synthetic datasets
Lønnsom og bærekraftig rehabilitering av 70-tallsblokker
Full text linkDenne masteroppgaven har undersøkt hvordan private boligeiere i boligblokker fra 1970-tallet kan ta beslutninger i rehabiliteringsprosesser som både er bærekraftige og økonomisk forsvarlige. Gjennom en litteraturstudie og analyse av referansebygget Orebakken Borettslag er det kartlagt hvordan fem tverrfaglige hensyn: energieffektivisering, klimafotspor, inneklima, arkitektonisk kvalitet og økonomi kan balanseres for å oppnå en bærekraftig rehabilitering.
En stor del av blokkene fra 1970-tallet har nådd en alder på 50 år og er moden for rehabilitering, ettersom den tekniske levetiden nærmer seg slutten. Dette gir et mulighetsvindu for å gjennomføre bærekraftige tiltak. Funnene viser at de mest bærekraftige tiltakene, de som gir størst energibesparelse og best inneklima, ikke er privatøkonomisk lønnsomme uten ekstern støtte. Selv med støtte kan tiltakene fortsatt være ulønnsomme, noe som skaper betydelige gjennomføringsbarrierer. Oppgaven påpeker behovet for videreutvikling av merkostnaden for å tydeliggjøre skillet mellom nødvendig rehabilitering og tilleggsinvesteringer for bærekraftige tiltak. For å overkomme barrierene knyttet til usikre innvesteringer, fremhever oppgaven viktigheten av et styrket beslutningsgrunnlag. Styret i borettslag spiller en nøkkelrolle som tillitsskapere og må selv ha god forståelse for rehabiliteringsprosessen for å kunne formidle troverdig og relevant informasjon.
Oppgaven understreker at tverrfaglig tilnærming gir et viktig rammeverk for å strukturere og prioritere tiltak basert på helhetlig bærekraft. Det pekes også på at teknologisk utvikling, som modulbaserte prefabrikkerte løsninger med integrerte tekniske systemer, kan revolusjonere rehabilitering ved å tilby ferdige, effektive og bærekraftige løsninger som ivaretar alle fem fokusområder samtidig. Slike moduler vil, dersom de introduseres i det norske markedet, kunne redusere kompleksiteten i planlegging og gjennomføring betydelig.
Samlet konkluderer oppgaven med at selv om det er betydelige utfordringer, finnes det muligheter for å fremme bærekraftige og helhetlige rehabiliteringer i form av energioppgraderinger. Forutsatt at boligeierne får støtte gjennom kunnskap, verktøy og konkrete insentiver som gjør det mulig å ta langsiktige og verdiskapende valg.This master’s thesis has examined how private homeowners in 1970s apartment blocks can make decisions in rehabilitation processes that are both sustainable and economically viable. Through a literature review and an analysis of the reference building Orebakken Housing Cooperative, the study has mapped how five interdisciplinary considerations, energy efficiency, climate footprint, indoor climate, architectural quality, and economy, can be balanced to achieve sustainable rehabilitation.
A large proportion of apartment blocks from the 1970s have now reached an age of 50 years and are due for rehabilitation, as their technical lifespan is nearing its end. This presents a window of opportunity for implementing sustainable measures. The findings show that the most sustainable measures, those that offer the greatest energy savings and best indoor climate, are not economically profitable for individual homeowners without external support. Even with support, such measures may remain unprofitable, creating significant barriers to implementation. The thesis highlights the need for further development of cost differentiation to clearly distinguish between necessary rehabilitation and additional investments in sustainable measures. To overcome the barriers related to uncertain investments, the thesis emphasizes the importance of a strengthened decision-making basis. The board of the housing cooperative plays a key role in building trust and must have a solid understanding of the rehabilitation process in order to communicate credible and relevant information.
The thesis highlights that an interdisciplinary approach provides an important framework for structuring and prioritizing measures based on holistic sustainability. It also points to technological developments, such as modular prefabricated solutions with integrated technical systems, as having the potential to revolutionize rehabilitation by offering ready-made, efficient, and sustainable solutions that address all five focus areas simultaneously. If introduced to the Norwegian market, such modules could significantly reduce the complexity of planning and implementation.
In conclusion, the thesis finds that although there are considerable challenges, there are also opportunities to promote sustainable and holistic rehabilitations in the form of energy upgrades, provided that homeowners are supported with knowledge, tools, and concrete incentives that enable them to make long-term, value-creating decision
Impact of Flood on Livelihoods of People living in Rural Areas of Pakistan
Full text linkFlooding is posing a more and more severe threat to rural livelihoods in agrarian economies, especially in the Kyber Pakhtunkhwa province of Pakistan, where climatic variability and strengthened monsoons meet the hydrological complexities of the Swat River basin. The study investigates the numerous impacts of floods on sustainable rural livelihoods in Kyber Pakhtunkhwa, with special reference to Swat Valley flood-prone and agriculturally dependent district. Using the Sustainable Livelihoods Framework (SLF), the research assesses the effects of floods on the five livelihood capitals: human, social, natural, physical, and financial.
The policy responses remain patchy and short-term in nature, despite the repeated occurrence of floods and their ever-increasing socio-economic consequences. A qualitative approach of semi-structured interviews with local villagers, NGO representatives, and government officials allowed for capturing qualitative data and contextual grounding regarding rural resilience and adaptation strategies. The degradation along all dimensions of livelihood capitals could be experienced with an immediate impact upon natural and financial assets for crop loss and income disruption. The rural populations are further vulnerable owing to poor infrastructure and little institutional support, which leads them into a vicious circle of debt, displacement, and disenfranchisement.
Further evaluation is carried out on the efficacy of present coping strategies and support interventions in addressing the gaps for long-term resilience planning. Through the triangulation of data sources and analytical application of the SLF, this research provides policy-relevant recommendations that promote sustainable adaptation strategies moving beyond reactive relief toward proactive risk reduction.
Therefore, this research adds to the climate-resilience discourse in South Asia and carries the burden of urgent rural development policies, which must be thought of in terms of climate, equity, and being rooted in the local contex
Planlegging av sensorer for overvåkning av klima mellom solcellepanelene i et APV-anlegg ved NMBU
Full text linkAgriphotovoltaics forkortet agriPV eller APV er en betegnelse på kombinasjonen
av landbruk og solenergi på samme areal. Ideen bak APV ble
først introdusert tidlig på 1980-tallet av Goetzberger og Zastrow [1], som
la grunnlaget for teknologien ved å foreslå bruk av dyrkbar mark mellom
rader med skråstilte solcellepaneler. Innen 2025 har APV-teknologien utviklet
og spredt seg over store deler av Europa i ulike utforminger [2].
Studier fra flere europeiske land viser at APV representerer en vinnvinn-
vinn-løsning for både landbrukssektoren, energisektoren og samfunnet
som helhet. APV kan gi økte inntekter for bønder og mer effektivt
arealbruk uten inngrep i uberørt natur [2].
APV-anlegget brukt som utgangspunkt i denne oppgaven består av
vertikalmonterte, tosidige solcellepaneler, og er montert ved Norges miljøog
biovitenskapelige universitet (NMBU). Målet med denne studien er å
identifisere hvilke type sensorer som trengs for å overvåke mikroklima
mellom panelradene, og vurdere ytelsen til anlegget som helhet.
Gjennom litteraturstudie av rapporter fra liknende APV-anlegg, kommunikasjon
med forskere fra NMBU og eksterne kontakter med relevant
erfaring, har det blitt undersøkt hvilke type sensorer som er egnet til måling
av mikroklima. Kommunikasjon med kontaktene fra NMBU har blitt
ekstra vektlagt da de har lang erfaring med måling av klima gjennom den
meteorologiske stasjonen på Søråsjordet [3].
Det skal måles solinnstråling i øst-vest-retning og celletemperatur på
panelene for å vurdere panelenes effektivitet. I tillegg skal det måles lufttemperatur
og -fuktighet, jordtemperatur og -fuktighet, vindhastighet og
-retning, albedo og fotosyntetisk aktiv stråling (PAR). Ved hjelp av en
LIDAR montert i nordenden av anlegget måles snødybde og gresshøyde.
Ved å installere passende sensorer for nøyaktige målinger som logger
kontinuerlig, vil det i fremtiden være mulig å trekke konklusjoner om hvordan
vertikalmonterte solcellepaneler og plantevekst påvirker hverandre i
nordisk klima.Agriphotovoltaics, abbreviated as agriPV or APV, refers to the combination
of agriculture and solar energy production on the same land
area. The concept of APV was first introduced in the early 1980s by
Goetzberger and Zastrow [1], who laid the foundation for the technology
by proposing the use of arable land between rows of tilted solar panels.
By 2025, APV has evolved and spread across much of Europe in various
configurations [2]. Studies from several European countries indicate
that APV represents a win–win–win solution for the agricultural sector,
the energy sector, and society at large. The technology aims to increase
farmers’ income and improve land-use efficiency without intervention on
untouched natural areas [2].
The APV system used as the basis for this thesis consists of vertically
mounted bifacial solar panels and is installed at the Norwegian University
of Life Sciences (NMBU). The aim of this study is to identify which types
of sensors are needed to monitor the microclimate between the panel rows
and assess the overall performance of the system.
Through a literature review of reports from similar APV installations,
as well as communication with researchers at NMBU and external experts
with relevant experience, suitable sensor types for microclimate monitoring
have been evaluated. Particular emphasis was placed on input from
NMBU contacts due to their experience with climate measurements at
the meteorological station on Søråsjordet [3].
Solar irradiance in the east–west direction will be measured, along
with panel cell temperature to evaluate efficiency. Additional measurements
include air temperature and humidity, soil temperature and moisture,
wind speed and direction, albedo, and photosynthetically active
radiation (PAR). A LIDAR sensor will be installed at the north end of
the facility to monitor snow depth and grass height.
By installing suitable sensors that enable continuous and accurate
data logging, it will be possible in the future to draw conclusions about
how vertically mounted solar panels affect, and are affected by, crop
growth under Nordic climate conditions
Operasjonell optimalisering av lakseoppdrett gjennom analyse av oksygenvariasjoner
Full text linkOppdrettsnæringen står overfor økende utfordringer knyttet til miljøstress, hvor lave oksygennivåer kan påvirke både fiskevelferd og produksjonsøkonomi. Denne masteroppgaven undersøker hvordan variasjoner i oksygenforhold kan modelleres og utnyttes for å optimalisere drift, særlig i perioder med økt risiko. Analysen bygger på data fra én produksjonssyklus ved én lokalitet hos en norsk oppdrettsbedrift.
Resultatene viser at lavt oksygennivå ofte sammenfaller med redusert tilvekst og dårligere fôrutnyttelse, og at slike forhold gjerne særlig oppstår i perioder eller merder med høy biomassetetthet og ugunstige strøm- og temperaturforhold. Gjennom enkle simuleringer og vurderinger av strategier indikeres det at fôringsstrategier tilpasset oksygenforhold kan redusere både risiko og kostnader.
Oppgaven konkluderer med at bedre forståelse av miljøvariasjoner gir mulighet for mer presise og lønnsomme valg i oppdrettsnæringen. Det anbefales videre arbeid for å validere funnene på tvers av lokaliteter, men studien peker på et tydelig potensial for forbedring innen både fiskevelferd og økonomisk ytelse.
Studien bidrar til økt bevissthet rundt miljømessige produksjonsgrenser i oppdrett og viser hvordan kunnskapsbasert tilpasning til naturlige variasjoner kan være et viktig verktøy i arbeidet mot mer bærekraftig og robust havbruk. Fremtidige strategier bør integrere miljødata tettere i driftsplanlegging og beslutningsstøtte.The aquaculture industry is facing increasing challenges related to environmental stress, where low oxygen levels can impact both fish welfare and production economics. This master´s thesis explores how variations in oxygen conditions can be modeled and utilized to optimize operations, particularly during periods of elevated risks. The analysis is based on data from one production cycle at a single site operated by a Norwegian aquaculture company.
The results show that low oxygen levels often coincide with reduced growth and feed conversion, especially in periods or cages with high biomass density and unfavorable current or temperature conditions. Through simple simulations and strategy assessments, the study indicates that feeding strategies adapted to oxygen conditions can reduce both risks and cost.
The thesis concludes that improved understanding of environmental variation enables more precise and profitable decisions. Further work is recommended to validate the findings across multiple sites, but the study highlights a clear potential for improvement in both fish welfare and economic performance.
The study contributes to greater awareness of environmental limits in aquaculture and demonstrates how knowledge-based adaptation to natural variability can serve as a valuable tool in the pursuit of more sustainable and resilient farming practices. Future strategies should integrate environmental data more closely into operational planning and decision support
Effekten av naturbaserte tiltak for å redusere kjelleroversvømmelse: En studie med fysisk modell og SWMM
Full text linkDagens ledningsnett har kapasitetsbegrensning, og med en økende nedbørsmengde blir den tradisjonelle tanken om at overvann skal ledes bort i ledninger utfordret. Det er viktig å finne gode løsninger for å håndtere overvann lokalt, hvis ikke er det å forvente store skadekostnader. Lokale løsninger for håndtering av overvann kan være naturbaserte tiltak. Imidlertid er det ennå et noe manglende kunnskapsgrunnlag av hvilken effekt naturbaserte tiltak har.
I undervisningen i dag introduseres flere datamodelleringsprogrammer, men de færreste blir knyttet opp til en fysisk modell. Norges miljø- og biovitenskapelige universitet har bygd en modellby under en nedbørssimulator. I denne oppgaven skal det sees på om naturbaserte tiltak kan ha en effekt på nivået på en kjelleroversvømmelse i en modellby. I tillegg skal det kalibreres en SWMM-modell, som skal kunne beskrive modellbyen.
Det er gjennomført forsøk av modellbyen både med og uten naturbaserte tiltak. Naturbaserte tiltak har en effekt på nivået på kjelleroversvømmelsen i modellbyen. Resultatene fra forsøkene viser at en regnhendelse på 4,7 mm over et minutt gir en kjelleroversvømmelse på 37 mm. Ved å etablere 25 naturbaserte tiltak som grønne tak, regnbed og permeabelt dekke reduseres oversvømmelsen til 21 mm. I tillegg blir det kalibrert en SWMM-modell, som ikke klarer å beskrive nivået på kjelleroversvømmelsen i modellbyen godt.Today’s pipeline network has capacity limitations, and with increasing rainfall, the traditional idea that stormwater should be directed into pipes is being challenged. It is important to find effective solutions for managing stormwater locally; otherwise, significant damage costs can be expected. Local solutions for stormwater management can include nature-based measures. However, there is still a somewhat lacking knowledge base regarding the effectiveness of such measures.
In today’s education, several data modeling programs are introduced, but few are connected to a physical model. The Norwegian University of Life Sciences has built a model city under a rainfall simulator. This thesis investigates whether nature-based measures can impact the level of basement flooding in the model city. Additionally, a SWMM model will be calibrated to describe the model city.
Experiments have been conducted on the model city both with and without nature-based measures. Nature-based measures do have an effect on the level of basement flooding in the model city. The results from the experiments show that a rainfall event of 4.7 mm over one minute causes a basement flood of 37 mm. By implementing 25 nature-based measures such as green roofs, rain gardens, and permeable surfaces, the flooding is reduced to 21 mm. Furthermore, a SWMM model was calibrated, but it failed to accurately describe the level of basement flooding in the model city
Helsefremming gjennom hundeassistert behandling: Behandlernes erfaringer
Full text linkSammendrag
Bakgrunn: Helsefremming er en måte å jobbe på der fokuset rettes mot faktorer som kan styrke ressursene for god helse. Jevnlige interaksjoner med en hund, enten gjennom å eie en hund eller gjennom langsiktige intervensjoner, er blitt assosiert med positive psykologiske utfall for mennesker. Dyreassisterte tiltak er ulike interaksjoner som tilrettelegges av fagfolk eller andre utøvere med formål om å styrke velferden til mennesker gjennom positiv kontakt med dyr. Hund er et av dyrene som ofte inkluderes i disse tiltakene, blant annet basert på at de er lette å trene og er spesielt gode på å forstå og kommunisere med mennesker. Denne masteroppgaven fokuserer på hundeassistert behandling. Dyreassistert behandling er en variant av dyreassisterte tiltak, som utføres av helsepersonell eller av andre utøvere med oppføling av helsepersonell, der dyr involveres i terapeutiske settinger, ofte som et supplement til annen behandling.
Formål: Formålet med denne masteroppgaven er å få kjennskap til hva de som utøver hundeassistert behandling erfarer om den mulige helsefremmende betydningen for de som mottar disse tiltakene.
Metode: Kvalitativ metode ble brukt for å få frem behandlernes erfaringer. Det ble gjennomført åtte semistrukturerte intervjuer med behandlere innen hundeassistert behandling. For å analysere datamaterialet ble systematisk tekstkondensering (STC) benyttet.
Resultater: Gjennom analyseringen av datamateriale kom jeg frem til følgende hovedtemaer: Enklere å etablere kontakt, stressdempende, positive følelser, en kommer styrket ut, styrker det sosiale fellesskapet, dyrevelferd i sentrum og å være forberedt.
Konklusjon: Behandlerne opplevde den hundeassisterte behandlingen som helsefremmende på mange ulike måter og for flere av brukergruppene de jobber med. Men var samtidig opptatt av at noen faktorer måtte være på plass for å få til «det gode resultatet». Spesielt det å ta vare på hunden i tiltaket ble trukket frem som viktig for å oppnå den effekten man ønsker av behandlingen.Abstract
Background: Health promotion is a way of working that focuses on factors that can strengthen the resources for good health. Regular interactions with a dog, either through owning a dog or through long-term interventions, have been linked with positive psychological outcomes for people. Animal-assisted services are various interactions that are facilitated by professionals or other practitioners with the aim of strengthening the welfare of people through positive contact with animals. Dogs are one of the animals that are often included in these interventions, some of the reasons for this is that they are easy to train and particularly good at understanding and communicating with people. This master’s thesis focuses on dog-assisted treatment. Animal-assisted treatment is a variant of animal-assisted interventions, which are carried out by health personnel or by other practitioners with supervision by health personnel, where animals are included in therapeutic settings, often as a supplement to other treatments.
Purpose: The purpose of this master's thesis is to gain knowledge about what those who practice dog-assisted treatment experience about the possible health promoting benefits for those who receive these interventions.
Method: Qualitative method was used to highlight the practitioners’ experiences. Eight semi-structured interviews were conducted with practitioners in dog-assisted treatment. Systematic text condensation (STC) was used to analyze the data.
Results: Through the analysis of the data, I identified the following main themes: easier to establish contact, stress relief, positive emotions, emerging stronger, strengthens the social community, focus on animal welfare and the importance of preparation.
Conclusion: The providers found dog-assisted treatment to be health promoting in many ways and for several of the groups they work with. However, they also highlighted that some factors had to be in place to achieve a successful outcome. In particular, ensuring the dog’s well-being during the intervention was seen as important to achieve the desired outcome of the treatment
Deep Learning for Automatic Diagnosis of Canine Hip Dysplasia
Full text linkArtifical Intelligence (AI), particularly Deep Learning (DL) has
recently become widely used in various fields. In the realm of medical
imaging and healthcare, AI enhances diagnostic accuracy and efficiency,
leading to improved disease monitoring, treatment, and overall patient
outcomes. A recurring challenge is the black-box issue, DL models
process and analyze vast amounts of data to produce an output. Still,
the reasoning behind the output often remains unclear, leading to an
incomplete understanding of the decision-making process. Therefore, the
main goal of this master thesis was to explore methods for uncertainty
and interpretability in the context of using DL for the automatic
diagnosis of canine hip dysplasia (HD).
HD is a hereditary, developemental defect in the hip joint. HD is
graded on a scale from A-E, where A and B are considered normal,
and C, D and E are considered abnormal. The disease is a part of
the Norwegian Kennel Club (NKK) screening program. This thesis
investigated the possibility of using EfficentNet B1 to B7 and Efficent-
NetV2 S, M, L models, a family of Convolutional Neural Networks
(CNNs) architectures, to classify hip radiographs as normal or abnormal.
The dataset included 5369 images of one-sided hips, preprocessed by
veterinarians from a larger dataset provided by the NKK. The dataset
was divided into training, validation, and test data at the ratios of 60%,
20%, 20%, respectively. The highest performing model was the ensemble
model, with an accuracy (ACC) of 0.913 and a Matthews Correlation
Coefficient (MCC) of 0.820 on the test set. An average performance
score (APS) was also calculated, yielding a value of 0.887 on the test data.
Six methods were considered for uncertainty analysis: probability
range, entropy, standard deviation, majority vote, test time augmenta-
tion (TTA) and Monte Carlo dropout (MC). Standard deviation, TTA
and MC were the most effective uncertainty methods for flagging correct
predictions as certain and flagging incorrect ones as uncertain. Regard-
ing interpretability, four methods, Variance of Gradients (VarGrad), Lo-
cally Interpretable Model-Agnostic Explainer (LIME), Gradient-weighted
Class Activation Mapping (Grad-CAM) and Excitation Backprop (EB),
were explored. VarGrad and LIME were chosen for their ability to pro-
duce the most consistent heatmaps and highlight the correct regions of
the images. However, since there are no numerical results to support a
decision on interpretability approaches, it is more up to each individ-
ual veterinarian’s preferances, as long as the method is consistent and
hightlights the correct regions