1,721,050 research outputs found
Automatic segmentation of bone, skin and synovitis in ultrasound images of finger joints
Get PDFMaster's thesis in Cybernetics and signal processingRheumatoid arthritis (RA) is estimated to affect between 0.3 to 1.5 % of the population. It tends to strike individuals between the ages of 35-50, which is their working age, with every third individual diagnosed with RA becoming work disabled, and up to 85% of the individuals who still can work losing almost 40 days per year on average. Therefore, an accurate measurement of disease activity is crucial to provide adequate treatment and care for patients. The first stage in RA is inflammation of the synovial membrane which is called synovitis. Using ultrasonography has proven to provide useful information regarding the disease activity. The assessment of disease activity has until now been done visually by doctors by grading the synovitis from 0-3 in the ultrasound images. Making a software to automate these assessments in order to reduce the number of human-dependent discrepancies can be advantageous.
Materials given in this thesis came from the Norwegian and Polish collaborative project, MEDUSA. They included ultrasound images of finger joints and manually annotated data which was used for similarity measurement. The objective of this thesis has been to segment the synovitis in the ultrasound images automatically. Since it develops from the joint area towards the skin, it was nec-essary to segment skin and bones first. Multiple image processing techniques were tested for the proposed system for segmentation of bone skin and synovitis. Novel methods for segmentation and location of these features were also developed. All the proposed methods were implemented using MATLAB.
The similarity measurement was done by computing the modified Hausdorff distance for bone and skin, whereas the Dice coefficient was used for comparing the synovitis with the annotation data. The results show that the proposed system for segmentation of bone and skin functioned well with 80% of the segmented bone and skin features having a distance under 20px to the annotation data. However, one of the two bones had only 55% under 20px, but had a median of 11px. The proposed system for segmentation of the synovitis gave an overall low Dice coefficient, with the best result giving a median and mean Dice of 58 and 54 respectively using Region growing. However, when inspecting the images visually, most of the segmented synovitis seemed descent.
It was concluded that even though the skin and bone segmentation was good, the proposed methods for segmentation of the synovitis did not yield satisfactory results for future grading of it
Classification of histological images of bladder cancer using deep learning
Get PDFMaster's thesis in Cybernetics and signal processingIn Norway bladder cancer is the fourth most common cancer type among men, with an almost 70 % increase in incidence the past four decades. For women, the increase has been about 40 %.
The histological images of bladder cancer are investigated by a pathologist to determine the grade and stage of cancer. In addition, the risk of recurrence and progression are also diagnosed. This is done manually by studying the histological images, but reproducibility of these results are low. To aid the pathologist, a proposed automatic system have been designed in this thesis consisting of six steps. Step one to four have been studied and experimented in detail, and step five and six are considered as future work.
The histological images are divided into smaller tiles, where each tile consists of one of several different categories; cancer tissue, damaged tissue, other tissue, blood or background. The aim is to make a system which automatically separates all tiles containing cancer tissue from the rest, as these have the potential to diagnose the cancer grade, stage, recurrence and progression.
To distinguish the different categories from each other, a classification system was constructed consisting of an autoencoder and a classifier trained in a semi-supervised fashion. The autoencoder was trained on 943,127 unlabeled tiles, extracted from seven histological images. Next, the encoder part of the autoencoder was connected to the classifier which was fine-tuned on 152,312 labeled images.
For evaluating the performance of the classifier, 10-fold cross-validation was calculated. Accuracy of the best classifier on a five class dataset was 97.7 % with a standard deviation of 3.2 %
Image processing on histopathological images of urothelial carcinoma – assessment of immune cells
Get PDFMaster's thesis in Automation and Signal processingBladder cancer is the 6th most common cancer in the world, where urothelial carcinoma is the most common one. Bladder cancer is one of the most economically expensive cancers to treat, as follow up is needed over a long period of time. Through extensive research, it has been indicated that the amount of tumor infiltrating lymphocytes(TIL) can have a positive impact on the relapse rate in conjunction with treatment.
This paper concentrates on image processing to identify, and analyze the amount of TIL cells in histological images of bladder tissue. The objective of this thesis is to locate all cells in a histological image, and to train a classifier to predict if a cell is a TIL or not. The end goal is to automatically determine the amount of TILs in an image which in turn can be used to predict the effectiveness of cancer treatment. A sub set of microscopic tissue samples has been derived from digitized samples, made available by Stavanger Universitetssykehus, to be able to analyze the quantitative performance of the proposed system.
Using a distance transform, in conjunction with pre-processing methods, to 93% of the cells in the histological images were found. A side effect was that there were wrongly located multiple cell centers for some cells, in addition to other non-cell objects in the histological images. Prediction of the located cells, using histogram features, was able to achieve 92% accuracy. Using local binary pattern features, the prediction accuracy was reduced to 73%. Synthetic over-sampling was introduced as the prediction showed a higher accuracy for correctly predicted non-TILs, but this proved to decrease the quantitative performance
Bildebehandling og dype nevrale nett for deteksjon av immunceller på histologiske bilder av blærekreft
Get PDFMaster's thesis in Automation and signal processingBladder cancer is the tenth most common cancer type, where urothelial carcinoma is the most common type of bladder cancer. Bladder cancer has been classified as the most expensive type of cancer per patient, as the need for post-treatment monitoring often lasts the rest of the patient’s life. A pathologist needs to diagnose and evaluate the risk of progression and relapse from analyzing histological images.
Recent research shows a correlation between the number of regulatory T-cells and which patients that get progression to a higher cancer grade. Today a computer randomly picks out a sub-set of cells, that is to be manually counted and classified; this will serve as an estimation for regulatory T-cells compared to other cells. This paper proposes a more automated solution to aid in analyzing histological images for the number of regulatory T-cells and other cells present.
The two proposed systems are using classical image processing to find and classify the cells based on color and using a convolutional neural network to detect and classify smaller parts of the images. Both systems will attempt to estimate the number of regulatory T-cells compared to other cells.
The classical image processing had an underestimation of 4.7% for regulatory T-cells while having a 4.5% overestimation of other cells. The convolutional neural network showed a correlation between the number of classifications and the actual amount of cells but requires further work to be usable
Diagnosis, Localization, and Prognosis of Melanoma in WSIs with a Complete Pipeline by Digital Pathology
Get PDFThe most dangerous and aggressive form of skin cancer is melanoma, responsible for 90% of skin cancer mortality. Early detection of melanoma plays a crucial role in the prognostic outcome. The diagnostic has to be performed by a pathologist, which is time- consuming. The recent increase in melanoma incidents indicates the growing demand for a more efficient diagnostic process.
This thesis’s main objective is to develop a pipeline utilizing two independent pre- trained models built on the VGG16 architecture. This pipeline consists of a diagnostic and a prognostic model. The diagnosis model is responsible for localizing malignant patches in WSIs and giving a patient-level diagnosis. The prognosis model uses the output from the diagnosis model to provide a patient-level prognosis. The complete pipeline provides both a prognostic and a diagnostic tool, which can be used by a pathologist when evaluating Whole Slide Images (WSIs). A total of 243 WSIs were provided by Stavanger University Hospital for this thesis. All have been provided a patient-level label. 203 of the WSIs were used for parameter tuning and 40 were used for testing.
The diagnosis model performed with a 100% accuracy when evaluated on the original test, which was provided together with the training set. The prognosis model also per- formed well on the original dataset, with an accuracy of 0.7885. The model’s capability to predict diagnosis and prognosis decreases significantly when being introduced to the new dataset. In addition to developing the pipeline, some parameters for the diagnosis model was found using a ROC cuve. By using the new parameters for the diagnosis model on the validation set, the performance of the diagnosis model increased when using the test set. The prognosis model performed relatively equally in all experiments. A correlation between the number of patches in a WSI and the number of patches predicted malignant was discovered and counteracted by altering the patient-level threshold calculation method.The most dangerous and aggressive form of skin cancer is melanoma, responsible for 90% of skin cancer mortality. Early detection of melanoma plays a crucial role in the prognostic outcome. The diagnostic has to be performed by a pathologist, which is time- consuming. The recent increase in melanoma incidents indicates the growing demand for a more efficient diagnostic process.
This thesis’s main objective is to develop a pipeline utilizing two independent pre- trained models built on the VGG16 architecture. This pipeline consists of a diagnostic and a prognostic model. The diagnosis model is responsible for localizing malignant patches in WSIs and giving a patient-level diagnosis. The prognosis model uses the output from the diagnosis model to provide a patient-level prognosis. The complete pipeline provides both a prognostic and a diagnostic tool, which can be used by a pathologist when evaluating Whole Slide Images (WSIs). A total of 243 WSIs were provided by Stavanger University Hospital for this thesis. All have been provided a patient-level label. 203 of the WSIs were used for parameter tuning and 40 were used for testing.
The diagnosis model performed with a 100% accuracy when evaluated on the original test, which was provided together with the training set. The prognosis model also per- formed well on the original dataset, with an accuracy of 0.7885. The model’s capability to predict diagnosis and prognosis decreases significantly when being introduced to the new dataset. In addition to developing the pipeline, some parameters for the diagnosis model was found using a ROC cuve. By using the new parameters for the diagnosis model on the validation set, the performance of the diagnosis model increased when using the test set. The prognosis model performed relatively equally in all experiments. A correlation between the number of patches in a WSI and the number of patches predicted malignant was discovered and counteracted by altering the patient-level threshold calculation method
Safer Births - Using Deep Neural Networks on Fetal Heart Rate Signals
Get PDFInfant death is a big issue, especially in Africa and parts of Asia where between 24 and 30 [21] in every thousand do not survive the first month. In Europe this number is only 5.9 in every thousand. Reading fetal heart rate signals requires specialists and is time consuming and tedious work.
The objective of this thesis is to determine if deep neural networks can detect birth complications based on fetal heart rate signals collected by using a Moyo fetal heart rate monitor as a part of the Safer Births project.
The best method found in this study included augmenting the data to get similarly sized classes, creating spectrogram images, and using a convolutional neural network for classification. The final method produced an F1-score of 0.13 and detected 21.875% of the births were bag-mask ventilation was needed immediately after. The proposed methods tested in this study have not been able to detect birth complications accurately, due to insufficient amounts of data, and low-quality signals missing important features for detection
Automatic Detection of the Intersection Point Between Left and Right Ventricle in LG-CMR Images
Get PDFMaster's thesis in Cybernetics and signal processingEach year Myocardial Infarction affects millions of people worldwide. Approximately 66% of those experiencing a Myocardial Infarction survives, but are at a higher risk of heart related complications and another Infarction. Approximately 30% of post Myocardial Infarction Patients will experience two- or more Infarctions.
A Myocardial Infarction causes damage to the heart muscles, which is later healed by forming a scar at the damaged area. Scarring on the heart impacts the hearts ability to pump blood around the body, how big of an impact depends on the scars size and localization.
When cardiac magnetic resonance images are used to study the scar, general anatomical reference points can be used to define the scars size and/or localization in terms of the angular position. Two meaningful reference points, similar in all patients are the left ventricles center point and the point where the left- and right ventricle intersect.
The reference points are often manually marked by a cardiologist. Since this is a time- consuming work, it is desirable to automate this process. Detection of the left- and right ventricles center point has not yet been automated. The objective of this thesis is therefore to determine whether the intersection point can be automatically found with use of image processing, in terms of image segmentation and point features, in terms of texture and location.
The system presented in this study showed promising results, with room for improvement. 109 images, from 15 different patients, with various contrast resolutions, placement of the scar and eventual enlargement of the heart were tested.
Results showed that 70% of the images where an intersection point was detected were classified as a satisfying result, while the average Euclidean distance between the automatically- and real intersection point was 10,46 pixels. After an interpolation method was used to find the missing intersection points, the results were somewhat worsened to 67% and 11,13 pixels
Deteksjon av kompresjoner ved bruk av lydopptak fra smarttelefon under trening på hjerte- lungeredning
Get PDFMaster's thesis in Cybernetics and signal processingTrening på hjerte- lungeredning (HLR) utføres ofte i skolesammenheng med MiniAnne-dukker. Disse lager dobbel klikkelyd når det komprimeres dypt nok. Dersom elevene kan få en tilbakemelding på hvor bra HLR utføres, kan dette skape konkurranse blant elevene og følgelig bedre kompresjoner. Ideen er at hver elev bruker sin egen smarttelefon til å detektere korrekte kompresjoner for sin dukke. Dataene kan presenteres for alle elevene på en felles skjerm. Det blir også lettere for veileder å vite hvilke elever som trenger mest hjelp.
I rapporten er det utforsket mulighetene til å detektere klikkelyder fra MiniAnne med en smarttelefon. Datamateriellet som brukes i eksperimentene er samlet inn underveis ved hjelp av ansatte på Laerdal, familie og kamerater. Totalt 7 datasett ble samlet inn underveis.
Totalt tre forskjellige metoder ble testet for å detektere kompresjoner. Template matching, blind source separation og adaptiv amplitude terskling. I template matching konvolveres lydsignalet med en mal før kompresjonene detekteres med adaptiv terskling. I adaptiv amplitude terskling detekteres toppunkt over en terskelverdi. Dette brukes til å finne doble klikkelyder som er tilfelle for en kompresjon som er dyp nok. I blind source separation skilles lyden fra hver dukke fra hverandre før det gjøres en adaptiv amplitude terskling.
Av de tre forskjellige metodene som ble testet gav blind source separation dårligst resultat. Det var vanskelig å skille lyden fra hver dukke fra hverandre. Bedre gikk det med Template matching hvor resultatene ble relativt gode. Her ble rundt 80% av kompresjonene ble detektert. Den enkleste metoden, adaptiv amplitude terskling gav best resultater hvor opp mot 100% av kompresjonene ble detektert.
Konklusjonen fra oppgaven er at det fungerer å detektere kompresjoner fra MiniAnne som er dype nok med en smarttelefon. En avstand mellom personer på over 1.2m i hver retning vil bli anbefalt for at kompresjonsdeteksjonen skal være tilstrekkelig robust
Cerebral vessel segmentation in contrast CT images
No full textMaster's thesis in Automation and signal processingPatients with cerebral stroke symptoms are typically examined using CT scanning due to its availability and efficiency. If the stroke is suspected to be ischemic, a contrast agent is injected to increase the contrast in the brain. Several images are acquired as the contrast enters the brain. This accentuates the vessels and shows how the contrast distributes in the brain. This helps to locate the stroke and the tissue that is affected by it, i.e. if is in risk of getting or already is irreversibly damaged.
This thesis uses contrast CT images of 11 stroke patients. The images are registered, and methods to remove skull and find vessel segments are developed and evaluated. The skull is removed by using wavelet coefficient decomposition to detect edges, and then perform thresholding and watershed segmentation to isolate the brain. To detect the vessel segments, two methods are implemented. One uses adaptive thresholding and the other uses unsupervised clustering based on the features that are typical for vessel segments. Both the skull removal and the vessel segmentation have visually been shown to be promising
Detection of heart rate and respiratory rate from video
Get PDFMaster's thesis in Cybernetics and signal processingI rapporten er mulighetene en har ved å benytte video til deteksjon av hjerte- og respirasjonsrate på nyfødte babyer utforsket. Datamaterialet som inngår i forsøkene og utviklingen av algoritmene er hentet inn ved Universitet i Stavanger (UIS). Testobjektene har vært voksne personer med ulik alder, kjønn, og hudfarge. Det er benyttet to kameratyper til opptakene, et speilreflekskamera i høy prisklasse og et webcamera i lav prisklasse. Ulike lysforhold i rommet hvor opptakene utføres er også testet ut. Løsningene som er utviklet er også testet ut i praksis, og her er datamateriale hentet inn fra nyfødt intensivavdeling ved Stavanger Universitetsykehus (SUS).
Hjerteratedeteksjon utføres ved å studere små endringer i fargekomponenter i bilderammene over tid. Ansiktet til testobjektene er utsnittsområdet deteksjonen foretas i. Det er sammenlignet 11 ulike metoder som tar utgangspunkt i fargekomponentene i bilderammene for å generere hjerteratesignalet.
I deteksjon av respirasjonsrate er det tatt utgangspunkt i metoden utviklet av Tveit [3], men hvor det er utført noen mindre modifikasjoner. Metoden studerer lokal fase og amplitude i bilderammer ved hjelp av Riesz transform, og genererer den globale bevegelsen i bilderammene ved å finne differansen i disse over tid.
Deteksjonene er følsomme for bevegelser, og en er avhengige av at personen ligger i ro for å kunne foreta en god deteksjon. Det er derfor implementert en forstyrrelsesdeteksjon som kutter ut deteksjon av hjerte- og respirasjonsrate dersom bevegelsene i bilderammene er for store. Denne metoden blir også benyttet for å detektere en mulig respirasjonsstopp. Dersom det i den siste perioden har vært for lite bevegelse mellom bilderammene gir detektoren alarm.
Resultatene er gode for deteksjon av både hjerte- og respirasjonsrate i testopptakene utført ved UIS. Kameratype spilte ingen vesentlig rolle og deteksjonene kan utføres med et webcamera i lav prisklasse. Lysskilder som forårsaker varierende lysforhold for bilderammene i opptaket gjør en hjerteratedeteksjon mer utfordrende, men fortsatt mulig. Da metodene ble testet ut i praksis på SUS fikk vi varierende resultat for hjerteratedeteksjon, og videre testopptak må her utføres for å kunne lokalisere støykilden som befant seg i to av opptakene. For respirasjonsdeteksjon var resultatene gode også ved SUS.
Kjøretiden på deteksjonene tilsier at en total sanntidsdeteksjon av både hjerte- og respirasjonsrate, forstyrrelse og respirasjonsstopp er mulig. Resultatene samlet sett er lovende for at detektoren skal kunne implementeres i en overvåkingsløsning av nyfødte
- …
