Faccioli, Caterina
(2022)
Spatial analysis in pathomics: a network based method applied on fluorescence microscopy.
[Laurea magistrale], Università di Bologna, Corso di Studio in
Physics [LM-DM270], Documento ad accesso riservato.
Documenti full-text disponibili:
![[img]](http://amslaurea.unibo.it/style/images/fileicons/application_pdf.png) |
Documento PDF (Thesis)
Full-text accessibile solo agli utenti istituzionali dell'Ateneo
Disponibile con Licenza: Salvo eventuali più ampie autorizzazioni dell'autore, la tesi può essere liberamente consultata e può essere effettuato il salvataggio e la stampa di una copia per fini strettamente personali di studio, di ricerca e di insegnamento, con espresso divieto di qualunque utilizzo direttamente o indirettamente commerciale. Ogni altro diritto sul materiale è riservato
Download (29MB)
| Contatta l'autore
|
Abstract
Recently, some applications of spatial statistics in histopathology have been explored, also thanks to the development of innovative digital imaging techniques and machine learning algorithms.
However, it seems that in all these studies only global spatial measures are considered, usually analysed in combination with other techniques that depart from spatial statistics.
In this thesis we developed a new spatial statistic based method for histopathological image analysis, which exploits local spatial features derived from coordinates in space and area of the cells. This features are mostly based on reciprocal distance between cells and also includes network-related measures. The dataset we analysed consisted of many sections of lymphoid tissue, for which also fluorescence measures obtained with a particular multiplexing technique were available. We performed clustering on these fluorescence features in order to obtain some reference labels for our points. Then we applied a supervised learning algorithm in order to predict fluorescence labels from the spatial features. We measured the performance of our predictions by computing the difference between the accuracy of the classifier we applied and of a random classifier.
What we obtained is that the accuracy score of our classifier was greater than the one of the dummy classifier in every image.
From a qualitative point of view, by comparing the achieved predictions and the clustering of fluorescence features of our images we obtained good results (verified by a senior histopathologist), often managing to identify the zone around the germinal centres of the lymph nodes and other structures.
We consider these results encouraging, since they prove the predictive capability of our spatial features towards biological structures. The potential of this work is big: these features could strongly enhance the results obtainable from fluorescence imaging, allowing to resolve previously undistinguishable structures.
Abstract
Recently, some applications of spatial statistics in histopathology have been explored, also thanks to the development of innovative digital imaging techniques and machine learning algorithms.
However, it seems that in all these studies only global spatial measures are considered, usually analysed in combination with other techniques that depart from spatial statistics.
In this thesis we developed a new spatial statistic based method for histopathological image analysis, which exploits local spatial features derived from coordinates in space and area of the cells. This features are mostly based on reciprocal distance between cells and also includes network-related measures. The dataset we analysed consisted of many sections of lymphoid tissue, for which also fluorescence measures obtained with a particular multiplexing technique were available. We performed clustering on these fluorescence features in order to obtain some reference labels for our points. Then we applied a supervised learning algorithm in order to predict fluorescence labels from the spatial features. We measured the performance of our predictions by computing the difference between the accuracy of the classifier we applied and of a random classifier.
What we obtained is that the accuracy score of our classifier was greater than the one of the dummy classifier in every image.
From a qualitative point of view, by comparing the achieved predictions and the clustering of fluorescence features of our images we obtained good results (verified by a senior histopathologist), often managing to identify the zone around the germinal centres of the lymph nodes and other structures.
We consider these results encouraging, since they prove the predictive capability of our spatial features towards biological structures. The potential of this work is big: these features could strongly enhance the results obtainable from fluorescence imaging, allowing to resolve previously undistinguishable structures.
Tipologia del documento
Tesi di laurea
(Laurea magistrale)
Autore della tesi
Faccioli, Caterina
Relatore della tesi
Correlatore della tesi
Scuola
Corso di studio
Indirizzo
Applied Physics
Ordinamento Cds
DM270
Parole chiave
spatial statistics,network,spatial,umap,DBSCAN,clustering,fluorescence,immunofluorescence,histopathology,pathomics,lymph nodes
Data di discussione della Tesi
18 Febbraio 2022
URI
Altri metadati
Tipologia del documento
Tesi di laurea
(NON SPECIFICATO)
Autore della tesi
Faccioli, Caterina
Relatore della tesi
Correlatore della tesi
Scuola
Corso di studio
Indirizzo
Applied Physics
Ordinamento Cds
DM270
Parole chiave
spatial statistics,network,spatial,umap,DBSCAN,clustering,fluorescence,immunofluorescence,histopathology,pathomics,lymph nodes
Data di discussione della Tesi
18 Febbraio 2022
URI
Statistica sui download
Gestione del documento: