| Title | Survival analysis incorporating medical imaging data / |
|---|---|
| Translation of Title | Medicininių vaizdų panaudojimas išgyvenamumo analizėje. |
| Authors | Jonaitytė, Ieva |
| Full Text |
|
| Pages | 40 |
| Keywords [eng] | Medicininių vaizdų analizė, histopatologinės nuotraukos, konvoliuciniai naurnų tinklai, išgyvenamumo analizė, Kokso proporcingų rizikų regresija, cenzūravimas Medical image analysis, Histopathology images, Convolutional Neural Networks, Survival analysis, Cox proportional hazard regression, censoring |
| Abstract [eng] | In this thesis the possibility of using medical images for survival analysis is investigated. In such analysis it is still common to use handcrafted features which needs lots of prior knowledge. Also such processing is quite different and subjective for each specific task. In this research we analyse how deep features affect survival model. Firstly, a few methods using different convolutional models are applied to obtain the representative covariates. Second, these image-related covariates are analysed by Cox proportional hazard regression. The aim is to find out whether proposed models are capable of extracting any meaningful covariates that are significant in survival analysis, without using explicit feature engineering or image labelling. During the work convolutional neural network models generating deep features from images are created. The experiments are run on breast cancer data set. |
| Dissertation Institution | Vilniaus universitetas. |
| Type | Master thesis |
| Language | English |
| Publication date | 2021 |