Emphysema is an incurable lung airway disease and a hallmark of Chronic Obstructive Pulmonary Disease (COPD). In recent decades, Computed Tomography (CT) has been used as a powerful tool for the detection and quantification of different diseases, including emphysema. The use of CT comes with a potential risk: ionizing radiation. It involves a trade-off between image quality and the risk of radiation exposure. However, early detection of emphysema is important as emphysema is an independent risk marker for lung cancer, and it also possesses evident qualities that make it a candidate for sub-classification of COPD. In this master's thesis, we use state-of-the-art deep learning models for pulmonary nodule detection to classify emphysema at an early stage of the disease's progression. We also demonstrate that deep learning denoising techniques can be applied to low-dose CT scans to improve the model's performance. We achieved an F-score of 0.66, an AUC score of 0.80, and an accuracy of 81.74%. The impact of denoising resulted in an increase of 1.57 percent units in accuracy and a 0.0332 increase in the F-score. In conclusion, this makes it possible to use low-dose CT scans for early detection of emphysema with State-of-The-Art deep-learning models for pulmonary nodule detection.
Identifer | oai:union.ndltd.org:UPSALLA1/oai:DiVA.org:umu-209948 |
Date | January 2023 |
Creators | Molin, Olov |
Publisher | Umeå universitet, Institutionen för datavetenskap |
Source Sets | DiVA Archive at Upsalla University |
Language | English |
Detected Language | English |
Type | Student thesis, info:eu-repo/semantics/bachelorThesis, text |
Format | application/pdf |
Rights | info:eu-repo/semantics/openAccess |
Relation | UMNAD ; 1395 |
Page generated in 0.0013 seconds