Showing 1–2 of 2 results for author: Chiti, A
-
Multi-stage intermediate fusion for multimodal learning to classify non-small cell lung cancer subtypes from CT and PET
Authors:
Fatih Aksu,
Fabrizia Gelardi,
Arturo Chiti,
Paolo Soda
Abstract:
Accurate classification of histological subtypes of non-small cell lung cancer (NSCLC) is essential in the era of precision medicine, yet current invasive techniques are not always feasible and may lead to clinical complications. This study presents a multi-stage intermediate fusion approach to classify NSCLC subtypes from CT and PET images. Our method integrates the two modalities at different st…
▽ More
Accurate classification of histological subtypes of non-small cell lung cancer (NSCLC) is essential in the era of precision medicine, yet current invasive techniques are not always feasible and may lead to clinical complications. This study presents a multi-stage intermediate fusion approach to classify NSCLC subtypes from CT and PET images. Our method integrates the two modalities at different stages of feature extraction, using voxel-wise fusion to exploit complementary information across varying abstraction levels while preserving spatial correlations. We compare our method against unimodal approaches using only CT or PET images to demonstrate the benefits of modality fusion, and further benchmark it against early and late fusion techniques to highlight the advantages of intermediate fusion during feature extraction. Additionally, we compare our model with the only existing intermediate fusion method for histological subtype classification using PET/CT images. Our results demonstrate that the proposed method outperforms all alternatives across key metrics, with an accuracy and AUC equal to 0.724 and 0.681, respectively. This non-invasive approach has the potential to significantly improve diagnostic accuracy, facilitate more informed treatment decisions, and advance personalized care in lung cancer management.
△ Less
Submitted 21 January, 2025;
originally announced January 2025.
-
Chest X-Rays Image Classification from beta-Variational Autoencoders Latent Features
Authors:
Leonardo Crespi,
Daniele Loiacono,
Arturo Chiti
Abstract:
Chest X-Ray (CXR) is one of the most common diagnostic techniques used in everyday clinical practice all around the world. We hereby present a work which intends to investigate and analyse the use of Deep Learning (DL) techniques to extract information from such images and allow to classify them, trying to keep our methodology as general as possible and possibly also usable in a real world scenari…
▽ More
Chest X-Ray (CXR) is one of the most common diagnostic techniques used in everyday clinical practice all around the world. We hereby present a work which intends to investigate and analyse the use of Deep Learning (DL) techniques to extract information from such images and allow to classify them, trying to keep our methodology as general as possible and possibly also usable in a real world scenario without much effort, in the future. To move in this direction, we trained several beta-Variational Autoencoder (beta-VAE) models on the CheXpert dataset, one of the largest publicly available collection of labeled CXR images; from these models, latent features have been extracted and used to train other Machine Learning models, able to classify the original images from the features extracted by the beta-VAE. Lastly, tree-based models have been combined together in ensemblings to improve the results without the necessity of further training or models engineering. Expecting some drop in pure performance with the respect to state of the art classification specific models, we obtained encouraging results, which show the viability of our approach and the usability of the high level features extracted by the autoencoders for classification tasks.
△ Less
Submitted 29 September, 2021;
originally announced September 2021.
