Primal-dual for classification with rejection (PD-CR): a novel method for classification and feature selection—an application in metabolomics studies

Chardin, David; Humbert, Olivier; Bailleux, Caroline; Burel‐Vandenbos, Fanny; Rigau, Valérie; Pourcher, Thierry; Barlaud, Michel

doi:10.1186/s12859-021-04478-w

“…Thus, two effects are at play during training: the classification loss trains the encoder to be a good classifier between malignant and benign nodules, and the reconstruction loss trains the encoder to learn a good representation of the data to facilitate the task of the decoder (which is also trained during this process). See [44], [45], [46], [47] and Fig. 2 for more details.…”

Section: A New Supervised Autoencoder Architecturementioning

confidence: 99%

Low-cost Thyroid Nodule Diagnosis Using a Supervised Autoencoder Neural network

Barlaud¹

2023

Preprint

0

View full text Add to dashboard Cite

Background: Thyroid Nodules are more and more widespread, with the frequency of thyroid cancers increasing accordingly. In routine practice, the practician's goal is to determine if a nodule is malignant or suspicious, and thus requires surgery. However, nearly one-third of thyroid nodules fall into the 'indeterminate' classification segment, often leading to the use of surgery as a diagnostic tool, when it could have been avoided with a better preoperative diagnosis. Molecular tests exist to address this issue, but they are markedly onerous. Results: In this paper, we present an accurate, cost-less method to predict the malignancy of thyroid nodules using a categorical thyroid nodules dataset. The prediction is achieved through the use of a new supervised autoencoder neural network. Results show that the Supervised Autoencoder Neural Network outperforms classical classification methods on all evaluation metrics. A proof-of-concept Graphical User Interface was developed to demonstrate the medical usefulness of the latent space of the trained network. Conclusions: Aside from the prediction itself, the Supervised Autoencoder Neural Network provides three medically interesting characteristics: a prediction confidence score, consistent feature ranking, and an interpretable latent space. All of those characteristics are valuable from the practician's point of view, as they allow for an efficient, cost-less and well-informed prognosis.

show abstract

“…Thus, two effects are at play during training: the classification loss trains the encoder to be a good classifier between malignant and benign nodules, and the reconstruction loss trains the encoder to learn a good representation of the data to facilitate the task of the decoder (which is also trained during this process). See [44], [45], [46], [47] and Fig. 2 for more details.…”

Section: A New Supervised Autoencoder Architecturementioning

confidence: 99%

Low-cost Thyroid Nodule Diagnosis Using a Supervised Autoencoder Neural network

Barlaud¹

2023

Preprint

0

View full text Add to dashboard Cite

Background: Thyroid Nodules are more and more widespread, with the frequency of thyroid cancers increasing accordingly. In routine practice, the practician's goal is to determine if a nodule is malignant or suspicious, and thus requires surgery. However, nearly one-third of thyroid nodules fall into the 'indeterminate' classification segment, often leading to the use of surgery as a diagnostic tool, when it could have been avoided with a better preoperative diagnosis. Molecular tests exist to address this issue, but they are markedly onerous. Results: In this paper, we present an accurate, cost-less method to predict the malignancy of thyroid nodules using a categorical thyroid nodules dataset. The prediction is achieved through the use of a new supervised autoencoder neural network. Results show that the Supervised Autoencoder Neural Network outperforms classical classification methods on all evaluation metrics. A proof-of-concept Graphical User Interface was developed to demonstrate the medical usefulness of the latent space of the trained network. Conclusions: Aside from the prediction itself, the Supervised Autoencoder Neural Network provides three medically interesting characteristics: a prediction confidence score, consistent feature ranking, and an interpretable latent space. All of those characteristics are valuable from the practician's point of view, as they allow for an efficient, cost-less and well-informed prognosis.

show abstract

A Novel Filter and Embedded Feature Selection Methods applied to High Dimensional Metabolomics Data in Enhancing Colorectal Cancer Classification

Ibrahim,

Ul-Saufie,

Tharmaratnam

et al. 2023

Preprint

0

View full text Add to dashboard Cite

Background Metabolomics is an emerging field, which focuses on the study of small molecules (metabolites) and their chemical processes. Metabolomics data are highly dimensional, with p>>n where p is the number of variables and n is the sample size of the cohort. Hence, feature selection is a key step in metabolomics studies to reduce the dimensionality in the dataset, removing redundant and unwanted features and mitigating overfitting. The t-test (T) and correlation sharing t-test method (corT) can be used as filter methods. Penalized regression, and in particular the embedded method least absolute shrinkage and selection operator (Lasso), have also been applied for feature selection with the aim of minimising the problem of overfitting that often affects prediction models in this field. These methods are here applied to datasets of volatile organic compounds (VOCs) from patients with colorectal cancer or non-cancer (aimed at discriminating between non-cancer vs colorectal cancer groups, and healthy control vs adenoma groups). Cross validation, with data split into two sets (80% for training and 20% for validation) was used to compare the performance of the feature selection methods in terms of classification accuracy, sensitivity, specificity, and area under ROC. Results As a result, for non-cancer and colorectal cancer discrimination, the T method showed the worst classification accuracy followed by Lasso. CorT achieved the best level of discrimination although this was still low (AUC of 0.60). For healthy control and adenoma discrimination however, methods corT showed the lowest AUC, followed by the T method. Lasso achieved the best level of discrimination, although this remained low (AUC of 0.65). However, there is limitation of these methods where the feature selection methods considered were not able to identify a set of VOCs with good levels of discrimination between colon cancer, adenoma and control cases. Conclusion This paper is being limited use to assist medical practitioners in earlier detection of colorectal cancer.

show abstract

Primal-dual for classification with rejection (PD-CR): a novel method for classification and feature selection—an application in metabolomics studies

Cited by 4 publications

References 41 publications

Low-cost Thyroid Nodule Diagnosis Using a Supervised Autoencoder Neural network

Low-cost Thyroid Nodule Diagnosis Using a Supervised Autoencoder Neural network

Low-cost Thyroid Nodule Diagnosis Using a Supervised Autoencoder Neural network

A Novel Filter and Embedded Feature Selection Methods applied to High Dimensional Metabolomics Data in Enhancing Colorectal Cancer Classification

Contact Info

Product

Resources

About