Study Proposes Acne Grading System, Out-Performs Previous Methods

June 2, 2022
Armand Butera

Armand Butera is the assistant editor for HCPLive. He attended Fairleigh Dickinson University and graduated with a degree in communications with a concentration in journalism. Prior to graduating, Armand worked as the editor-in-chief of his college newspaper and a radio host for WFDU. He went on to work as a copywriter, freelancer, and human resources assistant before joining HCPLive. In his spare time, he enjoys reading, writing, traveling with his companion and spinning vinyl records. Email him at abutera@mjhlifesciences.com.

Severe or inflammatory acne can result in prominent scarring in affected patients, which highlights the importance of grading acne in diagnosis of the skin condition.

A new proposal for grading acne via ensemble pruning had been shown to outperform previous state-of-the-art methods according to 2 new datasets from Jilin University.

Severe or inflammatory acne can result in prominent scarring in affected patients, which highlights the importance of grading acne in diagnosis of the skin condition. However, grading acne is often labor-intensive and can result in mistakes made by dermatologists.

As such, investigators led by Shuai Liu, MD, College of Computer Science and Technology at Jilin University, believed that it was imperative to develop considered and automatic diagnostic methods for grading acne.

With their study, Liu and colleagues proposed a novel ensemble classification framework to classify acne severity, which pertains to an ensemble pruning strategy designed to reduce the computational complexity of the trained ensemble classification model.

Additionally, the team used the prediction results of the base models as a new set of features and used the classifier to ensemble the results of the base models.

Liu and colleagues utilized the ACNE04 dataset, which contains 1457 images, in the evaluation of the proposed algorithm, AcneGrader, for the detection and grading of acne. This dataset annotated local lesions and global acne severity using the Hayashi criterion determined by professional dermatologists.

The investigators used 80% of randomly retrieved samples to train the model they developed, while the remaining 20% were used to test the model. Specifically, the dataset contained 3297 images, 2637 of which were used to for training and 660 for testing. A fivefold stratified cross-validation strategy was used to evaluate the prediction algorithms, and each of the fivefold were used at the test dataset iteratively.

Notably, a skin cancer dermoscopic image dataset was used to further verify the effectiveness of the model.

The study formulated the acne grading problem as a 4-class classification problem, and acne was grades as either mild, moderate, severe, and very severe.

Redundancy models were later removed by a feature selection algorithm, followed by the team integrating all base models by classifiers. Finally, the ensemble pruning algorithm was proposed to prune the deep learning models.

Additional training samples did not appear to lead to improved prediction performance, and 80% of the training dataset generated the best models.

However, investigators noted that pruning algorithms improve the ensemble learning models, overall, and the Kappa statistics achieved the best model performance using 22 base models, which was chosen as the default model.

The experimental data indicated that the ensemble pruned framework resulted in a prediction accuracy of 85.82% on the acne dataset, which investigators noted was better than existing studies.

"If our framework is used in an environment with limited computing resources, such as mobile devices, the ensemble model pruned by the error pruning strategy maybe considered," the team wrote.

The study, "AcneGrader: An ensemble pruning of the deep learning base models to grade acne," was published online in Skin Research & Technology.


x