Abstract:An automated computer-aided approach might aid radiologists in diagnosing breast cancer at a primary stage. This study proposes a novel decision support system to classify breast tumors into benign and malignant based on clinically important features, using ultrasound images. Nine handcrafted features, which align with the clinical markers used by radiologists, are extracted from the region of interest (ROI) of ultrasound images. To validate that these elected clinical markers have a significant impact on predicting the benign and malignant classes, ten machine learning (ML) models are experimented with resulting in test accuracies in the range of 96 to 99%. In addition, four feature selection techniques are explored where two features are eliminated according to the feature ranking score of each feature selection method. The Random Forest classifier is trained with the resultant four feature sets. Results indicate that even when eliminating only two features, the performance of the model is reduced for each feature selection technique. These experiments validate the efficiency and effectiveness of the clinically important features. To develop the decision support system, a probability density function (PDF) graph is generated for each feature in order to find a threshold range to distinguish benign and malignant tumors. Based on the threshold range of particular features, a decision support system is developed in such a way that if at least eight out of nine features are within the threshold range, the image will be denoted as true predicted. With this algorithm, a test accuracy of 99.38% and an F1 Score of 99.05% is achieved, which means that our decision support system outperforms all the previously trained ML models. Moreover, after calculating individual class-based test accuracies, for the benign class, a test accuracy of 99.31% has been attained where only three benign instances are misclassified out of 437 instances, and for the malignant class, a test accuracy of 99.52% has been attained where only one malignant instance is misclassified out of 210 instances. This system is robust, time-effective, and reliable as the radiologists' criteria are followed and may aid specialists in making a diagnosis.

Feature analysis of ultrasound breast tumor images

Breast tumor image recognition based on artificial immune system algorithm

Multi-Features-Based Automated Breast Tumor Diagnosis Using Ultrasound Image and Support Vector Machine

Breast tumors recognition based on edge feature extraction using support vector machine

Breast Tumor Classification Based on Shape Features of Ultrasonic Images

Automatic detection and classification of breast tumors in ultrasonic images using texture and morphological features.

Breast Cancer Classification Prediction Based on Ultrasonic Image Feature Recognition

Automatic Identification of Breast Ultrasound Image Based on Supervised Block-Based Region Segmentation Algorithm and Features Combination Migration Deep Learning Model

Application of Ultrasonographic Score Evaluation System in Differential Diagnosis of Breast Mass Lesions:Comparison with Pathological Data

Breast Tumor Classification Based on a Computerized Breast Imaging Reporting and Data System Feature System

An Automated Decision Support System to Analyze Malignancy Patterns of Breast Masses Employing Medically Relevant Features of Ultrasound Images

Tomographic Ultrasound Imaging in the Diagnosis of Breast Tumors under the Guidance of Deep Learning Algorithms

Breast Cancer Classification with Ultrasound Images Based on SLIC

Multi-feature Analysis for Automated Breast Lesion Classification from Ultrasonic Data

Improving the diagnosis of breast cancer based on biophysical ultrasound features utilizing machine learning

Benign and Malignant Classification of Breast Tumor Ultrasound Images Using Conventional Radiomics and Transfer Learning Features: A Multicenter Retrospective Study

Breast Tumor Classification in Ultrasound Images by Fusion of Deep Convolutional Neural Network and Shallow LBP Feature

Automatic Detection of the Region of Interest from Breast Tumor Ultrasound Images

Artificial intelligence-based classification of breast nodules: a quantitative morphological analysis of ultrasound images

Artificial Intelligence-Based Breast Cancer Diagnosis Using Ultrasound Images and Grid-Based Deep Feature Generator