Machine learning using ultrasound improves cardiac tumor diagnosis

Machine Learning


Machine learning could help improve the interpretation of echocardiograms for heart tumors, according to a study published July 1. Opening the door to medical informatics.

A research team led by Dr. Seyed Ali Sadegh Zadeh of Staffordshire University in the UK found that their machine learning model achieved high performance in diagnosing cardiac tumors, including a near-perfect area under the curve (AUC) score.

“These findings suggest that machine learning has the potential to revolutionize cardiac tumor diagnosis, providing a path toward a more accurate, non-invasive, and patient-centric diagnostic process,” Sadegh-Zadeh and his team wrote.

Although rare, cardiac tumors pose unique challenges to clinicians because their symptoms mimic other diseases. Advanced imaging is required to localize and characterize these tumors.

Although echocardiography is the primary imaging modality in this field, it has limited ability to distinguish between tumour types and determine malignancy, and the researchers highlighted that machine learning techniques may lead to improved diagnostic performance.

To improve diagnostic accuracy for cardiac tumors, Sadegh-Zadeh and his colleagues integrated echocardiographic imaging and pathology data with advanced machine learning techniques. They used support vector machines, random forests, and gradient boosting machines optimized for limited datasets from specialized medical domains.

The study included clinical data from 399 patients and evaluated the model's performance against traditional diagnostic criteria. The researchers reported that the random forest model outperformed other models in accurate diagnosis.

Performance of machine learning models in diagnosing cardiac tumors
measurement Support Vector Machine Gradient Boosting Machine Random Forest
Accuracy 71.25% 96.25% 96.25%
Accuracy (benign tumors) 78% 99% 99%
Accuracy (malignant tumors) 50% 88% 88%
Recall (benign) 43% 95% 95%
Recall (Malignant) 43% 99% 99%
F1 score (benign) 80.34 97.3% 97.3%
F1 score (malignant) 46.51 93.88% 93.88%
Commonwealth of Australia 0.72 0.98 0.99

The research team also identified important clinical predictors, including age, echo grade, and echo location, which they noted highlights the value of integrating multiple data types.

The random forest model was incorporated into clinical validation and achieved a diagnostic accuracy of 94% in a real-world setting.

The study authors emphasized that their findings demonstrate the ability of machine learning to improve the accuracy of cardiac tumor diagnosis. They added that the study “also lays the foundation for future exploration” of broader applications of the technology across various areas of medical diagnostics. The study highlights the need for larger datasets and external validation, the authors noted.

“Furthermore, examining implementation studies to understand the practical aspects of integrating these models into clinical practice, such as workflow integration, clinician training, and patient outcomes, will be essential for successful adoption,” the researchers wrote.

The full survey can be found here.



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *