The research developed and trained 2D and 3D deep learning models for distinguishing between atypical lipomatous tumors (ALTs) and lipomas in MRI images. Imaging was conducted preoperatively using MRI scanners with 1.5- or 3-Tesla magnets, covering various sequences, including T1-weighted, T2-weighted, and T1-weighted sequences after the application of contrast agent. The highest accuracy was achieved by the 2D model on T1w images, while 3D models performed slightly worse. Comparison with radiologists revealed that the DL model surpassed assistant doctors and yielded similar results to specialists. The study suggests promising applications of deep learning in clinical practice.
«
The research developed and trained 2D and 3D deep learning models for distinguishing between atypical lipomatous tumors (ALTs) and lipomas in MRI images. Imaging was conducted preoperatively using MRI scanners with 1.5- or 3-Tesla magnets, covering various sequences, including T1-weighted, T2-weighted, and T1-weighted sequences after the application of contrast agent. The highest accuracy was achieved by the 2D model on T1w images, while 3D models performed slightly worse. Comparison with radiol...
»