Preoperative Classification of Nerve Sheath Tumors Using Radiomics and Semi-Quantitative Parameters
Christianne Y.M.N. Jansma, BSc1,2, Ibtissam Acem, BSc2, Douwe J. Spaanderman, Msc3, Jan Jaap Visser, MD, PhD3, David Hanff, MD3, Walter Taal, MD, PhD4, J. Henk Coert, MD PhD1, Cornelis Verhoef, MD, PhD4, Enrico Martin, MD, PhD5 and Martijn P.A. Starmans, MD, PhD3, 1University Medical Center Utrecht, Utrecht, Utrecht, Netherlands, 2Erasmus MC Cancer Institute, Rotterdam, Zuid-Holland, Netherlands, 3Erasmus MC, Rotterdam, Zuid-Holland, Netherlands, 4Erasmus Medical Center, Rotterdam, Zuid-Holland, Netherlands, 5UMC Utrecht, Utrecht, Utrecht, Netherlands
Introduction: Malignant peripheral nerve sheath tumors (MPNSTs) are aggressive soft tissue sarcomas occurring in neurofibromatosis type 1 (NF1) patients. High-grade MPNSTs carry a high risk of metastasis and recurrence, making early recognition and surgical resection crucial for improved survival. The resection of high-grade MPNSTs often leads to postoperative complications, while neurofibromas can be adequately resected with minimal nerve damage. Therefore, preoperative differentiation between benign peripheral nerve sheath tumors (BPNSTs) and MPNSTs is important. However, current imaging tools may not always provide sufficient diagnostic accuracy, leading to the need for biopsies and associated burdens. Distinguishing features in magnetic resonance imaging (MRI) are not yet known. Radiomics provides a potential new tool in the diagnostic armamentarium, which may reduce the need for biopsies. To date, there are no studies incorporating T2-weighted MRI sequences in addition to T1-weighted MRI sequences. This study aims to develop a radiomics model that utilizes quantitative imaging features and machine learning to differentiate BPNSTs from MPNSTs based on T1- and T2-weighted MRI sequences.
Materials and Methods: We collected T1- and T2-weighted MRI sequences from BPNST and MPNST patients at our tertiary referral center for sarcoma. Clinical data included age, sex, neurogenic diagnosis, presence of spontaneous pain or preoperative motor deficits. Lesions were manually segmented on the MRI sequence where the tumor was best visible. Segmentations were warped to the other sequences using image registration. For each lesion, on each sequence, 564 radiomics features were extracted. For classification, the WORC algorithm was used, which includes a large set of commonly used radiomics methods and uses automated machine learning to determine their optimal combination based on the training set. Evaluation was performed using a 100x random-split cross-validation with 20% of the data for testing. Performance was compared to manual scoring by two radiologist who had access to the scans of the complete MRI sessions.
Results: A total of 35 MPNSTs and 73 BPNSTs were included. The radiomics models had a mean test area under the curve (AUC) of 0.71 on T1-weighted MRI, 0.68 on T2-weighted MRI, and 0.69 on the combination T1- and T2-weighted MRI. The two radiologist had AUCs of 0.75 and 0.60.
Conclusions: Radiomics based machine learning using T1- and T2 weighted MRI sequences can provide a valid tool for improved clinical decision-making in the management of these tumors. Further validation and refinement of the radiomics model are warranted to enhance its diagnostic accuracy and clinical utility.
Back to 2024 Abstracts