Automatic segmentation of femoral tumors by nnU-net.

BACKGROUND: Metastatic femoral tumors may lead to pathological fractures during daily activities. A CT-based finite element analysis of a patient's femurs was shown to assist orthopedic surgeons in making informed decisions about the risk of fracture and the need for a prophylactic fixation. Improving the accuracy of such analyses ruqires an automatic and accurate segmentation of the tumors and their automatic inclusion in the finite element model. We present herein a deep learning algorithm (nnU-Net) to automatically segment lytic tumors within the femur. METHOD: A dataset consisting of fifty CT scans of patients with manually annotated femoral tumors was created. Forty of them, chosen randomly, were used for training the nnU-Net, while the remaining ten CT scans were used for testing. The deep learning model's performance was compared to two experienced radiologists. FINDINGS: The proposed algorithm outperformed the current state-of-the-art solutions, achieving dice similarity scores of 0.67 and 0.68 on the test data when compared to two experienced radiologists, while the dice similarity score for inter-individual variability between the radiologists was 0.73. INTERPRETATION: The automatic algorithm may segment lytic femoral tumors in CT scans as accurately as experienced radiologists with similar dice similarity scores. The influence of the realistic tumors inclusion in an autonomous finite element algorithm is presented in (Rachmil et al., "The Influence of Femoral Lytic Tumors Segmentation on Autonomous Finite Element Analyses", Clinical Biomechanics, 112, paper 106192, (2024)).

The influence of femoral lytic tumors segmentation on autonomous finite element analysis.

BACKGROUND: The validated CT-based autonomous finite element system Simfini (Yosibash et al., 2020) is used in clinical practice to assist orthopedic oncologists in determining the risk of pathological femoral fractures due to metastatic tumors. The finite element models are created automatically from CT-scans, assigning to lytic tumors a relatively low stiffness as if these were a low-density bone tissue because the tumors could not be automatically identified. METHODS: The newly developed automatic deep learning algorithm which segments lytic tumors in femurs, presented in (Rachmil et al., 2023), was integrated into Simfini. Finite element models of twenty femurs from ten CT-scans of patients with femoral lytic tumors were analyzed three times using: the original methodology without tumor segmentation, manual segmentation of the lytic tumors, and the new automatic segmentation deep learning algorithm to identify lytic tumors. The influence of explicitly incorporating tumors in the autonomous finite element analysis on computed principal strains is quantified. These serve as an indicator of femoral fracture and are therefore of clinical significance. FINDINGS: Autonomous finite element models with segmented lytic tumors had generally larger strains in regions affected by the tumor. The deep learning and manual segmentation of tumors resulted in similar average principal strains in 19 regions out of the 23 regions within 15 femurs with lytic tumors. A high dice similarity score of the automatic deep learning tumor segmentation did not necessarily correspond to minor differences compared to manual segmentation. INTERPRETATION: Automatic tumor segmentation by deep learning allows their incorporation into an autonomous finite element system, resulting generally in elevated averaged principal strains that may better predict pathological femoral fractures.