Automatic detection, segmentation, and classification of primary bone tumors and bone infections using an ensemble multi-task deep learning framework on multi-parametric MRIs: a multi-center study.

OBJECTIVES: To develop an ensemble multi-task deep learning (DL) framework for automatic and simultaneous detection, segmentation, and classification of primary bone tumors (PBTs) and bone infections based on multi-parametric MRI from multi-center. METHODS: This retrospective study divided 749 patients with PBTs or bone infections from two hospitals into a training set (N = 557), an internal validation set (N = 139), and an external validation set (N = 53). The ensemble framework was constructed using T1-weighted image (T1WI), T2-weighted image (T2WI), and clinical characteristics for binary (PBTs/bone infections) and three-category (benign/intermediate/malignant PBTs) classification. The detection and segmentation performances were evaluated using Intersection over Union (IoU) and Dice score. The classification performance was evaluated using the receiver operating characteristic (ROC) curve and compared with radiologist interpretations. RESULT: On the external validation set, the single T1WI-based and T2WI-based multi-task models obtained IoUs of 0.71 ± 0.25/0.65 ± 0.30 for detection and Dice scores of 0.75 ± 0.26/0.70 ± 0.33 for segmentation. The framework achieved AUCs of 0.959 (95%CI, 0.955-1.000)/0.900 (95%CI, 0.773-0.100) and accuracies of 90.6% (95%CI, 79.7-95.9%)/78.3% (95%CI, 58.1-90.3%) for the binary/three-category classification. Meanwhile, for the three-category classification, the performance of the framework was superior to that of three junior radiologists (accuracy: 65.2%, 69.6%, and 69.6%, respectively) and comparable to that of two senior radiologists (accuracy: 78.3% and 78.3%). CONCLUSION: The MRI-based ensemble multi-task framework shows promising performance in automatically and simultaneously detecting, segmenting, and classifying PBTs and bone infections, which was preferable to junior radiologists. CLINICAL RELEVANCE STATEMENT: Compared with junior radiologists, the ensemble multi-task deep learning framework effectively improves differential diagnosis for patients with primary bone tumors or bone infections. This finding may help physicians make treatment decisions and enable timely treatment of patients. KEY POINTS: • The ensemble framework fusing multi-parametric MRI and clinical characteristics effectively improves the classification ability of single-modality models. • The ensemble multi-task deep learning framework performed well in detecting, segmenting, and classifying primary bone tumors and bone infections. • The ensemble framework achieves an optimal classification performance superior to junior radiologists' interpretations, assisting the clinical differential diagnosis of primary bone tumors and bone infections.

Anatomic characteristics of the knee influence the risk of suffering an isolated meniscal injury and the risk factors differ between women and men.

PURPOSE: To analyse the relationship between multiple anatomic characteristics of the knee (tibia and femur) and isolated meniscal injury in women and men. METHODS: Forty-seven patients with isolated medial meniscal injuries, 62 patients with isolated lateral meniscal injuries, and 70 control subjects were included. Medial posterior tibial slope (MTS), lateral posterior tibial slope (LTS), medial tibial plateau depth (MTD), coronal tibial slope (CTS), femoral notch width (NW), femoral condylar width (FCW), intercondylar notch depth (ND), femoral notch width index (NWI), intercondylar notch shape index (NSI), and cruciate ligaments tensity (CLT) were measured from magnetic resonance images. Anatomic characteristics differing between groups were compared, and risk factors for isolated meniscal injury were identified by multivariate forward stepwise logistic regression for men and women separately. RESULTS: Risk factors for an isolated medial meniscal injury were a steeper MTS and a lowered MTD in men, and a steeper MTS and an increased NWI in women. Risk factors for isolated lateral meniscal injury were a steeper LTS and an increased NW in men, and a steeper LTS and a lowered ND in women. Risk factors for both medial and lateral meniscal injuries were a higher CTS, an increased NWI, and a looser CLT in men, and a higher CTS, an increased NSI, and a looser CLT in women. CONCLUSION: The anatomic characteristics of the tibial plateau, femur, and cruciate ligaments influence the risk of suffering isolated meniscal injury, and the risk factors differ between men and women. This study provides a reference for developing identification criteria for those at risk of isolated meniscal injury. LEVEL OF EVIDENCE: III.

Evaluation of Mesorectal Microcirculation With Quantitative Dynamic Contrast-Enhanced MRI.

OBJECTIVE. The purpose of this study was to use quantitative dynamic contrast-enhanced MRI (DCE-MRI) to evaluate mesorectal microcirculation in patients with rectal cancer. MATERIALS AND METHODS. A total of 53 patients with semicircular rectal tumors underwent DCE-MRI with a 3-T MRI system before surgery. ROIs were manually delineated in the mesorectum that surrounded the tumor and the mesorectum that surrounded the normal rectal wall. DCE-MRI parameters including forward volume transfer constant (Ktrans), reverse volume transfer constant (kep), and fractional extravascular extracellular space volume (Ve) were estimated using computer software. Histopathologic analysis served as the standard reference. RESULTS. Mesorectum that surrounded the tumor showed significantly higher Ktrans val ues than mesorectum that surrounded normal rectal wall (mean, 0.069 ± 0.035 [SD] vs 0.039 ± 0.020 min-1; p < 0.001). The tumor-surrounding mesorectum also showed higher Ve values than normal mesorectum (p < 0.001). An opposite trend was observed for kep, but this was not significant (p = 0.077). A lower Ktrans of the tumor-surrounding mesorectum was observed in patients with malignant lymph nodes compared with those with benign lymph nodes (mean, 0.054 ± 0.027 vs 0.076 ± 0.036 min-1; p = 0.034). Although kep values for the tumor-surrounding mesorectum were higher in patients with tumors categorized as pathologic Tis (pTis) to pT2 than in those with pT3 tumors, the p value was close to 0.05 (p = 0.047). The tumor-surrounding mesorectum showed no significant differences in the aforementioned parameters between patients with positive MRI-detected extramural vascular invasion (mrEMVI) and those with negative mrEMVI. CONCLUSION. Mesorectum that surrounded rectal tumor had a higher blood flow than that close to the normal rectal wall. The blood flow decreased in the tumor-surrounding mesorectum when there was nodal involvement.

Using Three-Dimensional Printing to Create Individualized Cranial Nerve Models for Skull Base Tumor Surgery.

OBJECTIVE: Using three-dimensional (3D) printing to create individualized patient models of the skull base, the optic chiasm and facial nerve can be previsualized to help identify and protect these structures during tumor removal surgery. METHODS: Preoperative imaging data for 2 cases of sellar tumor and 1 case of acoustic neuroma were obtained. Based on these data, the cranial nerves were visualized using 3D T1-weighted turbo field echo sequence and diffusion tensor imaging-based fiber tracking. Mimics software was used to create 3D reconstructions of the skull base regions surrounding the tumors, and 3D solid models were printed for use in simulation of the basic surgical steps. RESULTS: The 3D printed personalized skull base tumor solid models contained information regarding the skull, brain tissue, blood vessels, cranial nerves, tumors, and other associated structures. The sphenoid sinus anatomy, saddle area, and cerebellopontine angle region could be visually displayed, and the spatial relationship between the tumor and the cranial nerves and important blood vessels was clearly defined. The models allowed for simulation of the operation, prediction of operative details, and verification of accuracy of cranial nerve reconstruction during the operation. Questionnaire assessment showed that neurosurgeons highly valued the accuracy and usefulness of these skull base tumor models. CONCLUSIONS: 3D printed models of skull base tumors and nearby cranial nerves, by allowing for the surgical procedure to be simulated beforehand, facilitate preoperative planning and help prevent cranial nerve injury.

Improving bone repair of femoral and radial defects in rabbit by incorporating PRP into PLGA/CPC composite scaffold with unidirectional pore structure.

In this study, a platelet-rich plasma poly(lactic-co-glycolic acid) (PRP-PLGA)/calcium phosphate cement (CPC) composite scaffold was prepared by incorporating PRP into PLGA/CPC scaffold with unidirectional pore structure, which was fabricated by the unidirectional freeze casting of CPC slurry and the following infiltration of PLGA. The results from in vitro cell experiments and in vivo implantation in femoral defects manifested that incorporation of PRP into PLGA/CPC scaffold improved in vitro cell response (cell attachment, proliferation, and differentiation), and markedly boosted bone formation, angiogenesis and material degradation. The incorporation of PRP into scaffold showed more outstanding improvement in osteogenesis as the scaffolds were used to repair the segmental radial defects, especially at the early stage. The new bone tissues grew along the unidirectional lamellar pores of scaffold. At 12 weeks postimplantation, the segmental radial defects treated with PRP-PLGA/CPC scaffold had almost recuperated, whereas treated with the scaffold without PRP was far from healed. Taken together, the PRP-PLGA/CPC scaffold with unidirectional pore structure is a promising candidate to repair bone defects at various sites.