Jia Ying

Background: The infrapatellar fat pad (IPFP) plays an important role in the incidence of knee osteoarthritis (OA). Magnetic resonance (MR) signal heterogeneity of the IPFP is related to pathologic changes. In this study, we aimed to investigate whether the IPFP radiomic features have predictive value for incident radiographic knee OA (iROA) 1 year prior to iROA diagnosis. Methods: Data used in this work were obtained from the osteoarthritis initiative (OAI). In this study, iROA was defined as a knee with a baseline Kellgren-Lawrence grade (KLG) of 0 or 1 that further progressed to KLG ≥2 during the follow-up visit. Intermediate-weighted turbo spin-echo knee MR images at the time of iROA diagnosis and 1 year prior were obtained. Five clinical characteristics-age, sex, body mass index, knee injury history, and knee surgery history-were obtained. A total of 604 knees were selected and matched (302 cases and 302 controls). A U-Net segmentation model was independently trained to automatically segment the IPFP. The prediction models were established in the training set (60%). Three main models were generated using (I) clinical characteristics; (II) radiomic features; (III) combined (clinical plus radiomic) features. Model performance was evaluated in an independent testing set (remaining 40%) using the area under the curve (AUC). Two secondary models were also generated using Hoffa-synovitis scores and clinical characteristics. Results: The comparison between the automated and manual segmentations of the IPFP achieved a Dice coefficient of 0.900 (95% CI: 0.891-0.908), which was comparable to that of experienced radiologists. The radiomic features model and the combined model yielded superior AUCs of 0.700 (95% CI: 0.630-0.763) and 0.702 (95% CI: 0.635-0.763), respectively. The DeLong test found no statistically significant difference between the receiver operating curves of the radiomic and combined models (P=0.831); however, both models outperformed the clinical model (P=0.014 and 0.004, respectively). Conclusions: Our results demonstrated that radiomic features of the IPFP are predictive of iROA 1 year prior to the diagnosis, suggesting that IPFP radiomic features can serve as an early quantitative prediction biomarker of iROA.

Abstract Lymph node involvement increases the risk of breast cancer recurrence. An accurate non-invasive assessment of nodal involvement is valuable in cancer staging, surgical risk, and cost savings. Radiomics has been proposed to pre-operatively predict sentinel lymph node (SLN) status; however, radiomic models are known to be sensitive to acquisition parameters. The purpose of this study was to develop a prediction model for preoperative prediction of SLN metastasis using deep learning-based (DLB) features and compare its predictive performance to state-of-the-art radiomics. Specifically, this study aimed to compare the generalizability of radiomics vs DLB features in an independent test set with dissimilar resolution. Dynamic contrast-enhancement images from 198 patients (67 positive SLNs) were used in this study. Of these subjects, 163 had an in-plane resolution of 0.7 × 0.7 mm 2 , which were randomly divided into a training set (approximately 67%) and a validation set (approximately 33%). The remaining 35 subjects with a different in-plane resolution (0.78 × 0.78 mm 2 ) were treated as independent testing set for generalizability. Two methods were employed: (1) conventional radiomics (CR), and (2) DLB features which replaced hand-curated features with pre-trained VGG-16 features. The threshold determined using the training set was applied to the independent validation and testing dataset. Same feature reduction, feature selection, model creation procedures were used for both approaches. In the validation set (same resolution as training), the DLB model outperformed the CR model (accuracy 83% vs 80%). Furthermore, in the independent testing set of the dissimilar resolution, the DLB model performed markedly better than the CR model (accuracy 77% vs 71%). The predictive performance of the DLB model outperformed the CR model for this task. More interestingly, these improvements were seen particularly in the independent testing set of dissimilar resolution. This could indicate that DLB features can ultimately result in a more generalizable model.

Abstract Presence of higher breast density (BD) and persistence over time are risk factors for breast cancer. A quantitatively accurate and highly reproducible BD measure that relies on precise and reproducible whole-breast segmentation is desirable. In this study, we aimed to develop a highly reproducible and accurate whole-breast segmentation algorithm for the generation of reproducible BD measures. Three datasets of volunteers from two clinical trials were included. Breast MR images were acquired on 3 T Siemens Biograph mMR, Prisma, and Skyra using 3D Cartesian six-echo GRE sequences with a fat-water separation technique. Two whole-breast segmentation strategies, utilizing image registration and 3D U-Net, were developed. Manual segmentation was performed. A task-based analysis was performed: a previously developed MR-based BD measure, MagDensity, was calculated and assessed using automated and manual segmentation. The mean squared error (MSE) and intraclass correlation coefficient (ICC) between MagDensity were evaluated using the manual segmentation as a reference. The test-retest reproducibility of MagDensity derived from different breast segmentation methods was assessed using the difference between the test and retest measures (Δ 2-1 ), MSE, and ICC. The results showed that MagDensity derived by the registration and deep learning segmentation methods exhibited high concordance with manual segmentation, with ICCs of 0.986 (95%CI: 0.974-0.993) and 0.983 (95%CI: 0.961-0.992), respectively. For test-retest analysis, MagDensity derived using the registration algorithm achieved the smallest MSE of 0.370 and highest ICC of 0.993 (95%CI: 0.982-0.997) when compared to other segmentation methods. In conclusion, the proposed registration and deep learning whole-breast segmentation methods are accurate and reliable for estimating BD. Both methods outperformed a previously developed algorithm and manual segmentation in the test-retest assessment, with the registration exhibiting superior performance for highly reproducible BD measurements.

Abstract Purpose Breast density (BD) is a significant risk factor for breast cancer, yet current assessment methods lack automation, quantification, and cross-platform consistency. This study aims to evaluate MagDensity, a novel magnetic resonance imaging (MRI)-based quantitative BD measure, for its validity and reliability across different imaging platforms. Methods Ten healthy volunteers participated in this prospective study, undergoing fat-water MRI scans on three scanners: 3T Siemens Prisma, 3T Siemens Biograph mMR, and 1.5T GE Signa. Great effort was made to schedule all scans within a narrow three-hour window on the same day to minimize any potential intraday variations, highlighting the logistical challenges involved. BD was assessed using the MagDensity technique, which included combining magnitude and phase images, applying a fat-water separation technique, employing an automated whole-breast segmentation algorithm, and quantifying the volumetric water fraction. The agreement between measures was analyzed using mean differences, two-tailed t-tests, Pearson’s correlation coefficients, and Bland-Altman plots. Results No statistically significant differences in BD measurements by MagDensity within the same field strength and vendor (3T Siemens), with high correlation (Pearson’s r > 0.99) and negligible mean differences (< 0.2%). Cross-platform comparison between the 3T Siemens and the 1.5T GE scanners showed mean differences of < 5%. After linear calibration, these variations were reduced to insignificant levels, yielding a strong correlation (Pearson’s r > 0.97) and mean differences within ±0.2%. Conclusion MagDensity, an MRI-based BD measure, exhibits robustness and reliability across diverse scanner models, vendors, and field strengths, marking a promising advancement towards standardizing BD measurements across multiple MRI platforms. It provides a valuable tool for monitoring subtle longitudinal changes in BD, which is vital for breast cancer prevention and personalized treatment strategies.

The infrapatellar fat pad (IPFP) plays an important role in the incidence of knee osteoarthritis (OA). However, whether the IPFP can serve as an independent biomarker for OA development is yet unknown. Radiomics is a powerful tool that can extract high-dimensional quantitative features for clinical outcomes. In this work, we proposed a prediction model for incident radiographic knee OA (iROA), using radiomic features of the IPFP, one year prior to diagnosis. The prediction performance was assessed, and our results from 604 knees demonstrated that MR-based radiomic features from the IPFP are predictive of iROA.