To develop and validate a deep learning algorithm that predicts the final diagnosis of Alzheimer disease (AD), mild cognitive impairment, or neither at fluorine 18 (18 F) fluorodeoxyglucose (FDG) PET of the brain and compare its performance to that of radiologic readers. Materials and Methods: Prospective 18 F-FDG PET brain images from the Alzheimer's Disease Neuroimaging Initiative (ADNI) (2109 imaging studies from 2005 to 2017, 1002 patients) and retrospective independent test set (40 imaging studies from 2006 to 2016, 40 patients) were collected. Final clinical diagnosis at follow-up was recorded. Convolutional neural network of InceptionV3 architecture was trained on 90% of ADNI data set and tested on the remaining 10%, as well as the independent test set, with performance compared to radiologic readers. Model was analyzed with sensitivity, specificity, receiver operating characteristic (ROC), saliency map, and t-distributed stochastic neighbor embedding. Results: The algorithm achieved area under the ROC curve of 0.98 (95% confidence interval: 0.94, 1.00) when evaluated on predicting the final clinical diagnosis of AD in the independent test set (82% specificity at 100% sensitivity), an average of 75.8 months prior to the final diagnosis, which in ROC space outperformed reader performance (57% [four of seven] sensitivity, 91% [30 of 33] specificity; P , .05). Saliency map demonstrated attention to known areas of interest but with focus on the entire brain. Conclusion: By using fluorine 18 fluorodeoxyglucose PET of the brain, a deep learning algorithm developed for early prediction of Alzheimer disease achieved 82% specificity at 100% sensitivity, an average of 75.8 months prior to the final diagnosis.

Hip fracture risk rises exponentially with age, but there is little knowledge about how fracture-related alterations in hip structure differ from those of aging. We employed computed tomography (CT) imaging to visualize the three-dimensional (3D) spatial distribution of bone mineral density (BMD) in the hip in relation to age- and incident hip fracture. We used inter-subject image registration to integrate 3D hip CT images into a statistical atlas comprising women aged 21-97 years (n=349) and a group of women with (n=74) and without (n=148) incident hip fracture 4-7 years after their imaging session. Voxel-based morphometry was used to generate Student’s t-test statistical maps from the atlas, which indicated regions that were significantly associated with age or with incident hip fracture. Scaling factors derived from inter-subject image registration were employed as measures of bone size. BMD comparisons of young, middle-aged, and older American women showed preservation of load-bearing cortical and trabecular structures with aging, whereas extensive bone loss was observed in other trabecular and cortical regions. In contrast, comparisons of older Icelandic fracture women with age-matched controls showed that hip fracture was associated with a global cortical bone deficit, including both the superior cortical margin and the load-bearing inferior cortex. Bone size comparisons showed larger dimensions in older compared to younger American women and in older Icelandic fracture women compared to controls. The results indicate that older Icelandic women who sustain incident hip fracture have a structural phenotype that cannot be described as an accelerated pattern of normal age-related loss. The fracture-related cortical deficit noted in this study may provide a biomarker of increased hip fracture risk that may be translatable to DXA and other clinical images.

Summary While type 2 diabetes (T2D) is associated with higher skeletal fragility, specific risk stratification remains incompletely understood. We found volumetric bone mineral density, geometry, and serum sclerostin differences between low-fracture risk and high-fracture risk T2D women. These features might help identify T2D individuals at high fracture risk in the future. Introduction Diabetic bone disease, an increasingly recognized complication of type 2 diabetes mellitus (T2D), is associated with high skeletal fragility. Exactly which T2D individuals are at higher risk for fracture, however, remains incompletely understood. Here, we analyzed volumetric bone mineral density (vBMD), geometry, and serum sclerostin levels in two specific T2D subsets with different fracture risk profiles. We examined a T2D group with prior history of fragility fractures (DMFx, assigned high-risk group) and a fracture-free T2D group (DM, assigned low-risk group) and compared their results to nondiabetic controls with (Fx) and without fragility fractures (Co). Methods Eighty postmenopausal women (n=20 per group) underwent quantitative computed tomography (QCT) to compute vBMD and bone geometry of the proximal femur. Additionally, serum sclerostin, vitamin D, parathyroid hormone (PTH), HbA1c, and glomerular filtration rate (GFR) levels were measured. Statistical analyses employed linear regression models. Results DMFx subjects exhibited up to 33 % lower femoral neck vBMD than DM subjects across all femoral sites (−19 %≤ΔvBMD≤−33 %, 0.008≤p≤0.021). Additionally, DMFx subjects showed significantly thinner cortices (−6 %, p=0.046) and a trend toward larger bone volume (+10 %, p= 0.055) relative to DM women and higher serum sclerostin levels when compared to DM (+31.4 %, p=0.013), Fx (+ 25.2 %, p=0.033), and control (+22.4 %, p=0.028) subjects. Conclusion Our data suggest that volumetric bone parameters by QCT and serum sclerostin levels can identify T2D individuals at high risk of fracture and might therefore show promise as clinical tools for fracture risk assessment in T2D. However, future research is needed to establish diabetes-specific QCTand sclerostin-reference databases.

Radiomics is an emerging technology for imaging biomarker discovery and disease-specific personalized treatment management. This paper aims to determine the benefit of using multi-modality radiomics data from PET and MR images in the characterization breast cancer phenotype and prognosis. Eighty-four features were extracted from PET and MR images of 113 breast cancer patients. Unsupervised clustering based on PET and MRI radiomic features created three subgroups. These derived subgroups were statistically significantly associated with tumor grade (p = 2.0 × 10−6), tumor overall stage (p = 0.037), breast cancer subtypes (p = 0.0085), and disease recurrence status (p = 0.0053). The PET-derived first-order statistics and gray level co-occurrence matrix (GLCM) textural features were discriminative of breast cancer tumor grade, which was confirmed by the results of L2-regularization logistic regression (with repeated nested cross-validation) with an estimated area under the receiver operating characteristic curve (AUC) of 0.76 (95% confidence interval (CI) = [0.62, 0.83]). The results of ElasticNet logistic regression indicated that PET and MR radiomics distinguished recurrence-free survival, with a mean AUC of 0.75 (95% CI = [0.62, 0.88]) and 0.68 (95% CI = [0.58, 0.81]) for 1 and 2 years, respectively. The MRI-derived GLCM inverse difference moment normalized (IDMN) and the PET-derived GLCM cluster prominence were among the key features in the predictive models for recurrence-free survival. In conclusion, radiomic features from PET and MR images could be helpful in deciphering breast cancer phenotypes and may have potential as imaging biomarkers for prediction of breast cancer recurrence-free survival.

Understanding the skeletal effects of resistance exercise involves delineating the spatially heterogeneous response of bone to load distributions from different muscle contractions. Bone mineral density (BMD) analyses may obscure these patterns by averaging data from tissues with variable mechano-response. To assess the proximal femoral response to resistance exercise, we acquired pre- and post-training quantitative computed tomography (QCT) images in 22 subjects (25-55 years, 9 males, 13 females) performing two resistance exercises for 16 weeks. One group (N=7) performed 4 sets each of squats and deadlifts, a second group (N=8) performed 4 sets each of standing hip abductions and adductions and a third (COMBO) performed two sets each of squat/deadlift and abduction/adduction exercise. Subjects exercised three times weekly, and the load was adjusted each session to maximum effort. We used voxel-based morphometry (VBM) to visualize BMD distributions. Hip strength computations used finite element modeling (FEM) with stance and fall loading conditions. Cortical and trabecular BMD, and cortical tissue volume employed QCT analysis. For muscle size and density, we analyzed the cross-sectional area (CSA) and mean Hounsfield Unit (HU) in the hip extensor, flexor, abductor and adductor muscle groups. While SQDL increased vertebral BMD, femoral neck cortical BMD and volume, and stance hip strength, ABADD increased trochanteric cortical volume. The COMBO group showed no changes in any parameter. VBM showed different effects of ABADD and SQDL exercise, with the former causing focal changes of trochanteric cortical bone, and the latter showing diffuse changes in the femoral neck and head. ABADD exercise increased adductor CSA and HU, while SQDL exercise increased the hip extensor CSA and HU. In conclusion, we observed different proximal femoral bone and muscle tissue responses to SQDL and ABADD exercise. This study supports VBM and vQCT to quantify the spatially heterogeneous effects of types of muscle contractions on bone.