VAEs with structured image covariance applied to compressed sensing MRI.

Objective.This paper investigates how generative models, trained on ground-truth images, can be used as priors for inverse problems, penalizing reconstructions far from images the generator can produce. The aim is that learned regularization will provide complex data-driven priors to inverse problems while still retaining the control and insight of a variational regularization method. Moreover, unsupervised learning, without paired training data, allows the learned regularizer to remain flexible to changes in the forward problem such as noise level, sampling pattern or coil sensitivities in MRI.Approach.We utilize variational autoencoders that generate not only an image but also a covariance uncertainty matrix for each image. The covariance can model changing uncertainty dependencies caused by structure in the image, such as edges or objects, and provides a new distance metric from the manifold of learned images.Main results.We evaluate these novel generative regularizers on retrospectively sub-sampled real-valued MRI measurements from the fastMRI dataset. We compare our proposed learned regularization against other unlearned regularization approaches and unsupervised and supervised deep learning methods.Significance.Our results show that the proposed method is competitive with other state-of-the-art methods and behaves consistently with changing sampling patterns and noise levels.

Survival of patients who develop solid tumors following hematopoietic stem cell transplantation.

Allogeneic hematopoietic cell transplantation is associated with late adverse effects of therapy, including secondary solid cancers. Most reports address risk factors; however, outcomes after secondary solid cancer development are incompletely described. Our objective was to estimate survival probabilities for transplant recipients dependent on secondary solid cancer subtype. We used a previously identified and published cohort who developed secondary solid cancers following allogeneic transplant. Follow-up for these 112 previously identified patients was extended and their survival probabilities were studied. Median duration of follow-up from the development of secondary cancer for survivors was 11.9 years (range: 0.8-23.4) and 75% were followed >7.0 years. The 5- and 10-year overall survival probabilities were 50% (95% confidence interval (CI): 41-60) and 46% (95% CI: 37-57), respectively. Overall survival varied by secondary cancer type. Secondary cancer was the cause of death in most patients who died following development of melanoma, central nervous system, oral cavity, thyroid, lung, lower gastrointestinal tract and bone cancers. Extended follow-up allowed for the most comprehensive longitudinal evaluation to date of this rare condition. These findings will enhance clinicians' ability to predict outcomes and counsel transplant survivors who develop secondary solid cancers.