Fidelity imposed network edit (FINE) for solving ill-posed image reconstruction

Deep learning (DL) is increasingly used to solve ill-posed inverse problems in medical imaging, such as reconstruction from noisy and/or incomplete data, as DL offers advantages over conventional methods that rely on explicit image features and hand engineered priors. However, supervised DL-based methods may achieve poor performance when the test data deviates from the training data, for example, when it has pathologies not encountered in the training data. Furthermore, DL-based image reconstructions do not always incorporate the underlying forward physical model, which may improve performance. Therefore, in this work we introduce a novel approach, called fidelity imposed network edit (FINE), which modifies the weights of a pre-trained reconstruction network for each case in the testing dataset. This is achieved by minimizing an unsupervised fidelity loss function that is based on the forward physical model. FINE is applied to two important inverse problems in neuroimaging: quantitative susceptibility mapping (QSM) and under-sampled image reconstruction in MRI. Our experiments demonstrate that FINE can improve reconstruction accuracy.

Admitted to treatment without diagnosis: The status of known diagnoses in US addictions treatment centres

BACKGROUND

Diagnostic and treatment gaps exist for those with substance use disorders (SUD). SUD diagnostics in treatment have been critically under-investigated. We investigated known diagnoses in US addiction treatment centres receiving state or federal funding.

METHODS

Cross-sectional risk factor analysis of variables associated with known diagnoses in US inpatient and outpatient treatment facilities receiving public funding from 2000 to 2014 (N = 8,692,362).

RESULTS

Of all clients, 59.68% had unknown diagnoses. However, 14 states reported zero diagnostic data, representing the majority of unknown diagnoses. Of the reporting states (n = 4,161,021), 16.05% had unknown diagnoses. Those with known diagnoses were represented by four subcategories of SUD (73.22%), No Diagnosis (0.81%), Alcohol Intoxication (1.36%), and Mental Health (8.57%). Uninsured status was one meaningful association and was less likely to have a known diagnosis. Insurance status, referral source, service setting, and wait to admission significantly predicted known diagnosis (p <  0.0001).

CONCLUSIONS

A large proportion of admissions are missing diagnoses. Many states do not report any diagnostics. Variables predictive of a known diagnosis highlight administrative inconsistencies. Policies requiring SUD diagnosis documentation can improve assessment of treatment appropriateness and level of care, programme evaluation, and inform the funding allocation for relevant treatment efforts.