What Magnetic Resonance Imaging Techniques got wrong, from 31 theses

Tried and failed

introducing spatial magnetic field gradient for encoding applied to portable low-field magnetic resonance imaging. Outcome: worse than baseline. Reason: gradient bandwidth exceeded RF excitation bandwidth, causing non-uniform excitation and severely degrading SNR

Handheld MRI for Point-of-Care and Educational Applications · MIT · PhD

Tried and failed

standard high-pass filtered phase processing applied to susceptibility-weighted magnetic resonance imaging. Outcome: worse than baseline. Reason: suboptimal echo time selection degraded lesion-to-vein contrast

ADVANCED ACQUISITION AND RECONSTRUCTION METHODS FOR CARDIAC QUANTITATIVE SUSCEPTIBILITY MAPPING AND CLINICAL APPLICATION · Cornell · PhD

Considered and rejected

Considered and rejected: Gradient-echo based functional imaging was rejected in favor of spin-echo based methods (HASTE) for spinal cord imaging due to magnetic field inhomogeneities caused by surrounding vertebrae and tissue interfaces.

Functional MRI Investigation of the Neural Basis of Human Pain Processing in the Brainstem and Spinal Cord; Descending Regulation and Chronic Pain · Queen's University · PhD

Tried and failed

T2* magnetic resonance relaxometry mapping applied to in vivo cardiac tissue characterisation. Reason: poor image quality and severe off-resonance artefacts prevented quantitative analysis

Extracellular vesiclinvestigation of post-myocardial infarction left ventricular remodelling by diffusion tensor cardiovascular magnetic resonance · Imperial · PhD

Tried and failed

gradient-echo shifting in multi-echo sequence applied to quantitative T2 relaxometry myelin water imaging. Outcome: worse than baseline. Reason: increased minimum echo time caused systematic overestimation of short T2 component fraction

Quantitative Microstructural Imaging for Clinical Use · EPFL · PhD

Tried and failed

time-of-flight magnetic resonance angiography applied to luminal defect volume estimation. Outcome: worse than baseline. Reason: flow artifacts and complex flow separation caused severe overestimation of tissue geometry compared to standard anatomical imaging

4D Flow Magnetic Resonance Imaging and Computational Fluid Dynamics for the Study of Complex Blood Flow Patterns in Carotid Webs · Georgia Tech · PhD

Tried and failed

sinusoidal gradient encoding with temporal shuffling applied to fast 3D magnetic resonance imaging. Reason: non-trivial intra-voxel dephasing caused severe image artifacts

On Improving the Acquisition and Reconstruction Of Spatio-Temporal Magnetic Resonance Imaging · MIT · PhD

Considered and rejected

Considered and rejected: Multi-echo T2 and mcDESPOT relaxometry were rejected for myelin water imaging due to model estimation instability and failure to account for compartmental exchange

Physiology and plasticity of myelinated long-range projections in the adult brain · Oxford · PhD

Tried and failed

diffusion-ordered NMR spectroscopy at low field strength applied to complex biofluid mixtures. Reason: lower magnetic field strength caused split-peak fitting artefacts in the diffusion dimension

Evaluation and application of diffusion coefficient measurements in NMR-based metabolomic analysis of human blood serum, urine and other biomaterials · Imperial · PhD

Lost to a baseline

In phantom regions with smooth, low-intensity tissue relaxation, an 8-phase-cycle discrete acquisition (DPC-8) achieved a lower coefficient of variation (CoV) than fm-bSSFP min-align due to reconstruction noise artifacts dominating imperceptible bands.

Debanding in Frequency-Modulated bSSFP MRI: High-Resolution Neuroimaging · Johns Hopkins · PhD

Tried and failed

strain correction model with simplified deformation assumptions applied to diffusion tensor magnetic resonance imaging. Outcome: did not generalise. Reason: oversimplified deformation assumptions caused deviations from measured experimental diffusion data

Computational modelling of diffusion tensor cardiovascular magnetic resonance: effects of membrane permeability, perfusion and strain · Imperial · PhD

Tried and failed

retrospectively undersampled training for accelerated imaging reconstruction applied to prospective magnetic resonance image reconstruction. Outcome: did not generalise. Reason: prospective acquisition contains spatial distortions and contrast shifts not simulated by retrospective undersampling

Combining Model Driven and Data Driven Approaches for Inverse Problems in Parameter Estimation and Image Reconstruction: From Modelling to Validation · EPFL · PhD

Tried and failed

Intensity attenuation correction at large rotation angles applied to Multi-orientation low-field magnetic resonance imaging. Outcome: no signal. Reason: Severe SNR loss beyond thirty degrees of rotation degraded image quality substantially

In vivo imaging of collagen fibre structures using dedicated low-field MRI · Imperial · PhD

Tried and failed

software-only in-plane super-resolution without physical displacement applied to magnetic resonance imaging. Reason: scanner hardware constraints prevent field-of-view shifts along required axes without physical target motion

Image Guided High Precision Robotic Positioning in MRI for Medical Applications · Georgia Tech · PhD

Tried and failed

polynomial fitting on coarse tissue segmentations applied to inhomogeneity calibration in magnetic resonance imaging. Outcome: did not generalise. Reason: failed to accurately model low-amplitude signal regions across heterogeneous tissue types

Imaging Glutamate In The Human Brain At Ultra-High Magnetic Field: Advances And Applications · Penn · PhD

Tried and failed

unrolled ADMM optimization with high penalty parameter applied to magnetic resonance quantitative susceptibility mapping. Outcome: unstable. Reason: excessive penalty parameter weighting caused divergent fidelity costs and shadow artifacts around high-contrast regions

PHYSICS-BASED DEEP LEARNING METHODS FOR MAGNETIC RESONANCE DATA SAMPLING, IMAGE RECONSTRUCTION AND QUANTITATIVE SUSCEPTIBILITY MAPPING · Cornell · PhD

Tried and failed

spatially-varying T2* correction applied to diffusion tensor magnetic resonance imaging. Outcome: worse than baseline. Reason: caused significant noise enhancement, reducing accuracy and precision of derived diffusion parameters

Development of novel magnetic resonance methods for advanced parametric mapping of the right ventricle · Imperial · PhD

Considered and rejected

Considered and rejected: Rejected imaging (MRI or ultrasound) as a primary diagnostic tool for mid-portion Achilles tendinopathy due to discrepancies between detectable imaging pathology and clinical symptoms.

Action Observation Therapy in the treatment of mid-portion Achilles Tendinopathy · University College Dublin · PhD

Considered and rejected

Considered and rejected: Supine cross-sectional imaging (CT and MR angiography) rejected as a primary dynamic diagnostic modality because supine positioning fails to reflect the functional thoracic outlet anatomy present in upright postures

Arterielles Thoracic Outlet Syndrome: Die Bedeutung der Farbkodierten Duplexsonografie in der vaskulären Diagnostik Arterial thoracic outlet syndrome: the importance of color-coded duplex sonography in vascular diagnostics · University of Marburg · PhD

Considered and rejected

Considered and rejected: Rejected routine use of MRI for post-treatment follow-up imaging because persistent signal elevations remain detectable long after clinical symptom resolution.

Chirurgische Behandlung und Outcome in einer konsekutiven Serie von 237 Patienten mit Spondylodiszitis Neurosurgical Management and Outcome Parameters in 237 Patients with Spondylodiscitis · University of Marburg · PhD

Tried and failed

front-line imaging screening instead of reflex testing applied to population cancer screening. Outcome: infeasible cost. Reason: front-line MRI strategies were dominated and not cost-effective compared to standard biomarker screening with reflex imaging

The Effect of TNF-α Inhibitors Therapy on the Growth of the Prostate Gland & Cost-Effectiveness of Biparametric Magnetic Resonance Imaging as a Front-line Test for Prostate Cancer · Harvard · Master's

Tried and failed

systolic fractional anisotropy via diffusion tensor imaging applied to chronic myocardial infarction segment detection. Outcome: no signal. Reason: poor diagnostic performance and moderate interobserver reproducibility in segment detection

Extracellular vesiclinvestigation of post-myocardial infarction left ventricular remodelling by diffusion tensor cardiovascular magnetic resonance · Imperial · PhD

Considered and rejected

Considered and rejected: Relying purely on MRI imaging for definitive diagnosis without biopsy was rejected due to high rates of misclassification.

Der tenosynoviale Riesenzelltumor - eine retroperspektive Analyse über 15 Jahre · Tübingen · PhD

Tried and failed

single structural imaging sequence screening applied to prostate cancer lesion classification. Reason: produced a high proportion of equivocal indeterminate scores, requiring multiparametric sequences for diagnostic certainty

Evaluation of an MRI-based screening pathway for prostate cancer · Imperial · PhD

Tried and failed

multimodal feature concatenation with neuroimaging applied to cognitive decline trajectory prediction. Outcome: worse than baseline. Reason: adding high-dimensional fMRI features degraded classifier performance compared to biomarker baselines

Developing machine learning and optimization models to classify Super-Agers and Cognitive Decliners · Iowa State · PhD

Considered and rejected

Considered and rejected: Rejected standard/global 1D clustering (e.g. k-means across all features/subjects) because it overgeneralizes and obscures localized, subpopulation-specific neuroimaging patterns

Granular factoring into neuroimaging dynamic across space, time, and modality · Georgia Tech · PhD

Tried and failed

2.5D fully convolutional network applied to cardiac magnetic resonance volumetric image segmentation. Outcome: worse than baseline. Reason: Inter-slice spatial correlations provided less effective features than pure 2D slice-wise representation

Automated cardiac segmentation pipeline and motion analysis in rodent models of pulmonary hypertension · Imperial · PhD

Tried and failed

linear regression for binary-derived continuous score prediction applied to neuroimaging biomarker prediction. Outcome: worse than baseline. Reason: produced very low specificity and high false positive rates compared to logistic classification

Statistical and Machine Learning Methods for Neuroimaging and Neurocognitive Data · Penn · PhD

Tried and failed

mass-univariate voxelwise testing with Bonferroni correction applied to spatial functional imaging data. Reason: overly conservative thresholding led to high false negative rates (>30%)

Next-generation functional data analysis for imaging and spatial data · Iowa State · PhD

Tried and failed

multimodal imaging fusion with circulating biomarker panels applied to disease progression survival prediction. Outcome: no signal. Reason: imaging features provided redundant predictive information relative to blood biomarkers, failing to improve AUROC significantly

Predicting Progression from Mild Cognitive Impairment to Alzheimer’s Disease using Blood-Based Biomarkers and Amyloid-Beta PET: A Survival Analysis Approach · Harvard · Master's

Tried and failed

mechanical rotation for spatial encoding applied to magnetic resonance imaging. Outcome: too slow. Reason: slow speed, high cost, and actuator precision limitations compared to electronic gradient coils

Handheld MRI for Point-of-Care and Educational Applications · MIT · PhD

Lost to a baseline

In small-sample finger imaging, peak sagittal SNR was higher for conventional LC loop (135.9 dB A.U.) and peak axial SNR for solenoid (129.9 dB A.U.) than solenoidal dipole (123 dB A.U.).

Advanced RF Hardware Solutions for Improved Imaging Performance and Safety in Ultra-High and Very-Low-Field MRI · University at Buffalo · PhD

Considered and rejected

Considered and rejected: Rejected PC-MRA and 3D TOF-MRA for abdominal vascular imaging due to excessive scan acquisition times incompatible with single breath-holds.

MRI changes in visceral fat in Crohn’s Disease · Nottingham · PhD

Left open

Determine whether weak fMRI functional dissociations across individuals stem from biological variation or imaging resolution limits. Blocker: Broad exploratory research question without a specified methodology or dataset

Examining the Organization and Functions of Parallel Association Networks within Individuals · Harvard · PhD

Left open

Test crossmodal compositional decoding in lmSTC using fMRI experiments with single sentence and image presentations containing delay periods. Blocker: Requires conducting new fMRI neuroimaging experiments with human subjects and specialized physical apparatus (MRI scanner).

The Neural Representation of Structured Complex Ideas in Different Formats · Harvard · PhD

Left open

Evaluate cross-participant fMRI semantic decoders on brain imaging data collected from individuals with brain lesions and language disorders like aphasia. Blocker: Requires fMRI acquisition apparatus, clinical access to patients with aphasia/brain lesions, and specialized human subjects data.

Modeling and decoding semantic representations in language and vision · UT Austin · PhD

Left open

Analyze fMRI BOLD activation outside the visual cortex to study cognitive modulation during geospatial mental rotation tasks. Blocker: Requires private 3T fMRI experimental neuroimaging dataset and participant behavioral data from the thesis study

Neurological Role of Cartographic Visual Contrast in Geospatial Cognition · University of Oregon · PhD

Left open

Adjudicate whether the supramarginal gyrus encodes phrasing-invariant semantic information, functional affordances, sequence order, or phonological representations using targeted fMRI MVPA experiments. Blocker: Requires collecting new human fMRI neuroimaging data or access to private trial-level fMRI datasets with specific linguistic controls.

The Neural Representation of Structured Complex Ideas in Different Formats · Harvard · PhD

Left open

Identify specific shared kinematic, physical, and semantic features encoded by common neural representations across actions and object events using fMRI. Blocker: Requires collecting new controlled fMRI experiment data or accessing proprietary neuroimaging datasets.

Agents, objects, and actions: Investigations into the neural representation of dynamic information · Harvard · PhD

Left open

Apply the pervasive bias eIRT model to multi-test cognitive battery datasets and functional neuroimaging data. Blocker: Access to specific multi-test cognitive battery and paired in situ functional neuroimaging datasets with demographic variables.

Explaining Racial Discrepancies in Verbal Memory Assessment with Joint Estimation of Life Course Social Inequalities and Measurement Bias · UT Southwestern

Left open

Conduct in vivo studies to evaluate the sensitivity and reliability of higher-order radiomic parameters in diagnosing tendinopathy. Blocker: Requires in vivo clinical/animal studies, medical ultrasound imaging, and physical subjects.

Radiomic Analysis of Sonograms for the Detection of Tendon Damage · Virginia Tech · Master's

Left open

Optimize a 3T MRI sequence balancing spatial resolution and acquisition time for imaging glenoid cartilage wear in shoulder hemiarthroplasty. Blocker: Requires access to a 3T MRI scanner and physical shoulder specimens/patients.

Hemiarthroplasty in patients aged 50-65 years old: In-vitro assessment of prosthetic materials and rotator cuff status on glenoid wear · Imperial · PhD

Left open

Perform muscular segmentation and analysis on high-resolution 3T MRI VIBE sequence spinal and lower-limb scans. Blocker: Requires private 3T MRI patient imaging data collected during the study.

Understanding spinal loading in persons with traumatic lower limb amputations · Imperial · PhD