About

The Laboratory of Mathematics in Imaging (LMI) is focused on the application of mathematical theory, analysis, modeling, and signal processing to medical imaging. Research projects within the group cover both novel theoretical contributions and translational clinical efforts. The research team combine strengths in computer science and mathematics with radiology, neuroscience, and novel MRI sequence developmentLearn more

Recent Publications

A Review and Experimental Evaluation of Deep Learning Methods for MRI Reconstruction

Pal A, Rathi Y. A Review and Experimental Evaluation of Deep Learning Methods for MRI Reconstruction. J Mach Learn Biomed Imaging. 2022;1 :001.Abstract
Following the success of deep learning in a wide range of applications, neural network-based machine-learning techniques have received significant interest for accelerating magnetic resonance imaging (MRI) acquisition and reconstruction strategies. A number of ideas inspired by deep learning techniques for computer vision and image processing have been successfully applied to nonlinear image reconstruction in the spirit of compressed sensing for accelerated MRI. Given the rapidly growing nature of the field, it is imperative to consolidate and summarize the large number of deep learning methods that have been reported in the literature, to obtain a better understanding of the field in general. This article provides an overview of the recent developments in neural-network based approaches that have been proposed specifically for improving parallel imaging. A general background and introduction to parallel MRI is also given from a classical view of k-space based reconstruction methods. Image domain based techniques that introduce improved regularizers are covered along with k-space based methods which focus on better interpolation strategies using neural networks. While the field is rapidly evolving with plenty of papers published each year, in this review, we attempt to cover broad categories of methods that have shown good performance on publicly available data sets. Limitations and open problems are also discussed and recent efforts for producing open data sets and benchmarks for the community are examined.
Read more

MR Fingerprinting with b-Tensor Encoding for Simultaneous Quantification of Relaxation and Diffusion in a Single Scan

Afzali M, Mueller L, Sakaie K, Hu S, Chen Y, Szczepankiewicz F, Griswold MA, Jones DK, Ma D. MR Fingerprinting with b-Tensor Encoding for Simultaneous Quantification of Relaxation and Diffusion in a Single Scan. Magn Reson Med. 2022.Abstract
PURPOSE: Although both relaxation and diffusion imaging are sensitive to tissue microstructure, studies have reported limited sensitivity and robustness of using relaxation or conventional diffusion alone to characterize tissue microstructure. Recently, it has been shown that tensor-valued diffusion encoding and joint relaxation-diffusion quantification enable more reliable quantification of compartment-specific microstructural properties. However, scan times to acquire such data can be prohibitive. Here, we aim to simultaneously quantify relaxation and diffusion using MR fingerprinting (MRF) and b-tensor encoding in a clinically feasible time. METHODS: We developed multidimensional MRF scans (mdMRF) with linear and spherical b-tensor encoding (LTE and STE) to simultaneously quantify T1, T2, and ADC maps from a single scan. The image quality, accuracy, and scan efficiency were compared between the mdMRF using LTE and STE. Moreover, we investigated the robustness of different sequence designs to signal errors and their impact on the maps. RESULTS: T1 and T2 maps derived from the mdMRF scans have consistently high image quality, while ADC maps are sensitive to different sequence designs. Notably, the fast imaging steady state precession (FISP)-based mdMRF scan with peripheral pulse gating provides the best ADC maps that are free of image distortion and shading artifacts. CONCLUSION: We demonstrated the feasibility of quantifying T1, T2, and ADC maps simultaneously from a single mdMRF scan in around 24 s/slice. The map quality and quantitative values are consistent with the reference scans.
Read more

Deep Learning-Based Lesion Subtyping and Prediction of Clinical Outcomes in COVID-19 Pneumonia Using Chest CT

Bermejo-Peláez D, San José Estépar R, Fernández-Velilla M, Palacios Miras C, Gallardo Madueño G, Benegas M, Gotera Rivera C, Cuerpo S, Luengo-Oroz M, Sellarés J, et al. Deep Learning-Based Lesion Subtyping and Prediction of Clinical Outcomes in COVID-19 Pneumonia Using Chest CT. Sci Rep. 2022;12 (1) :9387.Abstract
The main objective of this work is to develop and evaluate an artificial intelligence system based on deep learning capable of automatically identifying, quantifying, and characterizing COVID-19 pneumonia patterns in order to assess disease severity and predict clinical outcomes, and to compare the prediction performance with respect to human reader severity assessment and whole lung radiomics. We propose a deep learning based scheme to automatically segment the different lesion subtypes in nonenhanced CT scans. The automatic lesion quantification was used to predict clinical outcomes. The proposed technique has been independently tested in a multicentric cohort of 103 patients, retrospectively collected between March and July of 2020. Segmentation of lesion subtypes was evaluated using both overlapping (Dice) and distance-based (Hausdorff and average surface) metrics, while the proposed system to predict clinically relevant outcomes was assessed using the area under the curve (AUC). Additionally, other metrics including sensitivity, specificity, positive predictive value and negative predictive value were estimated. 95% confidence intervals were properly calculated. The agreement between the automatic estimate of parenchymal damage (%) and the radiologists' severity scoring was strong, with a Spearman correlation coefficient (R) of 0.83. The automatic quantification of lesion subtypes was able to predict patient mortality, admission to the Intensive Care Units (ICU) and need for mechanical ventilation with an AUC of 0.87, 0.73 and 0.68 respectively. The proposed artificial intelligence system enabled a better prediction of those clinically relevant outcomes when compared to the radiologists' interpretation and to whole lung radiomics. In conclusion, deep learning lesion subtyping in COVID-19 pneumonia from noncontrast chest CT enables quantitative assessment of disease severity and better prediction of clinical outcomes with respect to whole lung radiomics or radiologists' severity score.
Read more

"A Theta Burst Stimulation on Pre-SMA: Proof-of-Concept of Transcranial Magnetic Stimulation in Gambling Disorder"

Salerno L, Grassi E, Makris N, Pallanti S. "A Theta Burst Stimulation on Pre-SMA: Proof-of-Concept of Transcranial Magnetic Stimulation in Gambling Disorder". J Gambl Stud. 2022 :1-9.Abstract
Gambling Disorder (GD) is a condition constituting a public health concern, with a burden of harm which is much greater than that of drug addiction. Patients with GD are generally reluctant to pharmacologic treatment and seem to prefer nonpharmacological interventions. Therefore, this proof-of-concept study aimed to investigate the feasibility of continuous Theta Burst Stimulation (cTBS) on the pre-SMA in six patients (5 males, 1 female), aged 30-64 years, with a DSM-5 diagnosis of Gambling Disorder and no comorbid mood disorders. Participants received over 10 sessions of Continuous TBS (cTBS) over pre-SMA bilaterally and have been evaluated using rating scales, including the PG-YBOCS and the CGI, before treatment (T0), at day 10 of treatment (T1) and at day 30 after treatment (T2); cTBS intervention was safe and without side effects. Since the design of our study does not allow us to draw conclusions on the effectiveness of the intervention with respect to the improvement of the functioning of the subject with GD, a more in-depth study, including a sham condition, neurocognitive measures of disinhibition and decision making, and collecting follow-up data on the sustained effect of TBS over a longer period is ongoing.
Read more

Traction Bronchiectasis/Bronchiolectasis on CT Scans in Relationship to Clinical Outcomes and Mortality: The COPDGene Study

Hata A, Hino T, Putman RK, Yanagawa M, Hida T, Menon AA, Honda O, Yamada Y, Nishino M, Araki T, et al. Traction Bronchiectasis/Bronchiolectasis on CT Scans in Relationship to Clinical Outcomes and Mortality: The COPDGene Study. Radiology. 2022 :212584.Abstract
Background The clinical impact of interstitial lung abnormalities (ILAs) on poor prognosis has been reported in many studies, but risk stratification in ILA will contribute to clinical practice. Purpose To investigate the association of traction bronchiectasis/bronchiolectasis index (TBI) with mortality and clinical outcomes in individuals with ILA by using the COPDGene cohort. Materials and Methods This study was a secondary analysis of prospectively collected data. Chest CT scans of participants with ILA for traction bronchiectasis/bronchiolectasis were evaluated and outcomes were compared with participants without ILA from the COPDGene study (January 2008 to June 2011). TBI was classified as follows: TBI-0, ILA without traction bronchiectasis/bronchiolectasis; TBI-1, ILA with bronchiolectasis but without bronchiectasis or architectural distortion; TBI-2, ILA with mild to moderate traction bronchiectasis; and TBI-3, ILA with severe traction bronchiectasis and/or honeycombing. Clinical outcomes and overall survival were compared among the TBI groups and the non-ILA group by using multivariable linear regression model and Cox proportional hazards model, respectively. Results Overall, 5295 participants (median age, 59 years; IQR, 52-66 years; 2779 men) were included, and 582 participants with ILA and 4713 participants without ILA were identified. TBI groups were associated with poorer clinical outcomes such as quality of life scores in the multivariable linear regression model (TBI-0: coefficient, 3.2 [95% CI: 0.6, 5.7; P = .01]; TBI-1: coefficient, 3.3 [95% CI: 1.1, 5.6; P = .003]; TBI-2: coefficient, 7.6 [95% CI: 4.0, 11; P < .001]; TBI-3: coefficient, 32 [95% CI: 17, 48; P < .001]). The multivariable Cox model demonstrated that ILA without traction bronchiectasis (TBI-0-1) and with traction bronchiectasis (TBI-2-3) were associated with shorter overall survival (TBI-0-1: hazard ratio [HR], 1.4 [95% CI: 1.0, 1.9; P = .049]; TBI-2-3: HR, 3.8 [95% CI: 2.6, 5.6; P < .001]). Conclusion Traction bronchiectasis/bronchiolectasis was associated with poorer clinical outcomes compared with the group without interstitial lung abnormalities; TBI-2 and 3 were associated with shorter survival. © RSNA, 2022 Online supplemental material is available for this article. See also the editorial by Lee and Im in this issue.
Read more

Continuous Infusion of Factor VIII and von Willebrand Factor in Surgery: Trials with pdFVIII LFB or pdVWF LFB in Patients with Bleeding Disorders

Windyga J, Guillet B, Rugeri L, Fournel A, Stefanska-Windyga E, Chamouard V, Pujol S, Henriet C, Bridey F, Negrier C. Continuous Infusion of Factor VIII and von Willebrand Factor in Surgery: Trials with pdFVIII LFB or pdVWF LFB in Patients with Bleeding Disorders. Thromb Haemost. 2022.Abstract
INTRODUCTION: A plasma-derived factor VIII product (pdFVIII, Factane 100 or 200 IU/mL) and a plasma-derived von Willebrand factor product (pdVWF, Wilfactin 100 IU/mL) are approved for replacement therapy by intravenous bolus injections in haemophilia A (HA) and von Willebrand disease (VWD), respectively. However, in situations requiring intensive treatment, continuous infusion (CI) may be desirable to better control target plasma factor levels. AIM: To evaluate the perioperative haemostatic efficacy and safety of these concentrates administered by CI. METHODS: Three phase III trials were conducted. Adults with HA (FVIII:C <1%) (Studies 1 and 2) or VWD (VWF:RCo <20%) (Study 3) received a preoperative bolus followed by CI of undiluted concentrate for at least 6 days. Bolus doses and CI rates were based on individual recovery and clearance, respectively. Initial infusion rate had to be higher for 48 hours for HA and 24 hours for VWD patients to anticipate potential fluctuations of factor concentrations during major surgery. Target levels of FVIII:C in HA and VWF:RCo in VWD were 80 and 70 IU/dL, respectively. Efficacy was assessed using a global haemostatic efficacy score. RESULTS: Studies 1, 2, and 3 included 12, 4 and 6 patients, respectively. Efficacy outcomes were excellent/good in all 22 major surgeries including 18 orthopaedic procedures. Most daily measured FVIII and VWF levels (92%) were on target. No safety concerns, thrombotic events or inhibitors were identified. CONCLUSION: pdFVIII and pdVWF administered by CI represent an effective and safe alternative to bolus injections in patients with severe HA or VWD undergoing surgery.
Read more
More