Publications by Year: 2018

Stefanik L, Erdman L, Ameis SH, Foussias G, Mulsant BH, Behdinan T, Goldenberg A, O'Donnell LJ, Voineskos AN. Brain-Behavior Participant Similarity Networks Among Youth and Emerging Adults with Schizophrenia Spectrum, Autism Spectrum, or Bipolar Disorder and Matched Controls. Neuropsychopharmacology. 2018;43 (5) :1180-8.Abstract
There is considerable heterogeneity in social cognitive and neurocognitive performance among people with schizophrenia spectrum disorders (SSD), autism spectrum disorders (ASD), bipolar disorder (BD), and healthy individuals. This study used Similarity Network Fusion (SNF), a novel data-driven approach, to identify participant similarity networks based on relationships among demographic, brain imaging, and behavioral data. T1-weighted and diffusion-weighted magnetic resonance images were obtained for 174 adolescents and young adults (aged 16-35 years) with an SSD (n=51), an ASD without intellectual disability (n=38), euthymic BD (n=34), and healthy controls (n=51). A battery of social cognitive and neurocognitive tasks were administered. Data integration, cluster determination, and biological group formation were then obtained using SNF. We identified four new groups of individuals, each with distinct neural circuit-cognitive profiles. The most influential variables driving the formation of the new groups were robustly reliable across embedded resampling techniques. The data-driven groups showed considerably greater differentiation on key social and neurocognitive circuit nodes than groups generated by diagnostic analyses or dimensional social cognitive analyses. The data-driven groups were validated through functional outcome and brain network property measures not included in the SNF model. Cutting across diagnostic boundaries, our approach can effectively identify new groups of people based on a profile of neuroimaging and behavioral data. Our findings bring us closer to disease subtyping that can be leveraged toward the targeting of specific neural circuitry among participant subgroups to ameliorate social cognitive and neurocognitive deficits.Neuropsychopharmacology advance online publication, 6 December; doi:10.1038/npp.2017.274.
Zhang F, Savadjiev P, Cai W, Song Y, Rathi Y, Tunç B, Parker D, Kapur T, Schultz RT, Makris N, et al. Whole Brain White Matter Connectivity Analysis using Machine Learning: An Application to Autism. Neuroimage. 2018;172 :826-37.Abstract
In this paper, we propose an automated white matter connectivity analysis method for machine learning classification and characterization of white matter abnormality via identification of discriminative fiber tracts. The proposed method uses diffusion MRI tractography and a data-driven approach to find fiber clusters corresponding to subdivisions of the white matter anatomy. Features extracted from each fiber cluster describe its diffusion properties and are used for machine learning. The method is demonstrated by application to a pediatric neuroimaging dataset from 149 individuals, including 70 children with autism spectrum disorder (ASD) and 79 typically developing controls (TDC). A classification accuracy of 78.33% is achieved in this cross-validation study. We investigate the discriminative diffusion features based on a two-tensor fiber tracking model. We observe that the mean fractional anisotropy from the second tensor (associated with crossing fibers) is most affected in ASD. We also find that local along-tract (central cores and endpoint regions) differences between ASD and TDC are helpful in differentiating the two groups. These altered diffusion properties in ASD are associated with multiple robustly discriminative fiber clusters, which belong to several major white matter tracts including the corpus callosum, arcuate fasciculus, uncinate fasciculus and aslant tract; and the white matter structures related to the cerebellum, brain stem, and ventral diencephalon. These discriminative fiber clusters, a small part of the whole brain tractography, represent the white matter connections that could be most affected in ASD. Our results indicate the potential of a machine learning pipeline based on white matter fiber clustering.
Trinh TW, Glazer DI, Sadow CA, Sahni AV, Geller NL, Silverman SG. Bladder Cancer Diagnosis with CT Urography: Test Characteristics and Reasons for False-positive and False-negative Results. Abdom Radiol (NY). 2018;43 (3) :663-71.Abstract
PURPOSE: To determine test characteristics of CT urography for detecting bladder cancer in patients with hematuria and those undergoing surveillance, and to analyze reasons for false-positive and false-negative results. METHODS: A HIPAA-compliant, IRB-approved retrospective review of reports from 1623 CT urograms between 10/2010 and 12/31/2013 was performed. 710 examinations for hematuria or bladder cancer history were compared to cystoscopy performed within 6 months. Reference standard was surgical pathology or 1-year minimum clinical follow-up. False-positive and false-negative examinations were reviewed to determine reasons for errors. RESULTS: Ninety-five bladder cancers were detected. CT urography accuracy: was 91.5% (650/710), sensitivity 86.3% (82/95), specificity 92.4% (568/615), positive predictive value 63.6% (82/129), and negative predictive value was 97.8% (568/581). Of 43 false positives, the majority of interpretation errors were due to benign prostatic hyperplasia (n = 12), trabeculated bladder (n = 9), and treatment changes (n = 8). Other causes include blood clots, mistaken normal anatomy, infectious/inflammatory changes, or had no cystoscopic correlate. Of 13 false negatives, 11 were due to technique, one to a large urinary residual, one to artifact. There were no errors in perception. CONCLUSION: CT urography is an accurate test for diagnosing bladder cancer; however, in protocols relying predominantly on excretory phase images, overall sensitivity remains insufficient to obviate cystoscopy. Awareness of bladder cancer mimics may reduce false-positive results. Improvements in CTU technique may reduce false-negative results.
Hassanzadeh E, Alessandrino F, Olubiyi OI, Glazer DI, Mulkern RV, Fedorov A, Tempany CM, Fennessy FM. Comparison of Quantitative Apparent Diffusion Coefficient Parameters with Prostate Imaging Reporting and Data System V2 Assessment for Detection of Clinically Significant Peripheral Zone Prostate Cancer. Abdom Radiol (NY). 2018;43 (5) :1237-44.Abstract
PURPOSE: To compare diagnostic performance of PI-RADSv2 with ADC parameters to identify clinically significant prostate cancer (csPC) and to determine the impact of csPC definitions on diagnostic performance of ADC and PI-RADSv2. METHODS: We retrospectively identified treatment-naïve pathology-proven peripheral zone PC patients who underwent 3T prostate MRI, using high b-value diffusion-weighted imaging from 2011 to 2015. Using 3D slicer, areas of suspected tumor (T) and normal tissue (N) on ADC (b = 0, 1400) were outlined volumetrically. Mean ADCT, mean ADCN, ADCratio (ADCT/ADCN) were calculated. PI-RADSv2 was assigned. Three csPC definitions were used: (A) Gleason score (GS) ≥ 4 + 3; (B) GS ≥ 3 + 4; (C) MRI-based tumor volume >0.5 cc. Performances of ADC parameters and PI-RADSv2 in identifying csPC were measured using nonparametric comparison of receiver operating characteristic curves using the area under the curve (AUC). RESULTS: Eighty five cases met eligibility requirements. Diagnostic performances (AUC) in identifying csPC using three definitions were: (A) ADCT (0.83) was higher than PI-RADSv2 (0.65, p = 0.006); (B) ADCT (0.86) was higher than ADCratio (0.68, p < 0.001), and PI-RADSv2 (0.70, p = 0.04); (C) PI-RADSv2 (0.73) performed better than ADCratio (0.56, p = 0.02). ADCT performance was higher when csPC was defined by A or B versus C (p = 0.038 and p = 0.01, respectively). ADCratio performed better when csPC was defined by A versus C (p = 0.01). PI-RADSv2 performance was not affected by csPC definition. CONCLUSIONS: When csPC was defined by GS, ADC parameters provided better csPC discrimination than PI-RADSv2, with ADCT providing best result. When csPC was defined by MRI-calculated volume, PI-RADSv2 provided better discrimination than ADCratio. csPC definition did not affect PI-RADSv2 diagnostic performance.