Utilizing a Digital Swarm Intelligence Platform to Improve Consensus Among Radiologists and Exploring Its Applications.

Radiologists today play a central role in making diagnostic decisions and labeling images for training and benchmarking artificial intelligence (AI) algorithms. A key concern is low inter-reader reliability (IRR) seen between experts when interpreting challenging cases. While team-based decisions are known to outperform individual decisions, inter-personal biases often creep up in group interactions which limit nondominant participants from expressing true opinions. To overcome the dual problems of low consensus and interpersonal bias, we explored a solution modeled on bee swarms. Two separate cohorts, three board-certified radiologists, (cohort 1), and five radiology residents (cohort 2) collaborated on a digital swarm platform in real time and in a blinded fashion, grading meniscal lesions on knee MR exams. These consensus votes were benchmarked against clinical (arthroscopy) and radiological (senior-most radiologist) standards of reference using Cohen's kappa. The IRR of the consensus votes was then compared to the IRR of the majority and most confident votes of the two cohorts. IRR was also calculated for predictions from a meniscal lesion detecting AI algorithm. The attending cohort saw an improvement of 23% in IRR of swarm votes (k = 0.34) over majority vote (k = 0.11). Similar improvement of 23% in IRR (k = 0.25) in 3-resident swarm votes over majority vote (k = 0.02) was observed. The 5-resident swarm had an even higher improvement of 30% in IRR (k = 0.37) over majority vote (k = 0.07). The swarm consensus votes outperformed individual and majority vote decision in both the radiologists and resident cohorts. The attending and resident swarms also outperformed predictions from a state-of-the-art AI algorithm.

Longitudinal MR spectroscopy to detect progression in patients with lower-grade glioma in the surveillance phase.

Background: Monitoring lower-grade gliomas (LrGGs) for disease progression is made difficult by the limits of anatomical MRI to distinguish treatment related tissue changes from tumor progression. MR spectroscopic imaging (MRSI) offers additional metabolic information that can help address these challenges. The goal of this study was to compare longitudinal changes in multiparametric MRI, including diffusion weighted imaging, perfusion imaging, and 3D MRSI, for LrGG patients who progressed at the final time-point and those who remained clinically stable. Methods: Forty-one patients with LrGG who were clinically stable were longitudinally assessed for progression. Changes in anatomical, diffusion, perfusion and MRSI data were acquired and compared between patients who remained clinically stable and those who progressed. Results: Thirty-one patients remained stable, and 10 patients progressed. Over the study period, progressed patients had a significantly greater increase in normalized choline, choline-to-N-acetylaspartic acid index (CNI), normalized creatine, and creatine-to-N-acetylaspartic acid index (CRNI), than stable patients. CRNI was significantly associated with progression status and WHO type. Progressed astrocytoma patients had greater increases in CRNI than stable astrocytoma patients. Conclusions: LrGG patients in surveillance with tumors that progressed had significantly increasing choline and creatine metabolite signals on MRSI, with a trend of increasing T2 FLAIR volumes, compared to LrGG patients who remained stable. These data show that MRSI can be used in conjunction with anatomical imaging studies to gain a clearer picture of LrGG progression, especially in the setting of clinical ambiguity.

Clinical Assessment of Deep Learning-based Super-Resolution for 3D Volumetric Brain MRI.

Artificial intelligence (AI)-based image enhancement has the potential to reduce scan times while improving signal-to-noise ratio (SNR) and maintaining spatial resolution. This study prospectively evaluated AI-based image enhancement in 32 consecutive patients undergoing clinical brain MRI. Standard-of-care (SOC) three-dimensional (3D) T1 precontrast, 3D T2 fluid-attenuated inversion recovery, and 3D T1 postcontrast sequences were performed along with 45% faster versions of these sequences using half the number of phase-encoding steps. Images from the faster sequences were processed by a Food and Drug Administration-cleared AI-based image enhancement software for resolution enhancement. Four board-certified neuroradiologists scored the SOC and AI-enhanced image series independently on a five-point Likert scale for image SNR, anatomic conspicuity, overall image quality, imaging artifacts, and diagnostic confidence. While interrater κ was low to fair, the AI-enhanced scans were noninferior for all metrics and actually demonstrated a qualitative SNR improvement. Quantitative analyses showed that the AI software restored the high spatial resolution of small structures, such as the septum pellucidum. In conclusion, AI-based software can achieve noninferior image quality for 3D brain MRI sequences with a 45% scan time reduction, potentially improving the patient experience and scanner efficiency without sacrificing diagnostic quality. Keywords: MR Imaging, CNS, Brain/Brain Stem, Reconstruction Algorithms © RSNA, 2022.

Interinstitutional Portability of a Deep Learning Brain MRI Lesion Segmentation Algorithm.

PURPOSE: To assess how well a brain MRI lesion segmentation algorithm trained at one institution performed at another institution, and to assess the effect of multi-institutional training datasets for mitigating performance loss. MATERIALS AND METHODS: In this retrospective study, a three-dimensional U-Net for brain MRI abnormality segmentation was trained on data from 293 patients from one institution (IN1) (median age, 54 years; 165 women; patients treated between 2008 and 2018) and tested on data from 51 patients from a second institution (IN2) (median age, 46 years; 27 women; patients treated between 2003 and 2019). The model was then trained on additional data from various sources: (a) 285 multi-institution brain tumor segmentations, (b) 198 IN2 brain tumor segmentations, and (c) 34 IN2 lesion segmentations from various brain pathologic conditions. All trained models were tested on IN1 and external IN2 test datasets, assessing segmentation performance using Dice coefficients. RESULTS: The U-Net accurately segmented brain MRI lesions across various pathologic conditions. Performance was lower when tested at an external institution (median Dice score, 0.70 [IN2] vs 0.76 [IN1]). Addition of 483 training cases of a single pathologic condition, including from IN2, did not raise performance (median Dice score, 0.72; P = .10). Addition of IN2 training data with heterogeneous pathologic features, representing only 10% (34 of 329) of total training data, increased performance to baseline (Dice score, 0.77; P < .001). This final model produced total lesion volumes with a high correlation to the reference standard (Spearman r = 0.98). CONCLUSION: For brain MRI lesion segmentation, adding a modest amount of relevant training data from an external institution to a previously trained model supported successful application of the model to this external institution.Keywords: Neural Networks, Brain/Brain Stem, Segmentation Supplemental material is available for this article. © RSNA, 2021.

Neuro-Oncology Practice Clinical Debate: FDG PET to differentiate glioblastoma recurrence from treatment-related changes.

The ability to accurately differentiate treatment-related changes (ie, pseudoprogression and radiation necrosis) from recurrent glioma remains a critical diagnostic problem in neuro-oncology. Because these entities are treated differently and have vastly different outcomes, accurate diagnosis is necessary to provide optimal patient care. In current practice, this diagnostic quandary commonly requires either serial imaging or histopathologic tissue confirmation. In this article, experts in the field debate the utility of 2-deoxy-2[18F]fluoro-d-glucose positron emission tomography (FDG PET) as an imaging tool to distinguish tumor recurrence from treatment-related changes in a patient with glioblastoma and progressive contrast enhancement on magnetic resonance (MR) following chemoradiotherapy.

Imaging, Histopathologic, and Treatment Nuances of Pulmonary Carcinosarcoma.

A 76-year-old female with coronary artery disease, chronic obstructive pulmonary disease, diabetes mellitus type II, and 40 pack-year smoking history presented with a four-day history of cough, productive of green-yellow sputum. Chest X-ray revealed opacification of the left upper lung field, and computed tomography (CT) of the chest showed a large cavitary lesion invading the T2-T3 vertebral bodies, extending into the epidural space, giving rise to mild cord compression. Biopsy of the lesion revealed a poorly differentiated neoplasm composed of distinct epithelial and mesenchymal components, consistent with carcinosarcoma. A metastatic workup was negative. Primary lung carcinosarcoma is a rare tumour that can demonstrate an especially aggressive clinical course; diagnosis is often nuanced by limited sampling at initial presentation, especially in a setting of advanced disease and debility that precludes consideration for upfront resection or more extensive, invasive sampling.

It's not what it looks like: atypical rash in cryoglobulinaemic vasculitis.

A 48-year-old man with a history of intravenous drug use and chronic, untreated hepatitis C presented to the emergency room with acute bilateral lower extremity swelling, erythema and maculopapular rash. Serum C4 levels were low, but dermatology felt the rash was due to venous stasis dermatitis. The patient was discharged with compression stockings, but returned to the hospital 5 days later with no improvement in his symptoms. A more extensive laboratory workup revealed hepatitis C viral load of 4 million, elevated serum cryoglobulins, and skin biopsy showing leucocytoclastic vasculitis. He was treated with oral prednisone, with complete resolution of his symptoms after 2 weeks. He was scheduled for follow-up in gastroenterology clinic for treatment of his hepatitis C for definitive cure of his mixed cryoglobulinaemia, but failed to get insurance authorisation to begin treatment with Harvoni. He presented to the hospital 4 months later with diffuse alveolar haemorrhage.

Obturator Nerve Schwannoma as a Mimic of Ovarian Malignancy.

The obturator nerve is an extremely rare location for schwannomas to originate, and such diagnosis is typically not considered among the imaging diagnostic possibilities for a cystic-solid pelvic mass. A 63-year-old female with a known pelvic mass presented with increasing pelvic pain. The mass, which had been followed by serial imaging over five years, was described showing mixed solid and cystic components, likely arising from the left ovary. Although the key diagnosis to be excluded was a primary ovarian malignancy, the patient chose to pursue active surveillance. Over the five years of close observation, the lesion increased slowly, while her CA-125 level showed no significant elevation. Increase in size of the mass and worsening pain and concern for a gynecologic malignancy on MRI led her to ultimately consent to a hysterectomy with bilateral salpingooophorectomy. During the surgery, the mass was noted to be contiguous with the left obturator nerve. Pathologic evaluation revealed a schwannoma (WHO grade I). The patient's postsurgical course was uneventful, without residual weakness in the left adductor muscles.

Modifiers of ovarian function in girls and women with classic galactosemia.

CONTEXT: Classic galactosemia is a potentially lethal genetic disorder resulting from profound impairment of galactose-1P uridylyltransferase (GALT). More than 80% of girls and women with classic galactosemia experience primary or premature ovarian insufficiency despite neonatal diagnosis and rigorous lifelong dietary galactose restriction. OBJECTIVE: The goal of this study was to test the relationship between markers of ovarian reserve, cryptic residual GALT activity, and spontaneous pubertal development in girls with classic galactosemia. DESIGN AND SETTING: This was a cross-sectional study with some longitudinal follow-up in a university research environment. PATIENTS: Patients included girls and women with classic galactosemia and unaffected controls, <1 month to 30 years old. MAIN OUTCOME MEASURES: We evaluated plasma anti-Müllerian hormone (AMH) and FSH levels, antral follicle counts ascertained by ultrasound, and ovarian function as indicated by spontaneous vs assisted menarche. RESULTS: More than 73% of the pre- and postpubertal girls and women with classic galactosemia in this study, ages >3 months to 30 years, demonstrated AMH levels below the 95% confidence interval for AMH among controls of the same age, and both pre- and postpubertal girls and women with classic galactosemia also demonstrated abnormally low antral follicle counts relative to age-matched controls. Predicted residual GALT activity ≥ 0.4% significantly increased the likelihood that a girl with classic galactosemia would demonstrate an AMH level ≥ 0.1 ng/mL. CONCLUSIONS: A majority of girls with classic galactosemia demonstrate evidence of diminished ovarian reserve by 3 months of age, and predicted cryptic residual GALT activity is a modifier of ovarian function in galactosemic girls and women.

Misfolding of galactose 1-phosphate uridylyltransferase can result in type I galactosemia.

Type I galactosemia is a genetic disorder that is caused by the impairment of galactose-1-phosphate uridylyltransferase (GALT; EC 2.7.7.12). Although a large number of mutations have been detected through genetic screening of the human GALT (hGALT) locus, for many it is not known how they cause their effects. The majority of these mutations are missense, with predicted substitutions scattered throughout the enzyme structure and thus causing impairment by other means rather than direct alterations to the active site. To clarify the fundamental, molecular basis of hGALT impairment we studied five disease-associated variants p.D28Y, p.L74P, p.F171S, p.F194L and p.R333G using both a yeast model and purified, recombinant proteins. In a yeast expression system there was a correlation between lysate activity and the ability to rescue growth in the presence of galactose, except for p.R333G. Kinetic analysis of the purified proteins quantified each variant's level of enzymatic impairment and demonstrated that this was largely due to altered substrate binding. Increased surface hydrophobicity, altered thermal stability and changes in proteolytic sensitivity were also detected. Our results demonstrate that hGALT requires a level of flexibility to function optimally and that altered folding is the underlying reason of impairment in all the variants tested here. This indicates that misfolding is a common, molecular basis of hGALT deficiency and suggests the potential of pharmacological chaperones and proteostasis regulators as novel therapeutic approaches for type I galactosemia.

Cryptic residual GALT activity is a potential modifier of scholastic outcome in school age children with classic galactosemia.

Classic galactosemia is a potentially lethal disorder that results from profound deficiency of galactose-1-phosphate uridylyltransferase (GALT), the second enzyme in the Leloir pathway of galactose metabolism. Although early diagnosis and rigorous dietary restriction of galactose prevent or resolve the potentially lethal acute symptoms, patients are at markedly increased risk of long-term complications including significant cognitive, speech, and behavioral difficulties, among other problems. The mechanisms that underlie these long-term complications remain unclear, as do the factors that modify their severity. Here we explored the scholastic and behavioral outcomes experienced by a cohort of 54 school age children with classic galactosemia. Data collected included survey responses from parents and teachers, school records including standardized test scores, and GALT genotype data used to estimate predicted residual GALT activity based on a yeast expression system. As expected, many but not all of the children in our study demonstrated speech, scholastic, and behavioral difficulties. Perhaps most striking, we found that predicted cryptic residual GALT activity, often below the threshold of detection of clinical assays, appeared to modify scholastic outcome. These data raise the intriguing possibility that cryptic GALT activity might also influence the severity of other long-term complications in classic galactosemia.

N- and O-linked glycosylation of total plasma glycoproteins in galactosemia.

Classic galactosemia is a potentially lethal metabolic disorder that results from profound impairment of the enzyme galactose-1-phosphate uridylyltransferase (GALT); despite decades of research, the underlying mechanism of pathophysiology remains unclear. Previous studies of plasma and tissue samples from patients with classic galactosemia have revealed defects of protein and lipid glycosylation, however, the underlying bases for these defects and their clinical significance, if any, has remained unclear. As a step toward addressing these questions we characterized both the N- and O-linked glycomes of plasma proteins from neonates, infants, children, and adults with galactosemia using mass spectrometry and asked (1) whether similar or disparate defects exist for N-linked and O-linked modifications, (2) what factors correlate with the severity of these defects in different patients, and perhaps most important, (3) whether there is any apparent relationship between chronic glycosylation defects and long-term outcome in patients. We found that some but not all of the galactosemic neonates tested exhibited abnormal N- and O-linked glycosylation of plasma proteins. The types of abnormalities seen were similar between N- and O-linked moieties, but the extent of the defects varied between patients. Age, gender, GALT genotype, and predicted residual GALT activity all failed to explain the extent of the glycosylation defect in the samples studied. Dietary galactose restriction markedly normalized both the N- and O-linked glycosylation patterns for all infants tested; however, any remaining glycosylation defects evident in the plasma of older children or adults on galactose-restricted diets showed no correlation with clinical outcome. These data cannot rule out the possibility that subtle or localized glycosylation defects, not detectable by our methods or not reflected in plasma, may contribute to acute or long-term outcome severity.

Altered cofactor binding affects stability and activity of human UDP-galactose 4'-epimerase: implications for type III galactosemia.

Deficiency of UDP-galactose 4'-epimerase is implicated in type III galactosemia. Two variants, p.K161N-hGALE and p.D175N-hGALE, have been previously found in combination with other alleles in patients with a mild form of the disease. Both variants were studied in vivo and in vitro and showed different levels of impairment. p.K161N-hGALE was severely impaired with substantially reduced enzymatic activity, increased thermal stability, reduced cofactor binding and no ability to rescue the galactose-sensitivity of gal10-null yeast. Interestingly p.K161N-hGALE showed less impairment of activity with UDP-N-acetylgalactosamine in comparison to UDP-galactose. Differential scanning fluorimetry revealed that p.K161N-hGALE was more stable than the wild-type protein and only changed stability in the presence of UDP-N-acetylglucosamine and NAD(+). p.D175N-hGALE essentially rescued the galactose-sensitivity of gal10-null yeast, was less stable than the wild-type protein but showed increased stability in the presence of substrates and cofactor. We postulate that p.K161N-hGALE causes its effects by abolishing an important interaction between the protein and the cofactor, whereas p.D175N-hGALE is predicted to remove a stabilizing salt bridge between the ends of two α-helices that contain residues that interact with NAD(+). These results suggest that the cofactor binding is dynamic and that its loss results in significant structural changes that may be important in disease causation.

Ovarian function in Duarte galactosemia.

OBJECTIVE: To determine if girls with Duarte variant galactosemia (DG) have an increased risk of developing premature ovarian insufficiency based on prepubertal anti-Müllerian hormone (AMH) levels. DESIGN: Cross-sectional study. SETTING: University research laboratory. PATIENT(S): Study volunteers included 57 girls with DG, 89 girls with classic galactosemia (GG), and 64 control girls between the ages of <1 month and 10.5 years. INTERVENTION(S): Blood sampling. MAIN OUTCOME MEASURE(S): We determined AMH and FSH levels in study volunteers with and without Duarte variant or GG. RESULT(S): FSH levels were significantly higher and AMH levels significantly lower in girls with GG than in age-stratified control girls, but there was no significant difference between FSH and AMH levels in girls with DG and control girls. CONCLUSION(S): Although >80% of girls with GG in this study demonstrated low to undetectable AMH levels consistent with diminished ovarian reserve, 100% of girls with DG in our study demonstrated no apparent decrease in AMH levels or increase in FSH levels, suggesting that these girls are not at increased risk for premature ovarian insufficiency.