Neuroimaging in the Acute Psychiatric Setting: Associations With Neuropsychiatric Risk Factors.

OBJECTIVE: The appropriateness and clinical utility of neuroimaging in psychiatric populations has been long debated, and the ambiguity of guideline recommendations is well established. Most of the literature is focused on first-episode psychosis. The investigators aimed to review ordering practices and identify risk factors associated with neuroradiological MRI abnormalities and their clinical utility in a general psychiatric population. METHODS: A retrospective file review was undertaken for 100 consecutive brain MRI scans for adult psychiatric inpatients who received scanning as part of their clinical care in an Australian hospital. RESULTS: Brain MRI was abnormal in 79.0% of scans; in these cases, 72.2% of patients required further investigation or follow-up, with 17.7% requiring urgent referral within days to weeks, despite only 3.7% of admitted patients undergoing MRI during the study period. Psychiatrically relevant abnormalities were found in 32.0% of scans. Abnormalities were more likely to be found in the presence of cognitive impairment, older age, and longer duration of psychiatric disorder. Psychiatrically relevant abnormalities had further associations with older age at onset of the psychiatric disorder and a weak association with abnormal neurological examination. Multiple indications for imaging were present in 57.0% of patients; the most common indications were physical, neurological, and cognitive abnormalities. CONCLUSIONS: Brain MRI is a useful part of psychiatric management in the presence of certain neuropsychiatric risk factors. The present findings suggest that treating teams can judiciously tailor radiological investigations while limiting excessive imaging. Future research in larger cohorts across multiple centers may contribute to shaping more consistent neuroimaging guidelines in psychiatry.

Correlating MRI features with additional genetic markers and patient survival in histological grade 2-3 IDH-mutant astrocytomas.

PURPOSE: The increasing importance of molecular markers for classification and prognostication of diffuse gliomas has prompted the use of imaging features to predict genotype ("radiogenomics"). CDKN2A/B homozygous deletion has only recently been added to the diagnostic paradigm for IDH[isocitrate dehydrogenase]-mutant astrocytomas; thus, associated radiogenomic literature is sparse. There is also little data on whether different IDH mutations are associated with different imaging appearances. Furthermore, given that molecular status is now generally obtained routinely, the additional prognostic value of radiogenomic features is less clear. This study correlated MRI features with CDKN2A/B status, IDH mutation type and survival in histological grade 2-3 IDH-mutant brain astrocytomas. METHODS: Fifty-eight grade 2-3 IDH-mutant astrocytomas were identified, 50 with CDKN2A/B results. IDH mutations were stratified into IDH1-R132H and non-canonical mutations. Background and survival data were obtained. Two neuroradiologists independently assessed the following MRI features: T2-FLAIR mismatch (<25%, 25-50%, >50%), well-defined tumour margins, contrast-enhancement (absent, wispy, solid) and central necrosis. RESULTS: 8/50 tumours with CDKN2A/B results demonstrated homozygous deletion; slightly shorter survival was not significant (p=0.571). IDH1-R132H mutations were present in 50/58 (86%). No MRI features correlated with CDKN2A/B status or IDH mutation type. T2-FLAIR mismatch did not predict survival (p=0.977), but well-defined margins predicted longer survival (HR 0.36, p=0.008), while solid enhancement predicted shorter survival (HR 3.86, p=0.004). Both correlations remained significant on multivariate analysis. CONCLUSION: MRI features did not predict CDKN2A/B homozygous deletion, but provided additional positive and negative prognostic information which correlated more strongly with prognosis than CDKN2A/B status in our cohort.

Virtual conferences: results of an international survey on radiologist preferences and perspectives.

BACKGROUND: We explored perceptions and preferences regarding the conversion of in-person to virtual conferences as necessitated by travel and in-person meeting restrictions. METHODS: A 16-question online survey to assess preferences regarding virtual conferences during the COVID-19 pandemic and future perspectives on this subject was disseminated internationally online between June and August 2020. FINDINGS: A total of 508 responses were received from 73 countries. The largest number of responses came from Italy and the USA. The majority of respondents had already attended a virtual conference (80%) and would like to attend future virtual meetings (97%). The ideal duration of such an event was 2-3 days (42%). The preferred time format was a 2-4-h session (43%). Most respondents also noted that they would like a significant fee reduction and the possibility to attend a conference partly in-person and partly online. Respondents indicated educational sessions as the most valuable sections of virtual meetings. The reported positive factor of the virtual meeting format is the ability to re-watch lectures on demand. On the other hand, the absence of networking and human contact was recognized as a significant loss. In the future, people expressed a preference to attend conferences in person for networking purposes, but only in safer conditions. CONCLUSIONS: Respondents appreciated the opportunity to attend the main radiological congresses online and found it a good opportunity to stay updated without having to travel. However, in general, they would prefer these conferences to be structured differently. The lack of networking opportunities was the main reason for preferring an in-person meeting. KEY POINTS: • Respondents appreciated the opportunity to attend the main radiological meetings online, considering it a good opportunity to stay updated without having to travel. • In the future, it is likely for congresses to offer attendance options both in person and online, making them more accessible to a larger audience. • Respondents indicated that networking represents the most valuable advantage of in-person conferences compared to online ones.

Conventional MRI features can predict the molecular subtype of adult grade 2-3 intracranial diffuse gliomas.

PURPOSE: Molecular biomarkers are important for classifying intracranial gliomas, prompting research into correlating imaging with genotype ("radiogenomics"). A limitation of the existing radiogenomics literature is the paucity of studies specifically characterizing grade 2-3 gliomas into the three key molecular subtypes. Our study investigated the accuracy of multiple different conventional MRI features for genotype prediction. METHODS: Grade 2-3 gliomas diagnosed between 2007 and 2013 were identified. Two neuroradiologists independently assessed nine conventional MRI features. Features with better inter-observer agreement (κ ≥ 0.6) proceeded to consensus assessment. MRI features were correlated with genotype, classified as IDH-mutant and 1p/19q-codeleted (IDHmut/1p19qcodel), IDH-mutant and 1p/19q-intact (IDHmut/1p19qint), or IDH-wildtype (IDHwt). For IDHwt tumors, additional molecular markers of glioblastoma were noted. RESULTS: One hundred nineteen patients were included. T2-FLAIR mismatch (stratified as > 50%, 25-50%, or < 25%) was the most predictive feature across genotypes (p < 0.001). All 30 tumors with > 50% mismatch were IDHmut/1p19qint, and all seven with 25-50% mismatch. Well-defined margins correlated with IDHmut/1p19qint status on univariate analysis (p < 0.001), but this related to correlation with T2-FLAIR mismatch; there was no longer an association when considering only tumors with < 25% mismatch (p = 0.386). Enhancement (p = 0.001), necrosis (p = 0.002), and hemorrhage (p = 0.027) correlated with IDHwt status (especially "molecular glioblastoma"). Calcification correlated with IDHmut/1p19qcodel status (p = 0.003). A simple, step-wise algorithm incorporating these features, when present, correctly predicted genotype with a positive predictive value 91.8%. CONCLUSION: T2-FLAIR mismatch strongly predicts IDHmut/1p19qint even with a lower threshold of ≥ 25% mismatch and outweighs other features. Secondary features include enhancement, necrosis and hemorrhage (predicting IDHwt, especially "molecular glioblastoma"), and calcification (predicting IDHmut/1p19qcodel).

Bringing psychiatrists into the picture: Automated measurement of regional MRI brain volume in patients with suspected dementia.

OBJECTIVE: The volumes of various brain regions can be rapidly quantified using automated magnetic resonance imaging tools. While these appear to be useful at face value, their formal clinical utility is not yet understood, particularly for non-neuroradiologists and in patients presenting with suspected dementia. This study investigated the utility of an automated normative morphometry tool on determinations of brain atrophy by psychiatrists and radiologists in a tertiary hospital. METHODS: Consecutive magnetic resonance scans (n = 110) of patients referred with suspected neurodegenerative disorders were obtained retrospectively and rated by two neuroradiologists, two general radiologists and four psychiatrists over two sessions. First, conventional magnetic resonance sequences were shown. Then, morphometry colour-coded maps, which segmented T1-weighted magnetisation prepared rapid gradient echo images into brain regions and visualised these regions in colour according to their volumetric standard deviation from a normative population, were added to the second reading which occurred ⩾6 weeks later. Presence and laterality of atrophy in frontal, parietal and temporal lobes and hippocampal regions were measured using a digital checklist. The primary outcome of inter-rater agreement on atrophy was measured with Fleiss' Kappa (κ). We also evaluated the accuracy of the atrophy ratings for differentiating post hoc diagnosis of subjective cognitive impairment, mild cognitive impairment and dementia. RESULTS: Agreement among all raters was fair in frontal lobe and moderate in other regions with conventional method (κ = 0.362-0.555). With morphometry, higher agreement was seen in all regions (κ = 0.551-0.654), reaching significant improvement in the frontal and temporal lobes. No significant improvement was seen within the various disciplines, except in frontal lobes rated by psychiatrists. Accuracy of atrophy ratings on determining post hoc diagnosis was significantly improved for distinguishing subjective cognitive impairment versus dementia. CONCLUSION: In routine clinical assessment, automated normative morphometry complements the determination of regional atrophy and improves inter-rater agreement regardless of neuroradiology experience.

Correction to: Serum microRNA is a biomarker for post-operative monitoring in glioma.

For the reference citation '[57]' in the second paragraph of the Results section of the original article there was no corresponding entry in the References section. It should have referred to the below mentioned article by Ebrahimkhani et al. (2018).

Serum microRNA is a biomarker for post-operative monitoring in glioma.

PURPOSE: A circulating biomarker has potential to provide more accurate information for glioma progression post treatment, however no such biomarker is currently available. We aimed to discover a microRNA serum biomarker for longitudinal monitoring of glioma patients. METHODS: A prospectively collected cohort of 91 glioma patients and 17 healthy controls underwent pre and post-operative serum miRNA profiling using Nanostring®. Differentially expressed miRNAs were discovered using a machine learning random forest analysis. Candidate miRNAs were then assessed by droplet digital PCR in 11 patients with multiple follow up samples and compared to tumor volume based on magnetic resonance imaging. RESULTS: A 9-gene miRNA signature was identified that could distinguish between glioma and healthy controls with 99.8% accuracy. Two miRNAs miR-223 and miR-320e, best demonstrated dynamic changes that correlated closely with tumor volume in LGG and GBM respectively. Importantly, miRNA levels did not increase in two cases of pseudo-progression, indicating the potential utility of this test in guiding treatment decisions. CONCLUSIONS: We identified a highly accurate 9-miRNA signature associated with glioma serum. Additionally, we observed dynamic changes in specific miRNAs correlating with tumor volume over long-term follow up. These results support a large prospective validation study of serum miRNA biomarkers in glioma.

How to Create a Great Radiology Report.

The radiology report represents the sum of a radiologist's highest level of synthesis and insight into a patient's condition. It is the most important product that radiologists generate to help direct patient care. Despite the self-evident importance of clear and effective radiology reporting, radiologists usually receive little or no formal reporting education during training. Instead, it is learned in a piecemeal and often indirect fashion through occasional correction and imitating the reports of other radiologists. The audience of the radiology report extends far beyond the ordering provider and includes patients and their families, medical support staff, subspecialty providers, other radiologists, and research interests. Creating a report that fulfills the needs of this diverse group is a formidable if not quixotic ambition. However, there are certain key principles to reporting the imaging findings, impression, and recommendations that serve as a guide and promote careful consideration about how reports are understood. The findings section should emphasize short, informative, and factual observations while avoiding inappropriate interpretation, excessive use of terms of perception, and redundancy. The impression is the thoughtful synthesis of the meaning of the findings leading to a diagnosis, a differential diagnosis, and management recommendations. Creating a clear and impactful impression allows radiologists to provide the highest level of clinical care and direction but takes time and effort beyond simply restating the findings. The impression should use language that is understandable, memorable, and actionable. Reporting skills require ongoing attention and must adapt to the evolving practice patterns and communication styles in medicine. ©RSNA, 2020.

OCD symptoms in succinic semialdehyde dehydrogenase (SSADH) deficiency: a case report.

BACKGROUND: Succinic semialdehyde dehydrogenase (SSADH) deficiency is a rare neurometabolic disorder resulting in a heterogeneous clinical phenotype. Adolescent and adult patients with SSADH deficiency may present with OCD symptoms. There is minimal literature regarding the pathological basis of OCD symptoms and their management amongst SSADH deficiency patients. CASE PRESENTATION: A 26-year-old woman with SSADH deficiency experienced obsessional slowness and hesitancy in her activities of daily living, with motor rituals and stereotypies of her hands and face. Neuroimaging revealed T2 hyperintensities of the globi pallidi bilaterally. Commencement of the serotonergic escitalopram moderately improved her OCD symptoms. The addition of the dopaminergic pramipexole hydrochloride yielded further improvement, following unsuccessful trial of other adjuncts: risperidone, methylphenidate and mirtazapine. CONCLUSIONS: Pallidal pathology may explain the manifestation of OCD symptoms amongst individuals with SSADH deficiency. Serotonergic and concomitant dopaminergic therapy may be a viable treatment regimen for SSADH deficiency patients presenting with OCD symptoms.

Descriptive neuroradiology: beyond the hummingbird.

Radiology signs have long been described in ways that communicate the imagery around us to enhance our cognitive perception. Here, we describe the use and limitations of 10 such signs in neuroradiology, divided into three groups. The first are signs that are reliable for a specific diagnosis, such as the Medusa head sign indicating a developmental venous anomaly, or a racing car sign in agenesis of corpus callosum. The second group of signs helps us to diagnose rare conditions, such as the onion skin sign in Balo's concentric sclerosis. The third group is of unreliable signs that may lead clinicians astray. For example, the absence of a swallow-tail sign in Parkinson's disease or the presence of a hummingbird sign and Mickey Mouse sign in progressive supranuclear palsy. The appropriate use of these signs in clinical practice is essential.

Chest Radiographs in Congestive Heart Failure: Visualizing Neural Network Learning.

Purpose To examine Generative Visual Rationales (GVRs) as a tool for visualizing neural network learning of chest radiograph features in congestive heart failure (CHF). Materials and Methods A total of 103 489 frontal chest radiographs in 46 712 patients acquired from January 1, 2007, to December 31, 2016, were divided into a labeled data set (with B-type natriuretic peptide [BNP] result as a marker of CHF) and unlabeled data set (without BNP result). A generative model was trained on the unlabeled data set, and a neural network was trained on the encoded representations of the labeled data set to estimate BNP. The model was used to visualize how a radiograph with high estimated BNP would look without disease (a "healthy" radiograph). An overfitted model was developed for comparison, and 100 GVRs were blindly assessed by two experts for features of CHF. Area under the receiver operating characteristic curve (AUC), κ coefficient, and mixed-effects logistic regression were used for statistical analyses. Results At a cutoff BNP of 100 ng/L as a marker of CHF, the correctly trained model achieved an AUC of 0.82. Assessment of GVRs revealed that the correctly trained model highlighted conventional radiographic features of CHF as reasons for an elevated BNP prediction more frequently than the overfitted model, including cardiomegaly (153 [76.5%] of 200 vs 64 [32%] of 200, respectively; P < .001) and pleural effusions (47 [23.5%] of 200 vs 16 [8%] of 200, respectively; P = .003). Conclusion Features of congestive heart failure on chest radiographs learned by neural networks can be identified using Generative Visual Rationales, enabling detection of bias and overfitted models. © RSNA, 2018 See also the editorial by Ngo in this issue.

Understanding MRI in clinical psychiatry: perspectives from neuroimaging psychiatry registrars.

OBJECTIVE: Based on the experiences of neuroimaging psychiatry registrars, we describe several reflections on improving the understanding and integration of magnetic resonance imaging in clinical psychiatry. CONCLUSION: Better integration of magnetic resonance imaging into clinical psychiatry can be highly productive when our investments are focused on understanding the gaps in clinical knowledge and the systemic barriers involving the patient and relevant clinicians.

The incidence and significance of multicentric noncontrast-enhancing lesions distant from a histologically-proven glioblastoma.

Improvements in imaging are increasing the detection of multiple lesions in the setting of glioblastoma. Occasionally distant non-enhancing lesions may be identified which have the appearances of a multicentric low-grade glioma. We aimed to determine the incidence, prognostic significance and diagnostic value of this appearance in new glioblastoma patients. Pre-operative MRIs of patients with a new diagnosis of glioblastoma were reviewed to identify multicentric non-enhancing lesions, defined as areas of FLAIR hyperintensity and mass effect, without post-contrast enhancement, separate from the histologically-proven glioblastoma. Patient survival was compared to glioblastoma patients without these appearances, and follow-up imaging was reviewed. Nine of 151 patients (6 %) had multicentric non-enhancing lesions. Their median survival of 183 days was significantly worse than the 278 days for patients without multicentric nonenhancing lesions (p = 0.025). Follow-up MRIs were performed in four patients. In one patient, there were several additional lesions, one of which developed evidence of necrosis within 22 days of presentation. In the other three patients, the multicentric lesions developed enhancement and evidence of necrosis within 1 year, and became confluent on FLAIR with the dominant lesion. The appearance of a multicentric non-enhancing lesion is an uncommon finding in glioblastoma, but a poor prognostic feature. These lesions progress faster than expected for a low-grade glioma and are thus likely to represent more advanced lesions than their appearances suggest. Confluence with the dominant lesion developing with time suggests that the tumor is more extensive than appreciated on imaging.

Plasma ctDNA liquid biopsy of IDH1, TERTp, and EGFRvIII mutations in glioma.

Background: Circulating tumor DNA has emerging clinical applications in several cancers; however, previous studies have shown low sensitivity in glioma. We investigated if 3 key glioma gene mutations IDH1, TERTp, and EGFRvIII could be reliably detected in plasma by droplet digital polymerase chain reaction (ddPCR) thereby demonstrating the potential of this technique for glioma liquid biopsy. Methods: We analyzed 110 glioma patients from our biobank with a total of 359 plasma samples (median 4 samples per patient). DNA was isolated from plasma and analyzed for IDH1, TERTp, and EGFRvIII mutations using ddPCR. Results: Total cfDNA was significantly associated with tumor grade, tumor volume, and both overall and progression-free survival for all gliomas as well as the grade 4 glioblastoma subgroup, but was not reliably associated with changes in tumor volume/progression during the patients' postoperative time course. IDH1 mutation was detected with 84% overall sensitivity across all plasma samples and 77% in the preoperative samples alone; however, IDH1 mutation plasma levels were not associated with tumor progression or survival. IDH1m plasma levels were not associated with pre- or postsurgery progression or survival. The TERTp C228T mutation was detected in the plasma ctDNA in 88% but the C250T variant in only 49% of samples. The EGFRvIII mutation was detected in plasma in 5 out of 7 patients (71%) with tissue EGFRvIII mutations in tumor tissue. Conclusions: Plasma ctDNA mutations detected with ddPCR provide excellent diagnostic sensitivity for IDH1, TERTp-C228T, and EGFRvIII mutations in glioma patients. Total cfDNA may also assist with prognostic information. Further studies are needed to validate these findings and the clinical role of ctDNA in glioma.

Plasma ctDNA enables early detection of temozolomide resistance mutations in glioma.

Background: Liquid biopsy based on circulating tumor DNA (ctDNA) is a novel tool in clinical oncology, however, its use has been limited in glioma to date, due to low levels of ctDNA. In this study, we aimed to demonstrate that sequencing techniques optimized for liquid biopsy in glioma patients can detect ctDNA in plasma with high sensitivity and with potential clinical utility. Methods: We investigated 10 glioma patients with tumor tissue available from at least 2 surgical operations, who had 49 longitudinally collected plasma samples available for analysis. Plasma samples were sequenced with CAPP-seq (AVENIO) and tissue samples with TSO500. Results: Glioma-derived ctDNA mutations were detected in 93.8% of plasma samples. 25% of all mutations detected were observed in plasma only. Mutations of the mismatch repair (MMR) genes MSH2 and MSH6 were the most frequent circulating gene alterations seen after temozolomide treatment and were frequently observed to appear in plasma prior to their appearance in tumor tissue at the time of surgery for recurrence. Conclusions: This pilot study suggests that plasma ctDNA in glioma is feasible and may provide sensitive and complementary information to tissue biopsy. Furthermore, plasma ctDNA detection of new MMR gene mutations not present in the initial tissue biopsy may provide an early indication of the development of chemotherapy resistance. Additional clinical validation in larger cohorts is needed.

Radiogenomics of adult intracranial gliomas after the 2021 World Health Organisation classification: a review of changes, challenges and opportunities.

The classification of diffuse gliomas has undergone substantial changes over the last decade, starting with the 2016 World Health Organisation (WHO) classification, which introduced the importance of molecular markers for glioma diagnosis, in particular, isocitrate dehydrogenase (IDH) status and 1p/19-codeletion. This has spurred research into the correlation of imaging features with the key molecular markers, known as "radiogenomics" or "imaging genomics". Radiogenomics has a variety of possible benefits, including supplementing immunohistochemistry to refine the histological diagnosis and overcoming some of the limitations of the histological assessment. The recent 2021 WHO classification has introduced a variety of changes and continues the trend of increasing the importance of molecular markers in the diagnosis. Key changes include a formal distinction between adult- and paediatric-type diffuse gliomas, the addition of new diagnostic entities, refinements to the nomenclature for IDH-mutant (IDHmut) and IDH-wildtype (IDHwt) gliomas, a shift to grading within tumour types, and the addition of molecular markers as a determinant of tumour grade in addition to phenotype. These changes provide both challenges and opportunities for the field of radiogenomics, which are discussed in this review. This includes implications for the interpretation of research performed prior to the 2021 classification, based on the shift to first classifying gliomas based on genotype ahead of grade, as well as opportunities for future research and priorities for clinical integration.

Radiogenomics Provides Insights into Gliomas Demonstrating Single-Arm 1p or 19q Deletion.

BACKGROUND AND PURPOSE: IDH-mutant gliomas are further divided on the basis of 1p/19q status: oligodendroglioma, IDH-mutant and 1p/19q-codeleted, and astrocytoma, IDH-mutant (without codeletion). Occasionally, testing may reveal single-arm 1p or 19q deletion (unideletion), which remains within the diagnosis of astrocytoma. Molecular assessment has some limitations, however, raising the possibility that some unideleted tumors could actually be codeleted. This study assessed whether unideleted tumors had MR imaging features and survival more consistent with astrocytomas or oligodendrogliomas. MATERIALS AND METHODS: One hundred twenty-one IDH-mutant grade 2-3 gliomas with 1p/19q results were identified. Two neuroradiologists assessed the T2-FLAIR mismatch sign and calcifications, as differentiators of astrocytomas and oligodendrogliomas. MR imaging features and survival were compared among the unideleted tumors, codeleted tumors, and those without 1p or 19q deletion. RESULTS: The cohort comprised 65 tumors without 1p or 19q deletion, 12 unideleted tumors, and 44 codeleted. The proportion of unideleted tumors demonstrating the T2-FLAIR mismatch sign (33%) was similar to that in tumors without deletion (49%; P = .39), but significantly higher than codeleted tumors (0%; P = .001). Calcifications were less frequent in unideleted tumors (0%) than in codeleted tumors (25%), but this difference did not reach statistical significance (P = .097). The median survival of patients with unideleted tumors was 7.8 years, which was similar to that in tumors without deletion (8.5 years; P = .72) but significantly shorter than that in codeleted tumors (not reaching median survival after 12 years; P = .013). CONCLUSIONS: IDH-mutant gliomas with single-arm 1p or 19q deletion have MR imaging appearance and survival that are similar to those of astrocytomas without 1p or 19q deletion and significantly different from those of 1p/19q-codeleted oligodendrogliomas.

Evaluating the effectiveness of different perceptual training methods in a difficult visual discrimination task with ultrasound images.

Recent work has shown that perceptual training can be used to improve the performance of novices in real-world visual classification tasks with medical images, but it is unclear which perceptual training methods are the most effective, especially for difficult medical image discrimination tasks. We investigated several different perceptual training methods with medically naïve participants in a difficult radiology task: identifying the degree of hepatic steatosis (fatty infiltration of the liver) in liver ultrasound images. In Experiment 1a (N = 90), participants completed four sessions of standard perceptual training, and participants in Experiment 1b (N = 71) completed four sessions of comparison training. There was a significant post-training improvement for both types of training, although performance was better when the trained task aligned with the task participants were tested on. In both experiments, performance initially improves rapidly, with learning becoming more gradual after the first training session. In Experiment 2 (N = 200), we explored the hypothesis that performance could be improved by combining perceptual training with explicit annotated feedback presented in a stepwise fashion. Although participants improved in all training conditions, performance was similar regardless of whether participants were given annotations, or underwent training in a stepwise fashion, both, or neither. Overall, we found that perceptual training can rapidly improve performance on a difficult radiology task, albeit not to a comparable level as expert performance, and that similar levels of performance were achieved across the perceptual training paradigms we compared.

Impact of Different Artificial Intelligence User Interfaces on Lung Nodule and Mass Detection on Chest Radiographs.

Purpose: To explore the impact of different user interfaces (UIs) for artificial intelligence (AI) outputs on radiologist performance and user preference in detecting lung nodules and masses on chest radiographs. Materials and Methods: A retrospective paired-reader study with a 4-week washout period was used to evaluate three different AI UIs compared with no AI output. Ten radiologists (eight radiology attending physicians and two trainees) evaluated 140 chest radiographs (81 with histologically confirmed nodules and 59 confirmed as normal with CT), with either no AI or one of three UI outputs: (a) text-only, (b) combined AI confidence score and text, or (c) combined text, AI confidence score, and image overlay. Areas under the receiver operating characteristic curve were calculated to compare radiologist diagnostic performance with each UI with their diagnostic performance without AI. Radiologists reported their UI preference. Results: The area under the receiver operating characteristic curve improved when radiologists used the text-only output compared with no AI (0.87 vs 0.82; P < .001). There was no difference in performance for the combined text and AI confidence score output compared with no AI (0.77 vs 0.82; P = .46) and for the combined text, AI confidence score, and image overlay output compared with no AI (0.80 vs 0.82; P = .66). Eight of the 10 radiologists (80%) preferred the combined text, AI confidence score, and image overlay output over the other two interfaces. Conclusion: Text-only UI output significantly improved radiologist performance compared with no AI in the detection of lung nodules and masses on chest radiographs, but user preference did not correspond with user performance.Keywords: Artificial Intelligence, Chest Radiograph, Conventional Radiography, Lung Nodule, Mass Detection© RSNA, 2023.