Using Recurrent Neural Networks to Extract High-Quality Information From Lung Cancer Screening Computerized Tomography Reports for Inter-Radiologist Audit and Feedback Quality Improvement.

PURPOSE: Lung cancer screening programs generate a high volume of low-dose computed tomography (LDCT) reports that contain valuable information, typically in a free-text format. High-performance named-entity recognition (NER) models can extract relevant information from these reports automatically for inter-radiologist quality control. METHODS: Using LDCT report data from a longitudinal lung cancer screening program (8,305 reports; 3,124 participants; 2006-2019), we trained a rule-based model and two bidirectional long short-term memory (Bi-LSTM) NER neural network models to detect clinically relevant information from LDCT reports. Model performance was tested using F1 scores and compared with a published open-source radiology NER model (Stanza) in an independent evaluation set of 150 reports. The top performing model was applied to a data set of 6,948 reports for an inter-radiologist quality control assessment. RESULTS: The best performing model, a Bi-LSTM NER recurrent neural network model, had an overall F1 score of 0.950, which outperformed Stanza (F1 score = 0.872) and a rule-based NER model (F1 score = 0.809). Recall (sensitivity) for the best Bi-LSTM model ranged from 0.916 to 0.991 for different entity types; precision (positive predictive value) ranged from 0.892 to 0.997. Test performance remained stable across time periods. There was an average of a 2.86-fold difference in the number of identified entities between the most and the least detailed radiologists. CONCLUSION: We built an open-source Bi-LSTM NER model that outperformed other open-source or rule-based radiology NER models. This model can efficiently extract clinically relevant information from lung cancer screening computerized tomography reports with high accuracy, enabling efficient audit and feedback to improve quality of patient care.

Clinical Impact of Cardiac MRI T1 and T2 Parametric Mapping in Patients with Suspected Cardiomyopathy.

Background There are limited data on the incremental value of parametric mapping compared with core cardiac MRI protocols for suspected cardiomyopathy in routine clinical practice. Purpose To evaluate the impact of cardiac MRI T1 and T2 mapping in routine clinical practice with respect to diagnostic accuracy, reader diagnostic confidence, and downstream cardiac imaging utilization. Materials and Methods In this retrospective single-center study, consecutive clinical cardiac MRI scans obtained with and without T1 and T2 mapping for evaluation of suspected cardiomyopathy between January 2017 and October 2019 were evaluated. Diagnostic accuracy and reader diagnostic confidence were evaluated in a random subset. Downstream cardiac imaging utilization was analyzed in patients with a minimum of 1 year of clinical follow-up ending before January 2020. Results A total of 1876 patients (mean age, 51 years ± 17 [SD]; 1113 men) were evaluated. Of these, 751 (40%) underwent cardiac MRI with the core protocol and 1125 (60%) with the core protocol plus T1 and T2 mapping. In the mapping group, T1 and T2 were high in 280 (25%) and 47 patients (4%), respectively. In the subset evaluated for diagnostic utility (n = 450), the addition of T1 and T2 maps to the core protocol resulted in an improvement in reader diagnostic confidence in 174 patients (39%). Diagnostic sensitivity was higher with the core protocol plus mapping compared with the core protocol alone for myocarditis (89% [31 of 35 patients] vs 69% [24 of 35]; P = .008), Fabry disease (93% [13 of 14 patients] vs 50% [seven of 14]; P = .01), and amyloidosis (100% [16 of 16 patients] vs 63% [10 of 16]; P = .01). In the subset evaluated for downstream imaging utilization (n = 903), 47% of patients with mapping had at least one subsequent cardiac imaging test compared with 55% of patients without mapping (P = .01). Conclusion In patients with suspected cardiomyopathy, cardiac MRI with T1 and T2 mapping had high diagnostic utility and was associated with lower downstream cardiac imaging utilization. © RSNA, 2022 Online supplemental material is available for this article. See also the editorial by Jerosch-Herold and Coelho-Filho in this issue.

Diagnostic Performance of Abnormal Nulling on Cardiac Magnetic Resonance Imaging Look Locker Inversion Time Sequence in Differentiating Cardiac Amyloidosis Types.

PURPOSE: To evaluate the diagnostic utility of the Look Locker inversion time (TI) sequence on cardiac magnetic resonance imaging in patients with suspected cardiac amyloidosis and to evaluate whether there are differences in the nulling pattern between amyloid types. MATERIALS AND METHODS: A total of 144 patients with suspected cardiac amyloidosis who had undergone cardiac magnetic resonance imaging were included in this retrospective study. Sixty-four had cardiac amyloidosis (62.1±9.2 y, 70.3% male, 68.8% had light chain amyloid [AL], 18.8% had familial transthyretin amyloid caused by mutant genes [ATTRm], and 12.5% had wild-type transthyretin amyloid [ATTRwt]) and 80 did not have cardiac amyloidosis (61.3±13.3 y, 58.8% male). Time to myocardial and blood pool nulling on the Look Locker TI sequence was classified as normal if blood pool nulled before myocardium or abnormal if blood pool nulling was coincident with or after myocardial nulling. RESULTS: The nulling pattern was abnormal in 26 patients with cardiac amyloidosis compared with none of the patients without cardiac amyloidosis (40.6% vs. 0.0%, P<0.0001). Abnormal nulling had 40.6% sensitivity and 100% specificity for cardiac amyloidosis (area under the receiver operating characteristic curve: 0.703, 95% confidence interval: 0.642-0.764). All patients with cardiac amyloidosis with an abnormal nulling pattern demonstrated late gadolinium enhancement. Among patients with cardiac amyloidosis, there was no significant difference in abnormal nulling between AL, ATTRm, and ATTRwt amyloid types (31.8%, 58.3%, 62.5%, respectively, P=0.10). CONCLUSIONS: An abnormal nulling pattern on the Look Locker TI sequence is highly specific for cardiac amyloidosis when present. However, abnormal nulling is a late finding with low sensitivity and does not differentiate between amyloid types.

New cardiac manifestation of IgG4-related disease: a case report.

(IgG4-RD) is a systemic fibro-inflammatory immune-mediated disease, which has been defined in the past few years. IgG4-RD affects various organs and leads to a variety of clinical manifestations. As it is a relatively newly defined entity, new manifestations are now being recognised and reported. We describe a case involving the cardiovascular system.

CT-guided Microcoil Pulmonary Nodule Localization prior to Video-assisted Thoracoscopic Surgery: Diagnostic Utility and Recurrence-Free Survival.

Background CT-guided microcoil localization has been shown to reduce the need for thoracotomy or video-assisted thoracoscopic surgery (VATS) anatomic resection. However, only short-term follow-up after CT-guided microcoil localization and lung resection has been previously reported. Purpose To assess the diagnostic utility and recurrence-free survival over a minimum of 2 years following CT-guided microcoil localization and VATS. Materials and Methods Among 1950 VATS procedures performed in a single tertiary institution from October 2008 through April 2016, 124 consecutive patients with CT-guided microcoil localization were retrospectively evaluated. Patient demographics, nodule characteristics, and histopathologic findings were recorded. The primary end point was recurrence-free survival after 2 or more years of CT surveillance. Statistical analysis included Kaplan-Meier survival curves and Cox regression. Results In 124 patients (men, 35%; mean age, 65 years ± 12) with a nodule found at CT, microcoil localization and VATS resection were performed for a total of 126 nodules (mean size, 13 mm ± 6; mean distance to pleura, 20 mm ± 9). On presurgical CT evaluation, 42% (53 of 126) of nodules were solid, 33% (41 of 126) were ground glass, and 24% (30 of 126) were subsolid. VATS excisional biopsy altered cytopathologic diagnosis in 21% (five of 24) of patients with prior diagnostic premicrocoil CT-guided biopsy. At histopathologic examination, 17% (21 of 126) of the nodules were adenocarcinoma in situ, 17% (22 of 126) were minimally invasive adenocarcinoma, 30% (38 of 126) were invasive lung primary tumors, and 22% (28 of 126) were metastases. Among the 72 patients with malignancy at histopathologic examination and at least 2 years of CT surveillance, local recurrence occurred in 7% (five of 72), intrathoracic recurrence in 22% (16 of 72), and extrathoracic recurrence in 18% (13 of 72) after 2 or more years of CT surveillance. There was no recurrence for adenocarcinoma in situ, minimally invasive adenocarcinoma, or invasive lung tumors measuring less than 1 cm. After multivariable adjustment, nodule location at a distance greater than 10 mm from the pleura was an independent predictor of time to recurrence (hazard ratio, 2.9 [95% confidence interval: 1.1, 7.4]; P = .03). Conclusion CT-guided microcoil localization and video-assisted thoracoscopic surgical resection alter clinical management and were associated with excellent recurrence-free survival for superficial premalignant, minimally invasive, and small invasive lung tumors. © RSNA, 2019 Online supplemental material is available for this article.

Prevalence of obstructive coronary artery disease and prognosis in patients with stable symptoms and a zero-coronary calcium score.

AIMS: CT calcium scoring (CTCS) and CT cardiac angiography (CTCA) are widely used in patients with stable chest pain to exclude significant coronary artery disease (CAD). We aimed to resolve uncertainty about the prevalence of obstructive coronary artery disease and long-term outcomes in patients with a zero-calcium score (ZCS). METHODS AND RESULTS: Consecutive patients with stable cardiac symptoms referred for CTCS or CTCS and CTCA from chest pain clinics to a tertiary cardiothoracic centre were prospectively enrolled. In those with a ZCS, the prevalence of obstructive CAD on CTCA was determined. A follow-up for all-cause mortality was obtained from the NHS tracer service. A total of 3914 patients underwent CTCS of whom 2730 (69.7%) also had a CTCA. Half of the patients were men (50.3%) with a mean age of 56.9 years. Among patients who had both procedures, a ZCS was present in 52.2%, with a negative predictive value of 99.5% for excluding ≥70% stenosis on CTCA. During a mean follow-up of 5.2 years, the annual event rate was 0.3% for those with ZCS compared with 1.2% for CS ≥1. The presence of non-calcified atheroma on CTCA in patients with ZCS did not affect the prognostic value (P = 0.98). CONCLUSION: In patients with stable symptoms and a ZCS, obstructive CAD is rare, and prognosis over the long-term is excellent, regardless of whether non-calcified atheroma is identified. A ZCS could reliably be used as a 'gatekeeper' in this patient cohort, obviating the need for further more expensive tests.

Postoperative imaging after lung transplantation.

Lung transplantation (LT) is an established procedure for chronic end-stage lung diseases. Complications are frequent and diverse and are the consequence of the complex surgical technique, the severity of the initial pathology, and the deep state of posttransplantation immunosuppression. Complications following LT include primary graft dysfunction, rejection (hyperacute, acute, and chronic), infections, posttransplantation lymphoproliferative disease, pleural and airway complications, native lung complications, and recurrence of primary disease. An understanding of these complications, their temporal evolution, and the role of radiology and other diagnostic methods in their diagnosis and management will help reduce the morbidity and mortality associated with LT.