Using the Textual Content of Radiological Reports to Detect Emerging Diseases: A Proof-of-Concept Study of COVID-19.

Changes in the content of radiological reports at population level could detect emerging diseases. Herein, we developed a method to quantify similarities in consecutive temporal groupings of radiological reports using natural language processing, and we investigated whether appearance of dissimilarities between consecutive periods correlated with the beginning of the COVID-19 pandemic in France. CT reports from 67,368 consecutive adults across 62 emergency departments throughout France between October 2019 and March 2020 were collected. Reports were vectorized using time frequency-inverse document frequency (TF-IDF) analysis on one-grams. For each successive 2-week period, we performed unsupervised clustering of the reports based on TF-IDF values and partition-around-medoids. Next, we assessed the similarities between this clustering and a clustering from two weeks before according to the average adjusted Rand index (AARI). Statistical analyses included (1) cross-correlation functions (CCFs) with the number of positive SARS-CoV-2 tests and advanced sanitary index for flu syndromes (ASI-flu, from open-source dataset), and (2) linear regressions of time series at different lags to understand the variations of AARI over time. Overall, 13,235 chest CT reports were analyzed. AARI was correlated with ASI-flu at lag = + 1, + 5, and + 6 weeks (P = 0.0454, 0.0121, and 0.0042, respectively) and with SARS-CoV-2 positive tests at lag = - 1 and 0 week (P = 0.0057 and 0.0001, respectively). In the best fit, AARI correlated with the ASI-flu with a lag of 2 weeks (P = 0.0026), SARS-CoV-2-positive tests in the same week (P < 0.0001) and their interaction (P < 0.0001) (adjusted R2 = 0.921). Thus, our method enables the automatic monitoring of changes in radiological reports and could help capturing disease emergence.

What Influences the Way Radiologists Express Themselves in Their Reports? A Quantitative Assessment Using Natural Language Processing.

Although using standardized reports is encouraged, most emergency radiological reports in France remain in free-text format that can be mined with natural language processing for epidemiological purposes, activity monitoring or data collection. These reports are obtained under various on-call conditions by radiologists with various backgrounds. Our aim was to investigate what influences the radiologists' written expressions. To do so, this retrospective multicentric study included 30,227 emergency radiological reports of computed tomography scans and magnetic resonance imaging involving exactly one body region, only with pathological findings, interpreted from 2019-09-01 to 2020-02-28 by 165 radiologists. After text pre-processing, one-word tokenization and use of dictionaries for stop words, polarity, sentiment and uncertainty, 11 variables depicting the structure and content of words and sentences in the reports were extracted and summarized to 3 principal components capturing 93.7% of the dataset variance. In multivariate analysis, the 1st principal component summarized the length and lexical diversity of the reports and was significantly influenced by the weekday, time slot, workload, number of examinations previously interpreted by the radiologist during the on-call period, type of examination, emergency level and radiologists' gender (P value range: < 0.0001-0.0029). The 2nd principal component summarized negative formulations, polarity and sentence length and was correlated with the number of examination previously interpreted by the radiologist, type of examination, emergency level, imaging modality and radiologists' experience (P value range: < 0.0001-0.0032). The last principal component summarized questioning, uncertainty and polarity and was correlated with the type of examination and emergency level (all P values < 0.0001). Thus, the length, structure and content of emergency radiological reports were significantly influenced by organizational, radiologist- and examination-related characteristics, highlighting the subjectivity and variability in the way radiologists express themselves during their clinical activity. These findings advocate for more homogeneous practices in radiological reporting and stress the need to consider these influential features when developing models based on natural language processing.

Emergency teleradiological activity is an epidemiological estimator and predictor of the covid-19 pandemic in mainland France.

BACKGROUND: COVID-19 pandemic highlighted the need for real-time monitoring of diseases evolution to rapidly adapt restrictive measures. This prospective multicentric study aimed at investigating radiological markers of COVID-19-related emergency activity as global estimators of pandemic evolution in France. We incorporated two sources of data from March to November 2020: an open-source epidemiological dataset, collecting daily hospitalisations, intensive care unit admissions, hospital deaths and discharges, and a teleradiology dataset corresponding to the weekly number of CT-scans performed in 65 emergency centres and interpreted remotely. CT-scans specifically requested for COVID-19 suspicion were monitored. Teleradiological and epidemiological time series were aligned. Their relationships were estimated through a cross-correlation function, and their extremes and breakpoints were compared. Dynamic linear models were trained to forecast the weekly hospitalisations based on teleradiological activity predictors. RESULTS: A total of 100,018 CT-scans were included over 36 weeks, and 19,133 (19%) performed within the COVID-19 workflow. Concomitantly, 227,677 hospitalisations were reported. Teleradiological and epidemiological time series were almost perfectly superimposed (cross-correlation coefficients at lag 0: 0.90-0.92). Maximal number of COVID-19 CT-scans was reached the week of 2020-03-23 (1 086 CT-scans), 1 week before the highest hospitalisations (23,542 patients). The best valid forecasting model combined the number of COVID-19 CT-scans and the number of hospitalisations during the prior two weeks and provided the lowest mean absolute percentage (5.09%, testing period: 2020-11-02 to 2020-11-29). CONCLUSION: Monitoring COVID-19 CT-scan activity in emergencies accurately and instantly predicts hospitalisations and helps adjust medical resources, paving the way for complementary public health indicators.

Evaluation of ischemic lesion prevalence after endovascular treatment of intracranial aneurysms, as documented by 3-T diffusion-weighted imaging: a 2-year, single-center cohort study.

OBJECTIVE Even though published data exist concerning the prevalence of ischemic lesions detected by diffusion-weighted imaging (DWI) following endovascular treatment of intracranial aneurysms, a single-center cross-evaluation of the different endovascular techniques has been lacking. The authors sought to prospectively evaluate the prevalence and clinical significance of ischemic lesions occurring after endovascular treatment of intracranial aneurysms and to compare the safety and effectiveness of a broad spectrum of currently accepted endovascular techniques in a single-center setting. METHODS This was a prospective cohort study involving consecutive patients treated for intracranial aneurysms exclusively by endovascular means, excluding treatments in the acute rupture phase, in a center featuring an endovascular-only treatment policy for intracranial aneurysms. All patients underwent MRI, including a 3-directional DWI sequence, before treatment, 24 hours postprocedure, and 6 months following endovascular embolization. Selective angiography was performed at 6 months' follow-up. RESULTS From January 2012 through December 2013, 164 aneurysms were treated in 128 consecutive patients. Endovascular techniques included coiling (14.6%), balloon-assisted coiling (20.1%), stent-assisted coiling (3.7%), low-profile stent-assisted coiling, flow diversion (38.4%), and very complex treatments (6.1%) involving 2 stents in Y or T configurations. On postprocedure MRI, the rates of occurrence of new DWI-positive lesions were 64.3% for coiling, 54.5% for remodeling, 61.1% for stent-assisted coiling, 53.7% for flow-diverting stents, and 75% for very complex treatments (p = 0.4962). The 6-month procedure-related morbidity and mortality rates were 6.25% and 0%, respectively. At 6 months' follow-up, 93% of the patients had modified Rankin Scale (mRS) scores of 0-2. Very complex treatments offered a higher complete occlusion rate (100%) than all other techniques (66.7%-88.9%). Age and length of procedure were independent factors for DWI lesion occurrence. The diameter of DWI lesions on 24-hour postprocedure MRI was positively correlated with mRS score at discharge. Among the DWI-positive lesions measuring less than 2 mm in diameter on the 24-hour MRI, 44.12% had regressed at 6 months. CONCLUSIONS Procedure-related DWI lesions are far more often encountered in silent forms than they are clinically evident. They do not seem to be significantly correlated with procedure-related complications, nor do they seem to impair clinical outcome, regardless of the endovascular technique. Small lesions (< 2 mm in diameter) may regress within 6 months. The use of the most adapted technique, in terms of aneurysm configuration, results in significant total occlusion rates, with acceptable safety.