Tracking and predicting COVID-19 radiological trajectory on chest X-rays using deep learning.

Radiological findings on chest X-ray (CXR) have shown to be essential for the proper management of COVID-19 patients as the maximum severity over the course of the disease is closely linked to the outcome. As such, evaluation of future severity from current CXR would be highly desirable. We trained a repurposed deep learning algorithm on the CheXnet open dataset (224,316 chest X-ray images of 65,240 unique patients) to extract features that mapped to radiological labels. We collected CXRs of COVID-19-positive patients from an open-source dataset (COVID-19 image data collection) and from a multi-institutional local ICU dataset. The data was grouped into pairs of sequential CXRs and were categorized into three categories: 'Worse', 'Stable', or 'Improved' on the basis of radiological evolution ascertained from images and reports. Classical machine-learning algorithms were trained on the deep learning extracted features to perform immediate severity evaluation and prediction of future radiological trajectory. Receiver operating characteristic analyses and Mann-Whitney tests were performed. Deep learning predictions between "Worse" and "Improved" outcome categories and for severity stratification were significantly different for three radiological signs and one diagnostic ('Consolidation', 'Lung Lesion', 'Pleural effusion' and 'Pneumonia'; all P < 0.05). Features from the first CXR of each pair could correctly predict the outcome category between 'Worse' and 'Improved' cases with a 0.81 (0.74-0.83 95% CI) AUC in the open-access dataset and with a 0.66 (0.67-0.64 95% CI) AUC in the ICU dataset. Features extracted from the CXR could predict disease severity with a 52.3% accuracy in a 4-way classification. Severity evaluation trained on the COVID-19 image data collection had good out-of-distribution generalization when testing on the local dataset, with 81.6% of intubated ICU patients being classified as critically ill, and the predicted severity was correlated with the clinical outcome with a 0.639 AUC. CXR deep learning features show promise for classifying disease severity and trajectory. Once validated in studies incorporating clinical data and with larger sample sizes, this information may be considered to inform triage decisions.

Deep learning of chest X-rays can predict mechanical ventilation outcome in ICU-admitted COVID-19 patients.

The COVID-19 pandemic repeatedly overwhelms healthcare systems capacity and forced the development and implementation of triage guidelines in ICU for scarce resources (e.g. mechanical ventilation). These guidelines were often based on known risk factors for COVID-19. It is proposed that image data, specifically bedside computed X-ray (CXR), provide additional predictive information on mortality following mechanical ventilation that can be incorporated in the guidelines. Deep transfer learning was used to extract convolutional features from a systematically collected, multi-institutional dataset of COVID-19 ICU patients. A model predicting outcome of mechanical ventilation (remission or mortality) was trained on the extracted features and compared to a model based on known, aggregated risk factors. The model reached a 0.702 area under the curve (95% CI 0.707-0.694) at predicting mechanical ventilation outcome from pre-intubation CXRs, higher than the risk factor model. Combining imaging data and risk factors increased model performance to 0.743 AUC (95% CI 0.746-0.732). Additionally, a post-hoc analysis showed an increase performance on high-quality than low-quality CXRs, suggesting that using only high-quality images would result in an even stronger model.

Is contralateral breast ultrasound worthwhile in preoperative staging of breast cancer?

BACKGROUND: Women with invasive breast cancer are at higher risk of contralateral synchronous cancer. This study aimed at determining if contralateral breast ultrasound (CBUS) examination should be routinely performed in the preoperative evaluation of breast cancer patients. METHODS: This is a retrospective study of preoperative CBUS examinations performed between January 2012 and April 2015. The charts of patients presenting for biopsy of a Breast Imaging Reporting and Data System (BIRADS) 5 lesion and who had undergone a concomitant contralateral breast US examination were reviewed. Index tumor, lymph node status, American College of Radiology (ACR) breast density on mammogram, total scanning time, and results of CBUS were recorded. RESULTS: Of the 3007 patients who underwent breast biopsies during the study period, 360 patients met the inclusion criteria. Index mass size was 19 ± 10 mm. CBUS examination led to 76 biopsies, of which 12 were positive in 11 patients. Detection rate for mammographically occult contralateral invasive cancers was 3.1% (11/360). Contralateral lesion size was 13 ± 10 mm. Breast density was rated ACR C/D for nine women and ACR B for two. In the ACR C/D subgroup (82%), the contralateral cancer detection rate was 4.1%. Average additional scanning time spent required to perform CBUS examination was 3.1 ± 4.9 min. Patients diagnosed with contralateral invasive breast cancer underwent surgery and/or chemotherapy. The treatment strategy was changed in all 11 patients after the detection of a second primary cancer. CONCLUSION: Preoperative CBUS is effective and most beneficial with women presenting ACR C/D breast density. Given its impact on decreasing future morbidity, its routine use should be considered to improve quality healthcare for women diagnosed with breast cancer.

Vacuum-assisted complete excision of solid intraductal/intracystic masses and complex cysts: Is follow-up necessary?

INTRODUCTION: Management of complex cysts and benign intraductal/intracystic masses is controversial. The aim of this study was to determine if the complete removal of the complex cyst lesions with ultrasound-guided vacuum-assisted excision (US-VAE) is sufficient for their safe management when the histological diagnosis obtained at biopsy is benign. SUBJECTS AND METHODS: This is a single institution retrospective study performed on patients who underwent breast biopsy between April 2007 and September 2013. Patients with complete removal of complex cyst lesion of a BIRADS 4 lesion by US-VAE that obtained a benign diagnosis were included. Size, morphology, histological diagnosis, and surgical or imaging follow-up of the lesions were analyzed. RESULTS: During the study period, 131 lesions met the inclusion criteria. Benign papilloma represented 32% (42/131) of the lesions; the remaining lesions had various benign diagnoses. Mean size of the solid mass or the cysts' thickest septum was 7 mm (range, 2-24). Mean imaging follow-up was 34.9 months (24-99 months) in 115 lesions. No recurrence or malignancy in the post-biopsy bed were observed during follow-up. Eleven lesions (8.4%) underwent surgery as follow-up: no cancer was found, but two lesions demonstrated atypia. CONCLUSIONS: Complex cyst lesion image completely excised with US-VAE and with a benign histology at biopsy might not require further imaging follow-up or surgery and a return to routine screening can be safely recommended. In a world where healthcare delivery and accessibility is important, elimination of unnecessary follow-ups is pertinent given its lower cost and lesser social impact.

Multicenter evaluation of a new ultrasound-guided biopsy device: Improved ergonomics, sampling and rebiopsy rates.

The purpose of this study was to evaluate performance, ergonomics, and immediate rebiopsy rate of a new vacuum-assisted biopsy (VAB) device for ultrasound-guided breast biopsies. Between December 2002 and April 2003, 113 patients meeting study criteria were biopsied at four centers using the new 9 gauge VAB device. The device has a radiofrequency-tipped probe, 360 degrees vacuum, a circumferential cutter, and a coaxial cannula for multiple sampling. Patient and procedural data included breast composition, lesion characteristics, number of samples, procedure time, and complications. Quality of samples, lesion access, and ergonomic features were assessed qualitatively and compared with prior experience with other biopsy devices. Immediate rebiopsy rate included high-risk lesions requiring surgical excision (obligate rebiopsy) and lesions requiring rebiopsy due to discordance or insufficient samples yielding nondiagnostic material. Data were analyzed using the Wilcoxon signed-rank test. One hundred thirteen patients aged 20-83 years (mean 52) were successfully biopsied with dense/fibrous breast tissue in 60% and dense/fibrous lesions in 49%. Lesions measured 6-63 mm (mean 17); 97% were masses. Five circumferential specimens (range 2-19) were obtained in 6 minutes (range 2-20). Operators rated safety and comfort comparable with existing devices and rated sample quality, breast/lesion penetration, and positioning ease/accuracy superior (p < 0.01). Diagnoses included 37 cancers, 70 benign, and six high-risk lesions with one upgrade from atypical ductal hyperplasia to ductal carcinoma in situ at surgery. Excluding obligate excision in high-risk diagnoses, the immediate rebiopsy rate was 2%. No complications required intervention. The new VAB device provides diagnostic samples and reduces sampling error defined by immediate rebiopsy rate. Compared with other devices, it is more ergonomic to target and position for sampling, particularly in dense breast tissue or lesions.

Redefining ultrasound appearance criteria of positive axillary lymph nodes.

OBJECTIVE: To determine objective criteria for ultrasonography (US) appearance in the diagnosis of abnormal axillary lymph nodes (ALNs) and to assess retrospectively their accuracy in preoperative staging of patients with Breast Imaging Reporting and Data System (BI-RADS) Category 5 breast lesion. METHODS: US-guided axillary fine-needle aspiration biopsy (FNAB) was performed for preoperative staging of BI-RADS Category 5 lesions in breast cancer patients seen at the Ottawa Regional Women's Breast Health Centre. Retrospectively, ALN morphology was classified into 3 categories based on US appearance: mass-like appearance, focal nodular cortical thickening (FNCT), and diffuse cortical thickening (DCT), the latter 2 with fatty hilum preservation and cortical thickening greater than 2 mm. Cytologic results from the FNAB were compared with axillary lymph node dissection (ALND) or sentinel lymph node procedure (SLNP), when available. Retrospective image-based classification was performed blind to pathology results and analyzed with Pearson's correlation coefficient, receiver operating characteristic (ROC) curves, and chi-square and Hosmer-Lemeshow tests against the cytologic ground truth. RESULTS: Eighty-six ALNs in 84 patients were sampled with an 18-gauge needle under US guidance. The number of passes per sample was related to adequate sampling of positive ALNs having mass-like morphology but not DCT or FNCT (p < 0.001). The most frequent morphologic presentation of the 86 ALNs was mass-like appearance of lymph nodes (45%), followed by DCT (35%) and FNCT (20%). Metastatic involvement was found in 34 (87%) of 39 ALNs that were mass-like, in 9 (53%) of 17 ALNs with FNCT, and in 9 (30%) of 30 ALNs with DCT. The kappa agreement between FNAB and ALND or SLNP was 0.64 for mass-like lymph nodes, 0.7 for lymph nodes with DCT, and 1.0 for FNCT. Overall sensitivity and specificity were 94% and 89%, respectively, with a positive predictive value of 97% and negative predictive value of 80%. CONCLUSIONS: Using a 2-mm threshold, FNCT and DCT warrant US-guided FNAB by an experienced radiologist as much as do mass-like lymph nodes. When biopsy is positive, axillary dissection can be performed and the sentinel node technique avoided.

Breast biopsy: multicenter study of radiofrequency introducer with US-guided handheld system--initial experience.

PURPOSE: To evaluate the utility of a radiofrequency (RF) introducer in placement and use of a handheld breast biopsy system during ultrasonography (US)-guided breast biopsy. MATERIALS AND METHODS: In three institutions from September 2000 through June 2001, 99 patients were scheduled for US-guided breast biopsy with the handheld system. Patients were randomly assigned to undergo biopsy with the handheld system alone (control group, 50 patients) or with the RF introducer and the handheld system (treatment group, 49 patients). Investigators compared the ability of the radiologist and pathologist to obtain a histologic diagnosis, ease of tissue penetration, procedure time, and pain. Patients assessed their pain level after the procedure and at 1, 7, and 30 days after biopsy. RESULTS: The RF introducer provided easier penetration (treatment group mean score, 1.6; control group, 2.4), particularly in patients with dense breasts (treatment group, 1.0; control group, 2.9) on the basis of a five-point scale (1 = very easy, 5 = extremely difficult). A histologic diagnosis was obtained in all cases, and no RF-induced artifact was reported. Average biopsy procedure time for treatment and control groups was not significantly different. Average patient pain levels were comparable between groups for each evaluation interval. Clinical complications in four treatment patients and three control patients were minor and resolved spontaneously. CONCLUSION: The RF introducer may serve as a useful adjunct to the handheld biopsy system by facilitating penetration of breast tissue without causing any deleterious effects for the patient or the biopsy specimen.