Diagnostic histopathology is a gold standard for diagnosing hematopoietic malignancies. Pathologic diagnosis requires labor-intensive reading of a large number of tissue slides with high diagnostic accuracy equal or close to 100 percent to guide treatment options, but this requirement is difficult to meet. Although artificial intelligence (AI) helps to reduce the labor of reading pathologic slides, diagnostic accuracy has not reached a clinically usable level. Establishment of an AI model often demands big datasets and an ability to handle large variations in sample preparation and image collection. Here, we establish a highly accurate deep learning platform, consisting of multiple convolutional neural networks, to classify pathologic images by using smaller datasets. We analyze human diffuse large B-cell lymphoma (DLBCL) and non-DLBCL pathologic images from three hospitals separately using AI models, and obtain a diagnostic rate of close to 100 percent (100% for hospital A, 99.71% for hospital B and 100% for hospital C). The technical variability introduced by slide preparation and image collection reduces AI model performance in cross-hospital tests, but the 100% diagnostic accuracy is maintained after its elimination. It is now clinically practical to utilize deep learning models for diagnosis of DLBCL and ultimately other human hematopoietic malignancies.
Severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2) has spread rapidly throughout the whole world and caused significant difficulties in the prevention and control of the epidemic. In this case, several detection methods have been established based on nucleic acid diagnostic techniques and immunoassays to achieve sensitive and specific detection of SARS‐CoV‐2. However, most methods are still largely dependent on professional instruments, highly trained operators, and centralized laboratories. These limitations gravely diminish their practicality and portability. Herein, a clustered regularly interspaced short palindromic repeats (CRISPR) Cas12a based assay was developed for portable, rapid and sensitive of SARS‐CoV‐2. In this assay, samples were quickly pretreated and amplified by reverse transcription recombinase‐aided amplification under mild conditions. Then, by combining the CRISPR Cas12a system and a glucose‐producing reaction, the signal of the virus was converted to that of glucose, which can be quantitatively read by a personal glucose meter in a few seconds. Nucleocapsid protein gene was tested as a model target, and the sensitivity for quantitative detection was as low as 10 copies/μl, which basically meet the needs of clinical diagnosis. In addition, with the advantages of lower material cost, shorter detection time, and no requirement for professional instrument in comparison with quantitative reverse transcription‐polymerase chain reaction, this assay is expected to provide a powerful technical support for the early diagnosis and intervention during epidemic prevention and control.
ObjectiveTo develop and prospective validate an ultrasound (US) prediction model to differentiate between benign and malignant subpleural pulmonary lesions (SPLs).MethodsThis study was conducted retrospectively from July 2017 to December 2018 (development cohort [DC], n = 592) and prospectively from January to April 2019 (validation cohort [VC], n = 220). A total of 18 parameters of B-mode US and contrast-enhanced US (CEUS) were acquired. Based on the DC, a model was developed using binary logistic regression. Then its discrimination and calibration were verified internally in the DC and externally in the VC, and its diagnostic performance was compared with those of the existing US diagnostic criteria in the two cohorts. The reference criteria were from the comprehensive diagnosis of clinical-radiological-pathological made by two senior respiratory physicians.ResultsThe model was eventually constructed with 6 parameters: the angle between lesion border and thoracic wall, basic intensity, lung-lesion arrival time difference, ratio of arrival time difference, vascular sign, and non-enhancing region type. In both internal and external validation, the model provided excellent discrimination of benign and malignant SPLs (C-statistic: 0.974 and 0.980 respectively), which is higher than that of “lesion-lung AT difference ≥ 2.5 s” (C-statistic: 0.842 and 0.777 respectively, P <0.001) and “AT ≥ 10 s” (C-statistic: 0.688 and 0.641 respectively, P <0.001) and the calibration curves of the model showed good agreement between actual and predictive malignancy probabilities. As for the diagnosis performance, the sensitivity and specificity of the model [sensitivity: 94.82% (DC) and 92.86% (VC); specificity: 92.42% (DC) and 92.59% (VC)] were higher than those of “lesion-lung AT difference ≥ 2.5 s” [sensitivity: 88.11% (DC) and 80.36% (VC); specificity: 80.30% (DC) and 75.00% (VC)] and “AT ≥ 10 s” [sensitivity: 64.94% (DC) and 61.61% (VC); specificity: 72.73% (DC) and 66.67% (VC)].ConclusionThe prediction model integrating multiple parameters of B-mode US and CEUS can accurately predict the malignancy probability, so as to effectively differentiate between benign and malignant SPLs, and has better diagnostic performance than the existing US diagnostic criteria.Clinical Trial Registrationwww.chictr.org.cn, identifier ChiCTR1800019828.
The toxicity of doxorubicin (DOX), especially in terms of cardiotoxicity, has been a common problem in its clinical use.
To evaluate the application of machine learning for the detection of subpleural pulmonary lesions (SPLs) in ultrasound (US) scans, we propose a novel boundary-restored network (BRN) for automated SPL segmentation to avoid issues associated with manual SPL segmentation (subjectivity, manual segmentation errors, and high time consumption). In total, 1612 ultrasound slices from 255 patients in which SPLs were visually present were exported. The segmentation performance of the neural network based on the Dice similarity coefficient (DSC), Matthews correlation coefficient (MCC), Jaccard similarity metric (Jaccard), Average Symmetric Surface Distance (ASSD), and Maximum symmetric surface distance (MSSD) was assessed. Our dual-stage boundary-restored network (BRN) outperformed existing segmentation methods (U-Net and a fully convolutional network (FCN)) for the segmentation accuracy parameters including DSC (83.45 ± 16.60%), MCC (0.8330 ± 0.1626), Jaccard (0.7391 ± 0.1770), ASSD (5.68 ± 2.70 mm), and MSSD (15.61 ± 6.07 mm). It also outperformed the original BRN in terms of the DSC by almost 5%. Our results suggest that deep learning algorithms aid fully automated SPL segmentation in patients with SPLs. Further improvement of this technology might improve the specificity of lung cancer screening efforts and could lead to new applications of lung US imaging.
U S has been used not only to guide percutaneous lung biopsy but also to provide diagnostic evidence for differential diagnosis of subpleural pulmonary lesions (SPLs) for decades (1,2). We define subpleural lesions as those that touch or are very near the visceral pleural surface but are not in the pleural space (3-7). However, B-mode US has limited diagnostic value because benign and malignant SPLs have similar echo texture, shape, and outer margins (2).Contrast-enhanced (CE) US is a simple and safe ultrasonic technique capable of microcirculation visualization, and it has been used to improve the differential diagnosis of . It can display the perfusion pattern, intensity, time, and necrotic area of the lesion, among which the parameter time is believed to be the most potent indicator (6,7). Studies have indicated that malignant tumor tissues often (56%-87% of the time) invade the pulmonary artery of the affected lung segment, leading to pulmonary artery stenosis or occlusion, and the blood supply to the ischemic lung tissue is supplemented by the bronchial arteries (13). This abnormal change could be identified by the CE US time indicators (6,7).After intravenous injection, the US contrast agent arrives in the right side of the heart first, then goes into the pulmonary artery. After pulmonary circulation, the contrast agent enters the left side of the heart, then is pumped into the bronchial artery (Fig E1 [online]) ( 14). Therefore, the pulmonary artery enhances earlier than the bronchial artery (4,5). The arrival time (AT) is the time taken for US contrast agent to arrive at the target area after injection. It varies depending on the fraction of blood from the pulmonary artery and the bronchial artery, enabling the distinction of benign and malignant SPLs Background: US has proven valuable in the diagnosis of subpleural pulmonary lesions (SPLs); however, existing US indicators have limitations.Purpose: To propose and validate a revised contrast-enhanced (CE) US indicator for differential diagnosis of benign and malignant SPLs and to compare its performance with existing CE US diagnostic criteria. Materials and Methods:This prospective study (Chinese clinical trial registry, ChiCTR1800019828) enrolled patients with SPLs between May 2019 and August 2020. They were divided into a developmental cohort (DC) and a validation cohort (VC). In the DC, the optimal indicator was selected from five CE US indicators. In the VC, the selected indicator was compared with existing CE US diagnostic criteria using the area under the receiver operating characteristic curve (AUC). Pathologic analysis, microbial evidence, and clinical follow-up were used as reference standards for all SPLs.Results: A total of 902 participants (DC, 424 participants; VC, 478 participants) with SPLs (mean age, 56 years 17; 593 men) were evaluated. The arrival time (AT) difference ratio proved to be the optimal indicator to distinguish benign from malignant SPLs. In the overall (regardless of lesion size), large (vertical diameter 3 cm), and small ...
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.