scholarly journals Survival prediction in mesothelioma using a scalable Lasso regression model: instructions for use and initial performance using clinical predictors

2018 ◽  
Vol 5 (1) ◽  
pp. e000240 ◽  
Author(s):  
Andrew C Kidd ◽  
Michael McGettrick ◽  
Selina Tsim ◽  
Daniel L Halligan ◽  
Max Bylesjo ◽  
...  
Keyword(s):  
Prognostic Value ◽  
Computational Models ◽  
White Cell Count ◽  
Performance Status ◽  
Area Under The Curve ◽  
Clinical Information ◽  
Prognostic Models ◽  
Survival Prediction ◽  
Lasso Regression ◽  
Clinical Predictors

IntroductionAccurate prognostication is difficult in malignant pleural mesothelioma (MPM). We developed a set of robust computational models to quantify the prognostic value of routinely available clinical data, which form the basis of published MPM prognostic models.MethodsData regarding 269 patients with MPM were allocated to balanced training (n=169) and validation sets (n=100). Prognostic signatures (minimal length best performing multivariate trained models) were generated by least absolute shrinkage and selection operator regression for overall survival (OS), OS <6 months and OS <12 months. OS prediction was quantified using Somers DXY statistic, which varies from 0 to 1, with increasing concordance between observed and predicted outcomes. 6-month survival and 12-month survival were described by area under the curve (AUC) scores.ResultsMedian OS was 270 (IQR 140–450) days. The primary OS model assigned high weights to four predictors: age, performance status, white cell count and serum albumin, and after cross-validation performed significantly better than would be expected by chance (mean DXY0.332 (±0.019)). However, validation set DXY was only 0.221 (0.0935–0.346), equating to a 22% improvement in survival prediction than would be expected by chance. The 6-month and 12-month OS signatures included the same four predictors, in addition to epithelioid histology plus platelets and epithelioid histology plus C-reactive protein (mean AUC 0.758 (±0.022) and 0.737 (±0.012), respectively). The <6-month OS model demonstrated 74% sensitivity and 68% specificity. The <12-month OS model demonstrated 63% sensitivity and 79% specificity. Model content and performance were generally comparable with previous studies.ConclusionsThe prognostic value of the basic clinical information contained in these, and previously published models, is fundamentally of limited value in accurately predicting MPM prognosis. The methods described are suitable for expansion using emerging predictors, including tumour genomics and volumetric staging.

Survival prediction in terminally ill cancer patients by clinical estimates, laboratory and psychometric tests

2006 ◽  
Vol 24 (18_suppl) ◽  
pp. 8578-8578
Author(s):  
S. Gripp ◽  
S. Moeller ◽  
R. Willers
Keyword(s):  
Cancer Patients ◽  
Emotional Disorders ◽  
Advanced Cancer ◽  
Performance Status ◽  
Prognostic Models ◽  
Anxiety And Depression ◽  
Survival Prediction ◽  
Psychometric Tests ◽  
Life Table Analysis ◽  
Table Analysis

8578 Background: To improve general survival estimates in advanced cancer pat. we studied physicians’ clinical estimates, the impact of emotional disorders (anxiety and depression), and laboratory tests in palliative patients. Methods: From 12/03 to 7/04 patients with advanced cancer referred to radiation oncology for palliative treatment were invited to participate in this prospective cohort study. Pat. with adjuvant or curative treatment intent were not considered. The life span was independently estimated by two physicians and the institutional tumor board according to 3 categories (<1, 1–6, and >6 months). Agreement of survival predictions was analyzed with contingency tables and kappa statistics. Primary tumor, metastatic spread, performance status, pain, dyspnoea, weight loss, nausea, fatigue, serum enzymes (AP, LDH), function parameters (creatinine, bilirubin, CRP), and blood count (WBC, RBC) were also studied. Emotional disorders were measured using a validated psychometric self-assessment scale (Hospital Anxiety and Depression Scale, HADS). Life table analysis with log-rank test and stepwise Cox regression analysis with univariate significant variables were performed. Results: 216 pat. were enrolled and followed for at least 6 months. 580 prognoses were obtained. 94% (204) had complete blood tests. HADS questionnaires were completed by 71% (154). Survival was <1 mo in 15% (33), 1–6 mo in 36% (78), and >6 mo in 49% (105).Survival prediction was poor (kappa= 0.33) and consistently too optimistic (test of symmetry, p<0.0001). In life table analysis primary tumor (hazard ratio 2.0), brain metastases, performance status (HR 1.9), dyspnoea (HR 2.0), nausea (HR 2.0), LDH (HR 1.9), WBC (HR 2.1), fatigue, anxiety and depression (HADS) were highly significant (p< 0.0002). Conclusions: Physicians generally overestimated survival of advanced cancer patients emphasizing the need of objective prognostic models. Even short-term survival estimates (< 1 mo.) were unreliable. Combined objective variables may improve survival prediction. Psychometric tests are promising candidates to be incorporated in more accurate prognostic models. No significant financial relationships to disclose.

Stage-specific prognostic value of molecular markers in colon cancer: Results of the translational study on the PETACC 3-EORTC 40993-SAKK 60–00 trial

2009 ◽  
Vol 27 (15_suppl) ◽  
pp. 4002-4002 ◽  
Author(s):  
A. D. Roth ◽  
S. Tejpar ◽  
P. Yan ◽  
R. Fiocca ◽  
D. Dietrich ◽  
...  
Keyword(s):  
Colon Cancer ◽  
Molecular Markers ◽  
Prognostic Value ◽  
Cox Regression ◽  
Prognostic Models ◽  
Clinical Predictors ◽  
Exon 2 ◽  
Translational Study ◽  
T Stage ◽  
N Stage

4002 Background: We compared the incidence of molecular markers in stage II (SII) and III (SIII) colon cancer and tested their prognostic value per stage, using PETACC 3, an adjuvant trial with 3,278 patients. We included expression of P53, SMAD4, thymidylate synthetase (TS) and hTERT, mutations of KRAS and BRAF, microsatellite instability (MSI) and 18qLOH. Methods: 1,564 formalin fixed paraffin embedded tissue blocks were prospectively collected and DNA from normal and tumor tissue was extracted after macrodissection. High P53, TS and hTERT expression and SMAD4 loss were assessed by immunohistochemistry. MSI was studied with 10 markers. KRAS exon 2 and BRAF exon 15 mutations were analyzed by allele specific real time PCR. 18qLOH was studied by pyrosequencing 7 SNPs. Prognostic value of the markers was analysed per stage by Cox regression for Relapse Free Survival (RFS). Results: marker frequencies and stage specific p-values in prognostic models in 420 SII and 984 SIII patients are listed in the table . Significant differences in frequency per stage were found for all markers except KRAS and BRAF. An interaction test for differences between marker prognostic value for SII and SIII was significant for MSI (p=0.04) and 18qLOH (p=0.04) in SII. Multivariate analysis including markers, T stage, N stage (for SIII), Tu grade, age <60, sex, treatment arm, and Tu site found T stage (p=0.0001) and MSI (p=0.02) as independently significant clinical predictors in SII; N stage (p<0.0001), T stage (p<0.0001), SMAD4 (p<0.0001) and P53 (p=0.01) in SIII. Conclusions: Molecular markers in colon cancer have a stage specific prognostic value. The possibility that the stages represent different diseases, rather than sequential steps in the evolution of a single disease, needs to be considered. [Table: see text] [Table: see text]

scholarly journals Survival Prediction By Baseline Systemic Immune-Inflammation Index And Its Changes During First-Line Platinum-Based Treatment In A Caucasian Population Of Patients With Metastatic Urothelial Carcinoma

2021 ◽  
Author(s):  
Patrik Palacka ◽  
Jan Slopovsky ◽  
Jana Obertova ◽  
Michal Chovanec ◽  
Katarina Rejlekova ◽  
...  
Keyword(s):  
Urothelial Carcinoma ◽  
Prognostic Value ◽  
Karnofsky Performance Status ◽  
Performance Status ◽  
Progression Free Survival ◽  
Rank Test ◽  
Survival Prediction ◽  
First Line ◽  
Metastatic Urothelial Carcinoma ◽  
Immune Inflammation

Abstract Background: Systemic immune-inflammation index (SII) predicts survival in patients with various malignancies. This study explores the prognostic value of SII in metastatic urothelial carcinoma (MUC) subjects.Methods: We evaluated 181 consecutive MUC (148 bladder, 27 upper tract) patients (135 men) treated with first-line platinum-based therapy. Karnofsky performance status <80% and visceral metastasis were present in 18.2% and 46.4%, respectively. SII was based on platelet x neutrophil/lymphocyte counts. Study population was dichotomized by median into high SII and low SII groups before the initiation of chemotherapy (median 1326) and at week 6 (median 705). Progression-free survival (PFS) and overall survival (OS) were estimated by the Kaplan-Meier method and compared with log-rank test. Results: At median follow-up of 9.6 months (range 1.7-191.1 months), 174 patients (96.1%) experienced disease progression and 173 (95.6%) died. Subjects with low SII at baseline and at week 6 had significantly better PFS (HR 0.58; P = 0.0002 and HR 0.55; P < 0.0001) and OS (HR 0.54; P < 0.0001 and HR 0.54; P < 0.0001) compared to patients with high SII. Independent prognostic value of SII was confirmed in a multivariate analysis. Patients with low SII at baseline and at week 6 had significantly better PFS and OS compared to patients with high SII at both timepoints (P < 0.0001).Conclusion: High SII before chemotherapy that persists at week 6 negatively affects survival. SII at baseline can be used in stratification of patients within clinical trials and in clinical practice.

scholarly journals Development of a radiomics model to diagnose pheochromocytoma preoperatively: a multicenter study with prospective validation

2022 ◽  
Vol 20 (1) ◽  
Author(s):  
Jianqiu Kong ◽  
Junjiong Zheng ◽  
Jieying Wu ◽  
Shaoxu Wu ◽  
Jinhua Cai ◽  
...  
Keyword(s):  
External Validation ◽  
Area Under The Curve ◽  
Multivariable Logistic Regression Analysis ◽  
Lasso Regression ◽  
Clinical Predictors ◽  
Internal Validation ◽  
Clinical Model ◽  
Magnetic Resonance Imaging Mri ◽  
Individualized Prediction ◽  
Radiomics Signature

Abstract Background Preoperative diagnosis of pheochromocytoma (PHEO) accurately impacts preoperative preparation and surgical outcome in PHEO patients. Highly reliable model to diagnose PHEO is lacking. We aimed to develop a magnetic resonance imaging (MRI)-based radiomic-clinical model to distinguish PHEO from adrenal lesions. Methods In total, 305 patients with 309 adrenal lesions were included and divided into different sets. The least absolute shrinkage and selection operator (LASSO) regression model was used for data dimension reduction, feature selection, and radiomics signature building. In addition, a nomogram incorporating the obtained radiomics signature and selected clinical predictors was developed by using multivariable logistic regression analysis. The performance of the radiomic-clinical model was assessed with respect to its discrimination, calibration, and clinical usefulness. Results Seven radiomics features were selected among the 1301 features obtained as they could differentiate PHEOs from other adrenal lesions in the training (area under the curve [AUC], 0.887), internal validation (AUC, 0.880), and external validation cohorts (AUC, 0.807). Predictors contained in the individualized prediction nomogram included the radiomics signature and symptom number (symptoms include headache, palpitation, and diaphoresis). The training set yielded an AUC of 0.893 for the nomogram, which was confirmed in the internal and external validation sets with AUCs of 0.906 and 0.844, respectively. Decision curve analyses indicated the nomogram was clinically useful. In addition, 25 patients with 25 lesions were recruited for prospective validation, which yielded an AUC of 0.917 for the nomogram. Conclusion We propose a radiomic-based nomogram incorporating clinically useful signatures as an easy-to-use, predictive and individualized tool for PHEO diagnosis.

scholarly journals Comparing the characteristics and predicting the survival of patients with head and neck melanoma versus body melanoma: a population-based study

BMC Cancer ◽  
2021 ◽  
Vol 21 (1) ◽  
Author(s):  
Yuxin Ding ◽  
Runyi Jiang ◽  
Yuhong Chen ◽  
Jing Jing ◽  
Xiaoshuang Yang ◽  
...  
Keyword(s):  
Head And Neck ◽  
Health Management ◽  
External Validation ◽  
Area Under The Curve ◽  
Population Based ◽  
Survival Prediction ◽  
Cancer Specific Survival ◽  
Cox Models ◽  
Lentigo Maligna Melanoma ◽  
Clinical Tools

Abstract Background Previous studies reported cutaneous melanoma in head and neck (HNM) differed from those in other regions (body melanoma, BM). Individualized tools to predict the survival of patients with HNM or BM remain insufficient. We aimed at comparing the characteristics of HNM and BM, developing and validating nomograms for predicting the survival of patients with HNM or BM. Methods The information of patients with HNM or BM from 2004 to 2015 was obtained from the Surveillance, Epidemiology, and End Results (SEER) database. The HNM group and BM group were randomly divided into training and validation cohorts. We used the Kaplan-Meier method and multivariate Cox models to identify independent prognostic factors. Nomograms were developed via the rms and dynnom packages, and were measured by the concordance index (C-index), the area under the curve (AUC) of the receiver operating characteristic (ROC) curve and calibration plots. Results Of 70,605 patients acquired, 21% had HNM and 79% had BM. The HNM group contained more older patients, male sex and lentigo maligna melanoma, and more frequently had thicker tumors and metastases than the BM group. The 5-year cancer-specific survival (CSS) and overall survival (OS) rates were 88.1 ± 0.3% and 74.4 ± 0.4% in the HNM group and 92.5 ± 0.1% and 85.8 ± 0.2% in the BM group, respectively. Eight variables (age, sex, histology, thickness, ulceration, stage, metastases, and surgery) were identified to construct nomograms of CSS and OS for patients with HNM or BM. Additionally, four dynamic nomograms were available on web. The internal and external validation of each nomogram showed high C-index values (0.785–0.896) and AUC values (0.81–0.925), and the calibration plots showed great consistency. Conclusions The characteristics of HNM and BM are heterogeneous. We constructed and validated four nomograms for predicting the 3-, 5- and 10-year CSS and OS probabilities of patients with HNM or BM. These nomograms can serve as practical clinical tools for survival prediction and individual health management.

scholarly journals POS1439 A MULTI-PREDICTOR MODEL TO PREDICT RISK OF SCLERODERMA RENAL CRISIS IN SYSTEMIC SCLEROSIS

2021 ◽  
Vol 80 (Suppl 1) ◽  
pp. 1004.1-1004
Author(s):  
D. Xu ◽  
R. Mu
Keyword(s):  
Systemic Sclerosis ◽  
Prediction Model ◽  
Renal Involvement ◽  
Area Under The Curve ◽  
Individual Risk ◽  
Multivariable Logistic Regression Analysis ◽  
Scleroderma Renal Crisis ◽  
Lasso Regression ◽  
Patients At Risk ◽  
Renal Crisis

Background:Scleroderma renal crisis (SRC) is a life-threatening syndrome. The early identification of patients at risk is essential for timely treatment to improve the outcome[1].Objectives:We aimed to provide a personalized tool to predict risk of SRC in systemic sclerosis (SSc).Methods:We tried to set up a SRC prediction model based on the PKUPH-SSc cohort of 302 SSc patients. The least absolute shrinkage and selection operator (Lasso) regression was used to optimize disease features. Multivariable logistic regression analysis was applied to build a SRC prediction model incorporating the features of SSc selected in the Lasso regression. Then, a multi-predictor nomogram combining clinical characteristics was constructed and evaluated by discrimination and calibration.Results:A multi-predictor nomogram for evaluating the risk of SRC was successfully developed. In the nomogram, four easily available predictors were contained including disease duration <2 years, cardiac involvement, anemia and corticosteroid >15mg/d exposure. The nomogram displayed good discrimination with an area under the curve (AUC) of 0.843 (95% CI: 0.797-0.882) and good calibration.Conclusion:The multi-predictor nomogram for SRC could be reliably and conveniently used to predict the individual risk of SRC in SSc patients, and be a step towards more personalized medicine.References:[1]Woodworth TG, Suliman YA, Li W, Furst DE, Clements P (2016) Scleroderma renal crisis and renal involvement in systemic sclerosis. Nat Rev Nephrol 12 (11):678-91.Disclosure of Interests:None declared

scholarly journals Prognostic Value of Plasma hPG80 (Circulating Progastrin) in Metastatic Renal Cell Carcinoma

Cancers ◽  
2021 ◽  
Vol 13 (3) ◽  
pp. 375
Author(s):  
Manish Kohli ◽  
Winston Tan ◽  
Bérengère Vire ◽  
Pierre Liaud ◽  
Mélina Blairvacq ◽  
...  
Keyword(s):  
Renal Cell Carcinoma ◽  
Cell Carcinoma ◽  
Renal Cell ◽  
Prognostic Value ◽  
Risk Model ◽  
Predictive Accuracy ◽  
Area Under The Curve ◽  
Risk Groups ◽  
Prognostic Impact ◽  
Significant Difference

Precise management of kidney cancer requires the identification of prognostic factors. hPG80 (circulating progastrin) is a tumor promoting peptide present in the blood of patients with various cancers, including renal cell carcinoma (RCC). In this study, we evaluated the prognostic value of plasma hPG80 in 143 prospectively collected patients with metastatic RCC (mRCC). The prognostic impact of hPG80 levels on overall survival (OS) in mRCC patients after controlling for hPG80 levels in non-cancer age matched controls was determined and compared to the International Metastatic Database Consortium (IMDC) risk model (good, intermediate, poor). ROC curves were used to evaluate the diagnostic accuracy of hPG80 using the area under the curve (AUC). Our results showed that plasma hPG80 was detected in 94% of mRCC patients. hPG80 levels displayed high predictive accuracy with an AUC of 0.93 and 0.84 when compared to 18–25 year old controls and 50–80 year old controls, respectively. mRCC patients with high hPG80 levels (>4.5 pM) had significantly lower OS compared to patients with low hPG80 levels (<4.5 pM) (12 versus 31.2 months, respectively; p = 0.0031). Adding hPG80 levels (score of 1 for patients having hPG80 levels > 4.5 pM) to the six variables of the IMDC risk model showed a greater and significant difference in OS between the newly defined good-, intermediate- and poor-risk groups (p = 0.0003 compared to p = 0.0076). Finally, when patients with IMDC intermediate-risk group were further divided into two groups based on hPG80 levels within these subgroups, increased OS were observed in patients with low hPG80 levels (<4.5 pM). In conclusion, our data suggest that hPG80 could be used for prognosticating survival in mRCC alone or integrated to the IMDC score (by adding a variable to the IMDC score or by substratifying the IMDC risk groups), be a prognostic biomarker in mRCC patients.

scholarly journals Value of radiomics in differential diagnosis of chromophobe renal cell carcinoma and renal oncocytoma

2019 ◽  
Vol 45 (10) ◽  
pp. 3193-3201 ◽  
Author(s):  
Yajuan Li ◽  
Xialing Huang ◽  
Yuwei Xia ◽  
Liling Long
Keyword(s):  
Machine Learning ◽  
Differential Diagnosis ◽  
Cell Carcinoma ◽  
Area Under The Curve ◽  
Image Features ◽  
Renal Tumors ◽  
Support Vector ◽  
Svm Classifier ◽  
Renal Oncocytoma ◽  
Lasso Regression

Abstract Purpose To explore the value of CT-enhanced quantitative features combined with machine learning for differential diagnosis of renal chromophobe cell carcinoma (chRCC) and renal oncocytoma (RO). Methods Sixty-one cases of renal tumors (chRCC = 44; RO = 17) that were pathologically confirmed at our hospital between 2008 and 2018 were retrospectively analyzed. All patients had undergone preoperative enhanced CT scans including the corticomedullary (CMP), nephrographic (NP), and excretory phases (EP) of contrast enhancement. Volumes of interest (VOIs), including lesions on the images, were manually delineated using the RadCloud platform. A LASSO regression algorithm was used to screen the image features extracted from all VOIs. Five machine learning classifications were trained to distinguish chRCC from RO by using a fivefold cross-validation strategy. The performance of the classifier was mainly evaluated by areas under the receiver operating characteristic (ROC) curve and accuracy. Results In total, 1029 features were extracted from CMP, NP, and EP. The LASSO regression algorithm was used to screen out the four, four, and six best features, respectively, and eight features were selected when CMP and NP were combined. All five classifiers had good diagnostic performance, with area under the curve (AUC) values greater than 0.850, and support vector machine (SVM) classifier showed a diagnostic accuracy of 0.945 (AUC 0.964 ± 0.054; sensitivity 0.999; specificity 0.800), showing the best performance. Conclusions Accurate preoperative differential diagnosis of chRCC and RO can be facilitated by a combination of CT-enhanced quantitative features and machine learning.

scholarly journals Prognostic Value of Inflammatory Biomarkers in Patients with Severe COVID-19: A Single-Center Retrospective Study

2021 ◽  
Vol 16 ◽  
pp. 117727192110270
Author(s):  
Gönül Açıksarı ◽  
Mehmet Koçak ◽  
Yasemin Çağ ◽  
Lütfiye Nilsun Altunal ◽  
Adem Atıcı ◽  
...  
Keyword(s):  
Retrospective Study ◽  
Hospital Mortality ◽  
Prognostic Value ◽  
Current Knowledge ◽  
Area Under The Curve ◽  
Prognostic Nutritional Index ◽  
Inflammatory Biomarkers ◽  
Single Center ◽  
Discriminative Ability ◽  
Lymphocyte Ratio

Background: The current knowledge about novel coronavirus-2019 (COVID-19) indicates that the immune system and inflammatory response play a crucial role in the severity and prognosis of the disease. In this study, we aimed to investigate prognostic value of systemic inflammatory biomarkers including C-reactive protein/albumin ratio (CAR), prognostic nutritional index (PNI), neutrophil-to-lymphocyte ratio (NLR), lymphocyte-to-monocyte ratio (LMR), and platelet-to-lymphocyte ratio (PLR) in patients with severe COVID-19. Methods: This single-center, retrospective study included a total of 223 patients diagnosed with severe COVID-19. Primary outcome measure was mortality during hospitalization. Multivariate logistic regression analyses were performed to identify independent predictors associated with mortality in patients with severe COVID-19. Receiver operating characteristic (ROC) curve was used to determine cut-offs, and area under the curve (AUC) values were used to demonstrate discriminative ability of biomarkers. Results: Compared to survivors of severe COVID-19, non-survivors had higher CAR, NLR, and PLR, and lower LMR and lower PNI ( P < .05 for all). The optimal CAR, PNI, NLR, PLR, and LMR cut-off values for detecting prognosis were 3.4, 40.2, 6. 27, 312, and 1.54 respectively. The AUC values of CAR, PNI, NLR, PLR, and LMR for predicting hospital mortality in patients with severe COVID-19 were 0.81, 0.91, 0.85, 0.63, and 0.65, respectively. In ROC analysis, comparative discriminative ability of CAR, PNI, and NLR for hospital mortality were superior to PLR and LMR. Multivariate analysis revealed that CAR (⩾0.34, P = .004), NLR (⩾6.27, P = .012), and PNI (⩽40.2, P = .009) were independent predictors associated with mortality in severe COVID-19 patients. Conclusions: The CAR, PNI, and NLR are independent predictors of mortality in hospitalized severe COVID-19 patients and are more closely associated with prognosis than PLR or LMR.

scholarly journals Machine learning applied on chest x-ray can aid in the diagnosis of COVID-19: a first experience from Lombardy, Italy

2021 ◽  
Vol 5 (1) ◽  
Author(s):  
Isabella Castiglioni ◽  
Davide Ippolito ◽  
Matteo Interlenghi ◽  
Caterina Beatrice Monti ◽  
Christian Salvatore ◽  
...  
Keyword(s):  
Deep Learning ◽  
Area Under The Curve ◽  
Clinical Information ◽  
Training Dataset ◽  
Independent Dataset ◽  
X Ray ◽  
Learning Classifier ◽  
Chest X Ray ◽  
Polymerase Chain ◽  
Deep Learning Model

Abstract Background We aimed to train and test a deep learning classifier to support the diagnosis of coronavirus disease 2019 (COVID-19) using chest x-ray (CXR) on a cohort of subjects from two hospitals in Lombardy, Italy. Methods We used for training and validation an ensemble of ten convolutional neural networks (CNNs) with mainly bedside CXRs of 250 COVID-19 and 250 non-COVID-19 subjects from two hospitals (Centres 1 and 2). We then tested such system on bedside CXRs of an independent group of 110 patients (74 COVID-19, 36 non-COVID-19) from one of the two hospitals. A retrospective reading was performed by two radiologists in the absence of any clinical information, with the aim to differentiate COVID-19 from non-COVID-19 patients. Real-time polymerase chain reaction served as the reference standard. Results At 10-fold cross-validation, our deep learning model classified COVID-19 and non-COVID-19 patients with 0.78 sensitivity (95% confidence interval [CI] 0.74–0.81), 0.82 specificity (95% CI 0.78–0.85), and 0.89 area under the curve (AUC) (95% CI 0.86–0.91). For the independent dataset, deep learning showed 0.80 sensitivity (95% CI 0.72–0.86) (59/74), 0.81 specificity (29/36) (95% CI 0.73–0.87), and 0.81 AUC (95% CI 0.73–0.87). Radiologists’ reading obtained 0.63 sensitivity (95% CI 0.52–0.74) and 0.78 specificity (95% CI 0.61–0.90) in Centre 1 and 0.64 sensitivity (95% CI 0.52–0.74) and 0.86 specificity (95% CI 0.71–0.95) in Centre 2. Conclusions This preliminary experience based on ten CNNs trained on a limited training dataset shows an interesting potential of deep learning for COVID-19 diagnosis. Such tool is in training with new CXRs to further increase its performance.

Sign in / Sign up

Export Citation Format

Share Document