Chest X-ray Abnormality Detection by Using Artificial Intelligence: A Single-Site Retrospective Study of Deep Learning Model Performance
Chest X-ray (CXR) is one of the most common radiological examinations for both nonemergent and emergent clinical indications, but human error or lack of prioritization of patients can hinder timely interpretation. Deep learning (DL) algorithms have proven to be useful in the assessment of various ab...
Main Authors: | , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2023-01-01
|
Series: | BioMedInformatics |
Subjects: | |
Online Access: | https://www.mdpi.com/2673-7426/3/1/6 |
_version_ | 1797613295313616896 |
---|---|
author | Daniel Kvak Anna Chromcová Marek Biroš Robert Hrubý Karolína Kvaková Marija Pajdaković Petra Ovesná |
author_facet | Daniel Kvak Anna Chromcová Marek Biroš Robert Hrubý Karolína Kvaková Marija Pajdaković Petra Ovesná |
author_sort | Daniel Kvak |
collection | DOAJ |
description | Chest X-ray (CXR) is one of the most common radiological examinations for both nonemergent and emergent clinical indications, but human error or lack of prioritization of patients can hinder timely interpretation. Deep learning (DL) algorithms have proven to be useful in the assessment of various abnormalities including tuberculosis, lung parenchymal lesions, or pneumothorax. The deep learning–based automatic detection algorithm (DLAD) was developed to detect visual patterns on CXR for 12 preselected findings. To evaluate the proposed system, we designed a single-site retrospective study comparing the DL algorithm with the performance of five differently experienced radiologists. On the assessed dataset (n = 127) collected from the municipal hospital in the Czech Republic, DLAD achieved a sensitivity (Se) of 0.925 and specificity (Sp) of 0.644, compared to bootstrapped radiologists’ Se of 0.661 and Sp of 0.803, respectively, with statistically significant difference. The negative likelihood ratio (NLR) of the proposed software (0.12 (0.04–0.32)) was significantly lower than radiologists’ assessment (0.42 (0.4–0.43), <i>p</i> < 0.0001). No critical findings were missed by the software. |
first_indexed | 2024-03-11T06:53:51Z |
format | Article |
id | doaj.art-bee9d41bbafd48c0b4197a0ccf4a0163 |
institution | Directory Open Access Journal |
issn | 2673-7426 |
language | English |
last_indexed | 2024-03-11T06:53:51Z |
publishDate | 2023-01-01 |
publisher | MDPI AG |
record_format | Article |
series | BioMedInformatics |
spelling | doaj.art-bee9d41bbafd48c0b4197a0ccf4a01632023-11-17T09:48:40ZengMDPI AGBioMedInformatics2673-74262023-01-01318210110.3390/biomedinformatics3010006Chest X-ray Abnormality Detection by Using Artificial Intelligence: A Single-Site Retrospective Study of Deep Learning Model PerformanceDaniel Kvak0Anna Chromcová1Marek Biroš2Robert Hrubý3Karolína Kvaková4Marija Pajdaković5Petra Ovesná6Carebot, Ltd., 128 00 Prague, Czech RepublicCarebot, Ltd., 128 00 Prague, Czech RepublicCarebot, Ltd., 128 00 Prague, Czech RepublicCarebot, Ltd., 128 00 Prague, Czech RepublicCarebot, Ltd., 128 00 Prague, Czech RepublicCarebot, Ltd., 128 00 Prague, Czech RepublicInstitute of Biostatistics and Analysis, Ltd., 602 00 Brno, Czech RepublicChest X-ray (CXR) is one of the most common radiological examinations for both nonemergent and emergent clinical indications, but human error or lack of prioritization of patients can hinder timely interpretation. Deep learning (DL) algorithms have proven to be useful in the assessment of various abnormalities including tuberculosis, lung parenchymal lesions, or pneumothorax. The deep learning–based automatic detection algorithm (DLAD) was developed to detect visual patterns on CXR for 12 preselected findings. To evaluate the proposed system, we designed a single-site retrospective study comparing the DL algorithm with the performance of five differently experienced radiologists. On the assessed dataset (n = 127) collected from the municipal hospital in the Czech Republic, DLAD achieved a sensitivity (Se) of 0.925 and specificity (Sp) of 0.644, compared to bootstrapped radiologists’ Se of 0.661 and Sp of 0.803, respectively, with statistically significant difference. The negative likelihood ratio (NLR) of the proposed software (0.12 (0.04–0.32)) was significantly lower than radiologists’ assessment (0.42 (0.4–0.43), <i>p</i> < 0.0001). No critical findings were missed by the software.https://www.mdpi.com/2673-7426/3/1/6artificial intelligencecomputer-aided detectiondeep learningchest X-raypatient prioritization |
spellingShingle | Daniel Kvak Anna Chromcová Marek Biroš Robert Hrubý Karolína Kvaková Marija Pajdaković Petra Ovesná Chest X-ray Abnormality Detection by Using Artificial Intelligence: A Single-Site Retrospective Study of Deep Learning Model Performance BioMedInformatics artificial intelligence computer-aided detection deep learning chest X-ray patient prioritization |
title | Chest X-ray Abnormality Detection by Using Artificial Intelligence: A Single-Site Retrospective Study of Deep Learning Model Performance |
title_full | Chest X-ray Abnormality Detection by Using Artificial Intelligence: A Single-Site Retrospective Study of Deep Learning Model Performance |
title_fullStr | Chest X-ray Abnormality Detection by Using Artificial Intelligence: A Single-Site Retrospective Study of Deep Learning Model Performance |
title_full_unstemmed | Chest X-ray Abnormality Detection by Using Artificial Intelligence: A Single-Site Retrospective Study of Deep Learning Model Performance |
title_short | Chest X-ray Abnormality Detection by Using Artificial Intelligence: A Single-Site Retrospective Study of Deep Learning Model Performance |
title_sort | chest x ray abnormality detection by using artificial intelligence a single site retrospective study of deep learning model performance |
topic | artificial intelligence computer-aided detection deep learning chest X-ray patient prioritization |
url | https://www.mdpi.com/2673-7426/3/1/6 |
work_keys_str_mv | AT danielkvak chestxrayabnormalitydetectionbyusingartificialintelligenceasinglesiteretrospectivestudyofdeeplearningmodelperformance AT annachromcova chestxrayabnormalitydetectionbyusingartificialintelligenceasinglesiteretrospectivestudyofdeeplearningmodelperformance AT marekbiros chestxrayabnormalitydetectionbyusingartificialintelligenceasinglesiteretrospectivestudyofdeeplearningmodelperformance AT roberthruby chestxrayabnormalitydetectionbyusingartificialintelligenceasinglesiteretrospectivestudyofdeeplearningmodelperformance AT karolinakvakova chestxrayabnormalitydetectionbyusingartificialintelligenceasinglesiteretrospectivestudyofdeeplearningmodelperformance AT marijapajdakovic chestxrayabnormalitydetectionbyusingartificialintelligenceasinglesiteretrospectivestudyofdeeplearningmodelperformance AT petraovesna chestxrayabnormalitydetectionbyusingartificialintelligenceasinglesiteretrospectivestudyofdeeplearningmodelperformance |