Your browser doesn't support javascript.
DR-MIL: deep represented multiple instance learning distinguishes COVID-19 from community-acquired pneumonia in CT images.
Qi, Shouliang; Xu, Caiwen; Li, Chen; Tian, Bin; Xia, Shuyue; Ren, Jigang; Yang, Liming; Wang, Hanlin; Yu, Hui.
  • Qi S; College of Medicine and Biological Information Engineering, Northeastern University, Shenyang, China; Key Laboratory of Intelligent Computing in Medical Image, Ministry of Education, Northeastern University, Shenyang, China.
  • Xu C; College of Medicine and Biological Information Engineering, Northeastern University, Shenyang, China.
  • Li C; College of Medicine and Biological Information Engineering, Northeastern University, Shenyang, China.
  • Tian B; Department of Radiology, The Second People's Hospital of Guiyang, Guiyang, China.
  • Xia S; Department of Respiratory Medicine, Central Hospital Affiliated to Shenyang Medical College, Shenyang, China.
  • Ren J; Department of Radiology, The Affiliated Hospital of Guizhou Medical University, Guiyang, China.
  • Yang L; Department of Radiology, The Affiliated Hospital of Guizhou Medical University, Guiyang, China.
  • Wang H; Department of Radiology, General Hospital of the Yangtze River Shipping, Wuhan, China. Electronic address: 75288763@qq.com.
  • Yu H; Department of Radiology, The Affiliated Hospital of Guizhou Medical University, Guiyang, China. Electronic address: 331693861@qq.com.
Comput Methods Programs Biomed ; 211: 106406, 2021 Nov.
Article in English | MEDLINE | ID: covidwho-1401346
ABSTRACT
BACKGROUND AND

OBJECTIVE:

Given that the novel coronavirus disease 2019 (COVID-19) has become a pandemic, a method to accurately distinguish COVID-19 from community-acquired pneumonia (CAP) is urgently needed. However, the spatial uncertainty and morphological diversity of COVID-19 lesions in the lungs, and subtle differences with respect to CAP, make differential diagnosis non-trivial.

METHODS:

We propose a deep represented multiple instance learning (DR-MIL) method to fulfill this task. A 3D volumetric CT scan of one patient is treated as one bag and ten CT slices are selected as the initial instances. For each instance, deep features are extracted from the pre-trained ResNet-50 with fine-tuning and represented as one deep represented instance score (DRIS). Each bag with a DRIS for each initial instance is then input into a citation k-nearest neighbor search to generate the final prediction. A total of 141 COVID-19 and 100 CAP CT scans were used. The performance of DR-MIL is compared with other potential strategies and state-of-the-art models.

RESULTS:

DR-MIL displayed an accuracy of 95% and an area under curve of 0.943, which were superior to those observed for comparable methods. COVID-19 and CAP exhibited significant differences in both the DRIS and the spatial pattern of lesions (p<0.001). As a means of content-based image retrieval, DR-MIL can identify images used as key instances, references, and citers for visual interpretation.

CONCLUSIONS:

DR-MIL can effectively represent the deep characteristics of COVID-19 lesions in CT images and accurately distinguish COVID-19 from CAP in a weakly supervised manner. The resulting DRIS is a useful supplement to visual interpretation of the spatial pattern of lesions when screening for COVID-19.
Subject(s)
Keywords

Full text: Available Collection: International databases Database: MEDLINE Main subject: Pneumonia / Deep Learning / COVID-19 Type of study: Diagnostic study / Prognostic study Limits: Humans Language: English Journal: Comput Methods Programs Biomed Journal subject: Medical Informatics Year: 2021 Document Type: Article Affiliation country: J.cmpb.2021.106406

Similar

MEDLINE

...
LILACS

LIS


Full text: Available Collection: International databases Database: MEDLINE Main subject: Pneumonia / Deep Learning / COVID-19 Type of study: Diagnostic study / Prognostic study Limits: Humans Language: English Journal: Comput Methods Programs Biomed Journal subject: Medical Informatics Year: 2021 Document Type: Article Affiliation country: J.cmpb.2021.106406