Chest X-Ray Classifier

Quibim

Quibim’s Chest X-ray Classifier aims to help radiologists deal with the large volumes of chest radiographs by prioritizing potentially pathological cases. This AI-fueled app is able to automatically identify PA/AP acquisitions and estimate the presence probability of 15 different findings in chest radiographs. Once these 15 probabilities are calculated, the classifier combines them to quantify the final abnormal probability of an image. The suite provides heatmaps that are displayed over the analyzed radiographs and that show the level of influence of each region of an image in the final abnormality score. By segmenting both lungs, the tool extracts textural features to foster posterior radiomic studies.

Product specifications

Information source: Vendor
Last updated: May 17, 2020

General
Product name	Chest X-Ray Classifier
Company	Quibim
Subspeciality	Chest
Modality	X-ray
Disease targeted	Atelectasis, Cardiomegaly, Consolidation, Edema, Emphysema, Enlarged cardiomediastinum, Fibrosis, Fracture, Hernia, Lung lesion, Lung opacity, Pleural effusion, Pleural thickening, Pneumothorax
Key-features	Abonormal probability %, work list triage
Suggested use	Before: adapting worklist order, flagging acute findings During: report suggestion Without interference of a radiologist: AI-only diagnosis
Data characteristics
Population	All chest x-rays
Input	Chest X-ray
Input format
Output	One-page structured report: abnormality heatmap, abnormality probability, most probable findings
Output format	DICOM Secondary Capture / PDF
Technology
Integration	Integration in standard reading environment (PACS), Integration RIS (Radiological Information System), Integration via AI marketplace or distribution platform, Stand-alone third party application, Stand-alone webbased
Deployment	Locally on dedicated hardware, Locally virtualized (virtual machine, docker), Cloud-based, Hybrid solution
Trigger for analysis	Automatically, right after the image acquisition
Processing time	10 - 60 seconds
Certification
CE	Certified, Class IIa , MDD
FDA	No or not yet
Market presence
On market since	05-2019
Distribution channels
Countries present (clinical, non-research use)	4
Paying clinical customers (institutes)	5+
Research/test users (institutes)	<5
Pricing
Pricing model	License
Based on	Number of installations, Number of analyses
Evidence
Peer reviewed papers on performance	Identifying pulmonary nodules or masses on chest radiography using deep learning: external validation and strategies to improve clinical practice. (read)
Non-peer reviewed papers on performance
Other relevant papers