AI-Driven Model for Automatic Emphysema Detection in Low-Dose Computed Tomography Using Disease-Specific Augmentation

Yeshaswini Nagaraj,Hendrik Joost Wisselink,Rozemarijn Vliegenthart,Raymond Veldhuis,Matthijs Oudkerk,Jiali Cai,Peter Van Ooijen,Grigory Sidorenkov,Mieneke Rook,Sunil Belur Nagaraj

doi:10.1007/s10278-022-00599-7

Abstract

The objective of this study is to evaluate the feasibility of a disease-specific deep learning (DL) model based on minimum intensity projection (minIP) for automated emphysema detection in low-dose computed tomography (LDCT) scans. LDCT scans of 240 individuals from a population-based cohort in the Netherlands (ImaLife study, mean age ± SD = 57 ± 6 years) were retrospectively chosen for training and internal validation of the DL model. For independent testing, LDCT scans of 125 individuals from a lung cancer screening cohort in the USA (NLST study, mean age ± SD = 64 ± 5 years) were used. Dichotomous emphysema diagnosis based on radiologists’ annotation was used to develop the model. The automated model included minIP processing (slab thickness range: 1 mm to 11 mm), classification, and detection maps generation. The data-split for the pipeline evaluation involved class-balanced and imbalanced settings. The proposed DL pipeline showed the highest performance (area under receiver operating characteristics curve) for 11 mm slab thickness in both the balanced (ImaLife = 0.90 ± 0.05) and the imbalanced dataset (NLST = 0.77 ± 0.06). For ImaLife subcohort, the variation in minIP slab thickness from 1 to 11 mm increased the DL model’s sensitivity from 75 to 88% and decreased the number of false-negative predictions from 10 to 5. The minIP-based DL model can automatically detect emphysema in LDCTs. The performance of thicker minIP slabs was better than that of thinner slabs. LDCT can be leveraged for emphysema detection by applying disease specific augmentation.

Highlights

Chronic obstructive pulmonary disease (COPD) is among the leading causes of early deaths worldwide [1], and 70% of COPD is estimated to be under-diagnosed [2, 3]
The notable contributions of this study are as follows: (1) we used an unsupervised deep learning (DL) algorithm to address the annotation-less and class-imbalanced scenarios that normally characterize lung cancer screening Low-dose computed tomography (LDCT); (2) we tested the feasibility of using clinical domain knowledge such as minimum intensity projection (minIP) to emphasize the minimal differences of emphysema regions in LDCT for unsupervised learning; (3) we explored the effects of different slab thickness of minIP on DL algorithm; (4) we generated detection maps to interpret the model predictions and to serve as a quality check; and (5) we validated our model on lung cancer screening data to check the efficacy of the proposed DL algorithm in a real use-case
The Imaging in Lifelines (ImaLife) subcohort used for training and internal validation consisted of 240 participants; the mean age ± SD at enrollment was 57 ± 6 years

Summary

Introduction

Chronic obstructive pulmonary disease (COPD) is among the leading causes of early deaths worldwide [1], and 70% of COPD is estimated to be under-diagnosed [2, 3]. Emphysema is a key component of COPD that is characterized by the destruction of lung parenchyma [4]. Emphysema is diagnosed at the later stages of the disease’s progression and is itself an independent risk factor for lung cancer [5]. Low-dose computed tomography (LDCT) has been shown to be capable of detecting lung cancer and provides an opportunity to detect comorbidities like emphysema in early stages [6]. LDCT contains inherent noise, and screening asymptomatic participants means there are more normal scans than abnormal ones, making emphysema

Objectives

Methods

Results

Discussion

Conclusion