Electric noses can smell atopy in asthma patients
Electric noses (eNose) can accurately detect atopy in patients with asthma and wheezing regardless of age, according to a recent study. These devices may prove valuable for asthma phenotyping.
“By using a composite of benchmarking supervised and unsupervised analysis techniques to analyse data on different eNose platforms used with four independent cohorts, we have shown that different eNoses can appropriately discriminate between individuals with atopic versus nonatopic asthma,” the researchers said.
“To our knowledge, this is the first study to investigate the ability of eNose technology to detect atopy in patients with asthma across different age groups,” they added.
A total of 655 individuals from four separate cohorts participated in the present study. While most (n=503) were asthmatic adults, 98 were school-aged children with asthma and 54 were preschoolers with wheezing. Atopy, determined through a skin prick test or immunoglobulin measurements, had prevalence rates ranging from 63 percent to 83 percent. [J Allerg Clin Immunol 2020;146:1045-1055]
Three different machine learning algorithms were used to distinguish atopic from nonatopic asthma in each of the four cohorts using the exhaled breath profiles. The resulting areas under the receiver operating characteristic curves (AUROCCs) were at least 0.85. In addition, the machine learning models showed relatively good accuracy in two separate validation sets.
Two devices were used to train the models in two separate sets. In the largest cohort (n=429), a SpiroNose device, a real-time eNose device, was used to train the machine learning algorithms. Validation was subsequently carried out in 25 percent of this cohort.
On the other hand, an offline eNose was used to train and validate the algorithms in a pooled set consisting of the remaining three cohorts.
Validation showed comparable results between the two cohorts. AUROCCs in the largest and pooled sets were relatively high, at 0.91 and 0.72, respectively. Excluding patients who were sensitized with nonaroallergens led to no meaningful change in the principal testing and validation results.
“We have shown that observations made by using offline eNose technology are replicated with the real-time SpiroNose technology where sensors are placed in line with standard equipment for spirometry,” the researchers said, pointing out that contrary to the per-cohort analysis, the AUROCC was lowered after pooling three data sets.
“This decrease may be related to differences in study populations with respect to age and asthma-associated characteristics and, possibly, to different eNose batch versions within subjects of the pooled cohorts, which may introduce more diverse VOC patterns,” they explained.
Nevertheless, the findings are encouraging and show that through supervised and unsupervised machine learning approaches, exhaled breath can be used reliably to detect atopy.
“The findings presented here suggest that exhaled breath analysis by eNose allows meaningful phenotyping of patients with asthma and may therefore be used in personalized clinical decisions related to asthma,” the researchers said.