Deep learning diagnostic and risk-stratification pattern detection for COVID-19 in digital lung auscultations: clinical protocol for a case-control and prospective cohort study.

BACKGROUND: Lung auscultation is fundamental to the clinical diagnosis of respiratory disease. However, auscultation is a subjective practice and interpretations vary widely between users. The digitization of auscultation acquisition and interpretation is a particularly promising strategy for diagnosing and monitoring infectious diseases such as Coronavirus-19 disease (COVID-19) where automated analyses could help decentralise care and better inform decision-making in telemedicine. This protocol describes the standardised collection of lung auscultations in COVID-19 triage sites and a deep learning approach to diagnostic and prognostic modelling for future incorporation into an intelligent autonomous stethoscope benchmarked against human expert interpretation. METHODS: A total of 1000 consecutive, patients aged ≥ 16 years and meeting COVID-19 testing criteria will be recruited at screening sites and amongst inpatients of the internal medicine department at the Geneva University Hospitals, starting from October 2020. COVID-19 is diagnosed by RT-PCR on a nasopharyngeal swab and COVID-positive patients are followed up until outcome (i.e., discharge, hospitalisation, intubation and/or death). At inclusion, demographic and clinical data are collected, such as age, sex, medical history, and signs and symptoms of the current episode. Additionally, lung auscultation will be recorded with a digital stethoscope at 6 thoracic sites in each patient. A deep learning algorithm (DeepBreath) using a Convolutional Neural Network (CNN) and Support Vector Machine classifier will be trained on these audio recordings to derive an automated prediction of diagnostic (COVID positive vs negative) and risk stratification categories (mild to severe). The performance of this model will be compared to a human prediction baseline on a random subset of lung sounds, where blinded physicians are asked to classify the audios into the same categories. DISCUSSION: This approach has broad potential to standardise the evaluation of lung auscultation in COVID-19 at various levels of healthcare, especially in the context of decentralised triage and monitoring. TRIAL REGISTRATION: PB_2016-00500, SwissEthics. Registered on 6 April 2020.

High-resolution imaging of progressive articular cartilage degeneration.

The objective of this study was to develop and verify a new technique for monitoring the progression of osteoarthritis (OA) by combining a rat model with the imaging modality optical coherence tomography (OCT). Time-sequential, in vivo, OCT imaging was performed on the left femoral condyles of 12 Wistar rats following sodium-iodoacetic acid-induced OA progression. The right femoral condyles (untreated) were also imaged and served as controls. Imaging was performed on days 0, 10, 20, 30, and 60 with an OCT system capable of acquiring images at four frames per second and an axial resolution of 5 microm. Progressive changes were analyzed using an OA scoring system. OCT successfully identified progressive cartilage degeneration as well as alteration of the cartilage/bone interface. Significant changes to both of these structures were observed in the sodium-iodoacetic acid-injected condyles. Structural changes detected with OCT were confirmed histologically. OCT in combination with a well-known model used in arthritis research represents a powerful tool for following degenerative joint disease progression in a given animal by detecting changes to the cartilage/bone interface and articular cartilage.

Ultrahigh-resolution and 3-dimensional optical coherence tomography ex vivo imaging of the large and small intestines.

BACKGROUND: Ultrahigh-resolution optical coherence tomography (OCT) has an axial resolution of <5 microm, 2 to 3 times finer than standard OCT. This study investigates ultrahigh-resolution and three-dimensional OCT for ex vivo imaging of the large and small intestines and correlates images with histology. METHODS: Ultrahigh-resolution OCT imaging was performed on fresh surgical specimens from the large and small intestines in the pathology laboratory, and images were correlated with histology. OCT was performed at 1.3-microm wavelength with 4.5-microm axial x 11-microm transverse resolution and at 1.1-microm wavelength with 3.5-microm axial x 6-microm transverse resolution. Three-dimensional OCT also was investigated. RESULTS: Normal and pathologic areas from 23 surgical specimens of the large and small intestines were imaged. Ultrahigh-resolution OCT distinguished the epithelial layer of the mucosa and visualized individual villi, glands, and crypts. Finer transverse resolutions improved visualization of features, e.g., the epithelium, but reduced the depth of field. Architectural distortion of glands from inflammatory and neoplastic processes was observed. Three-dimensional rendering enabled visualization of surface pit pattern and mucosal folds as well as subsurface crypt microstructure. CONCLUSIONS: This study evaluates new OCT technology and can provide a baseline for interpreting future ultrahigh-resolution endoscopic OCT studies.

Multiple scattering in optical coherence tomography. I. Investigation and modeling.

We present a new model of optical coherence tomography (OCT) taking into account multiple scattering. A theoretical analysis and experimental investigation reveals that in OCT, despite multiple scattering, the field backscattered from the sample is generally spatially coherent and that the resulting interference signal with the reference field is stationary relative to measurement time. On the basis of this result, we model an OCT signal as a sum of spatially coherent fields with random-phase arguments--constant during measurement time--caused by multiple scattering. We calculate the mean of such a random signal from classical results of statistical optics and a Monte Carlo simulation. OCT signals predicted by our model are in very good agreement with a depth scan measurement of a sample consisting of a mirror covered with an aqueous suspension of microspheres. We discuss other comprehensive OCT models based on the extended Huygens-Fresnel principle, which rest on the assumption of partially coherent interfering fields.

Multiple scattering in optical coherence tomography. II. Experimental and theoretical investigation of cross talk in wide-field optical coherence tomography.

We present a comprehensive study of multiple-scattering effects in wide-field optical coherence tomography (OCT) realized with spatially coherent illumination. Imaging a sample made of a cleaved mirror embedded in an aqueous suspension of microspheres revealed that, despite temporal coherence gating, multiple scattering can induce significant coherent optical cross talk. The latter is a serious limitation to the method, since it prevents shot-noise-limited detection and diffraction-limited imaging in scattering samples. We investigate the dependence of cross talk on important system design parameters, as well as on some relevant sample properties. The agreement between theoretical and experimental results for the wide range of parameters investigated was very good, in both the lateral and the axial dimensions. This further confirms the validity of the model developed in our companion paper [J. Opt. Soc. Am. A 22, 1369-1379 (2005)].