Project Title

Value of lung ultrasound, chest radiographs, chest CT and artificial intelligence to diagnose acute heart failure in unselected breathless patients: an interdisciplinary prospective study

Background

The diagnosis of heart failure may be difficult in the acute setting and requires both signs and symptoms augmented by clinical judgement, biomarkers, pulmonary-, and cardiac imaging. Cardiology expertise is not always available in the emergency department, and an echocardiography is not feasible for all patients in the acute setting. Therefore, the heart failure diagnosis is initially often made by general internists with little training in cardiology, using pulmonary congestion to confirm the diagnosis in comorbid dyspnoeic patients. The PhD project is designed to investigate methods to objectify pulmonary congestion as a sign of heart failure in consecutive acute breathless patients.

Aim

Article 1: to determine if there is a higher probability of detecting pulmonary congestion with chest CT compared to chest radiographs in consecutive adult patients with dyspnoea in the emergency department.
Article 2: to identify the intravascular and interstitial congestion parameters that is most closely associated with patient outcomes, including signs by lung ultrasound, brain natriuretic peptides, chest X-ray and chest CT.
Article 3: to externally validate an artificial intelligence algorithm that is currently being developed by “Institut for Datalogi Københavns Universitet” (DIKU) to diagnose heart failure among unselected dyspnoeic patients from CT scans of thorax.

Methods

The study will be based on a prospective observational study

Preliminary results

Results from article 1:

Of 228 dyspnoeic patients, 64 (28%) had adjudicated AHF, and 79 (35%) had echo-bnp AHF. The probability to confirm adjudicated AHF and echo-bnp AHF were up to four times greater using chest CT compared to chest radiographs (conditional odds ratio (cOR): 3.89 [2.15,7.06] and cOR: 2.52 [1.45,4.38], and the interrater agreement was higher using chest CT (kappa 0.88 [95% CI: 0.81, 0.95] versus 0.73 [95% CI:0.63, 0.82]). As first line imaging modality, chest CT will find one additional adjudicated AHF in 12.5 patients and, prevent one false-positive in 20 patients. Similar results were demonstrated for echo-bnp AHF.