Rapid detection of NAFLD and its evolutionary stages toward cirrhosis at the targeted population through multiparametric liver ultrasonographic screening (MLUS) and artificial intelligence with fibrosis risk stratification by family physicians.
Mihai Iacob, Ana Remes, Madalina Stoican
Keywords: Multiparametric Liver Ultrasound (MLUS), NAFLD screening in primary healthcare, Morfometric-ultrasound, APRI Score, artificial intelligence, Point of Care Ultrasonography(POCUS), .
Background:
NAFLD is a global public health issue, which progressively covers a spectrum of liver pathology, including steatosis, steatohepatitis, fibrosis, and cirrhosis, and their incidence increases exponentially. This study aimed to evaluate the diagnostic accuracy of the multiparametric-liver-ultrasonographic-screening with uses of artificial-intelligence performed by family doctors, compared to the evaluation performed by a specialist, at the targeted patients with a high-risk of NAFLD/NASH.
Research questions:
How can we improve the early diagnosis of NAFLD progression to NASH/fibrosis and cirrhosis to high-risk-patients in primary healthcare?
Method:
We conducted a multiparametric-liver-ultrasound-screening(MLUS) on 4751patients, with a high-risk of NAFLD/NASH, which presented as inclusion criteria: mixed dyslipidemia, obesity(BMI≥30), type2-diabetes, metabolic-syndrome(NCEP-criteria), chronic-lithiasis-cholecystitis, liver cirrhosis, chronic-hepatitis-B/hepatitis-C. APRI-score was initially calculated to stratify the fibrosis risk.
We use "standard-protocol", which could improve reproducibility and facilitate dynamic comparison, in grayscale, color/power-Doppler-US, and Strain-Elastography in standard-liver-scans as:transverse,oblique,and longitudinal-views. We established the cut-off/median-values(morphometric-ultrasound) of normal-ratios, between the anterior-posterior-diameters of the normal-liver-segments(Couinaud)/lobes, with the kidney/spleen-long-axis-ratio(not influenced by fatty-tissue-loading).
The high-risk-patients identified with NAFLD were first examined by a experienced-family-doctor subsequently compared with ultrasound-review by the specialist. We have developed a Smart-Computerized-Diagnostic-Algorithm of NAFLD/ NASH-pathology for US-diagnosis by family-physicians. The agreement between family-physicians and specialists on each finding was evaluated using Cohen’s-kappa-coefficient.
Results:
We identified 4751-patients with NAFLD/NASH,or cirrhosis and subsequently confirmed by the specialist. The positive-results of this screening were:2592 steatosis, NASH/steatofibrosis 971persons, and 22cases with Cirrhosis. The accuracy of liver-US-screening by family-physicians was:95,87% with95%CI=95.27%to96.42%,Sensitivity:97,12%,Specificity:91,59%, which were subsequently confirmed by the specialist as the"Gold-Standard"-method through fibroscan. The prevalence of liver-pathology was:77,48% with 95%CI:76,26%at78.66%. Reports of the two groups of specialists for identifying NAFLD/NASH showed a very-good strength of agreement-k=0.875;95%CI=0.864–0.887,standard-error:0,005.
Conclusions:
The uses of Multiparametric-Liver-Ultrasound-Screening(MLUS), morphometric-US(MUS), and artificial-intelligence(AI), performed by trained-family-physicians are comparable to diagnostic performed by the gastroenterologist. The use of a diagnostic-algorithm based on ratios between the axes of organs, using artificial-intelligence can identify early fatty-liver.
Points for discussion:
Is it possible to perform liver multiparametric ultrasound in a multidisciplinary screening team by family doctors specially trained in this regard?
How can artificial intelligence help us together with ultrasound technology as a diagnostic method in the practice of family doctors?
Can we consider clinical-morfometric-ultrasound as a means of diagnosing NAFLD(Nonalcoholic fatty liver disease) or NASH(Nonalcoholic-steatohepatitis) pathology by family physicians?