[en] Introduction: Skin prick test (SPT) is the gold standard for identifying allergic sensitization in individuals with
a suspected airborne allergy. A novel device, Skin Prick Automated Test (SPAT), previously showed reduced
variability and less inconsistent test results compared to manual SPT (1,2). Additionally, an artificial
intelligence (AI)-assisted readout method showed high accuracy in supporting physicians to interpret skin
reactions following SPAT (3).
Objective: In the current study, we evaluated the impact of real-world confounders on the accuracy of the AI
model.
Methods: The images of the forearm of a validation cohort consisting of 217 patients undergoing SPAT (2604
wheals) were analysed for skin tone, presence of hair, scar tissue, tattoos, hyperpigmentation and imaging
artefacts. The individual typology angle (ITA) was used to measure apparent skin tone. Other confounders
were qualitatively assessed by visual inspection of the images.
Results: Darker skin tone (ITA <25) was observed in 16.1% of patients. Clear presence of hair, scar tissue,
tattoos, hyperpigmentation and imaging artefacts were observed in respectively 19.4, 3.7, 4.1, 25.3 and 6.0%
of patients. In patients with a darker skin tone (ITA -50 to <-25, -25 to <0, ITA 0 to 25) accuracy was
decreased to respectively 88.9%, 95.0% and 91.7% compared to other ITA categories (96.4-96.6%). In
patients with tattoo marks in between the prick locations (1.8% of patients), accuracy dropped to 85.4%
(compared to 95.7-98.3%). For the other confounders, accuracy remained >90% irrespective of the
presence or absence of the confounding variable.
Conclusion: The SPAT AI-assisted readout method showed high accuracy in most patients irrespective of
presence of hair, hyperpigmentation or scar tissue. Darker skin tone and the presence of tattoo marks were
observed in some patients and impacted the performance of the AI model. Enrichment of these patient
populations in the AI training cohort is needed to further optimize the AI mod
Disciplines :
Otolaryngology
Author, co-author :
Roux, Karolien; UZ Leuven - Universitair Ziekenhuis Leuven > Otorhinolaryngology-Head and Neck Surgery
Seys, Sven; Hippo Dx, Aarschot, Belgium
Hox, Valerie; UCL Saint-Luc - Brussels Saint-Luc University Hospital > ENT
Chaker, Adam; Klinikum rechts der Isar, Munich, Germany
De Greve, Glynnis; GZA Sint-Augustinus, Antwerp, Belgium > ENT
Lemmens, Winde; ZOL, Genk, Belgium > ENT
Poirrier, Anne-Lise ; Université de Liège - ULiège > Département des sciences cliniques > Oto-rhino-laryngologie et audiophonologie ; Centre Hospitalier Universitaire de Liège - CHU > > Service d'ORL, d'audiophonologie et de chir. cervico-faciale
Beckers, Eline; ZOL Genk > Otorhinolaryngology-Head and Neck Surgery
Daems, Rembert; Hippo Dx, Aarschot, Belgium
Diamant, Zuzana; KU Leuven - Katholieke Universiteit Leuven
Dierckx, Carmen; ZOL, Genk > Otorhinolaryngology-Head and Neck Surgery
Hellings, Peter; KU Leuven - Katholieke Universiteit Leuven
Huart, Caroline; UCL - Université Catholique de Louvain
Jerin, Claudia; Klinikum rechts der Isar, Munich, Germany > Otolaryngology, head and neck surgery
Jorissen, Marc; KU Leuven - Katholieke Universiteit Leuven
Loeckx, Dirk; Hippo Dx, Aarschot, Belgium
Oscé, Hanne; GZA Sint-Augustinus, Antwerp, Belgium > Otorhinolaryngology-Head and Neck Surgery
Thompson, Marc; Zurich University of Applied Sciences, Zurich, Switzerland
TOMBU, Sophie ; Centre Hospitalier Universitaire de Liège - CHU > > Service d'ORL, d'audiophonologie et de chir. cervico-faciale
Uyttebroek, Saartje; KU Leuven - Katholieke Universiteit Leuven
Zarowski, Andrzej; GZA Sint-Augustinus, Antwerp, Belgium > Otorhinolaryngology-Head and Neck Surgery
Gorris, Senne; Hippo Dx, Aarschot, Belgium
Van Gerven, Laura; KU Leuven - Katholieke Universiteit Leuven
