Applying an artificial neural network model for developing a severity score for patients with hereditary amyloid polyneuropathy.

Autor: Novis S; a Universidade Federal do Rio de Janeiro, Hospital Universitário Clementino Fraga Filho (HUCFF-UFRJ) , Rio de Janeiro , Brazil., Machado F; b Signal Processing Lab, COPPE/Poli , Universidade Federal do Rio de Janeiro , Rio de Janeiro , Brazil., Costa VB; b Signal Processing Lab, COPPE/Poli , Universidade Federal do Rio de Janeiro , Rio de Janeiro , Brazil., Foguel D; c Instituto de Bioquímica Médica Leopoldo de Meis, Programa de Biologia Estrutural , Universidade Federal do Rio de Janeiro , Rio de Janeiro , Brazil., Cruz MW; a Universidade Federal do Rio de Janeiro, Hospital Universitário Clementino Fraga Filho (HUCFF-UFRJ) , Rio de Janeiro , Brazil., de Seixas JM; b Signal Processing Lab, COPPE/Poli , Universidade Federal do Rio de Janeiro , Rio de Janeiro , Brazil.
Jazyk: angličtina
Zdroj: Amyloid : the international journal of experimental and clinical investigation : the official journal of the International Society of Amyloidosis [Amyloid] 2017 Sep; Vol. 24 (3), pp. 153-161. Date of Electronic Publication: 2017 Jul 18.
DOI: 10.1080/13506129.2017.1343714
Abstrakt: Hereditary (familial) amyloid polyneuropathy (FAP) is a systemic disease that includes a sensorimotor polyneuropathy related to transthyretin (TTR) mutations. So far, a scale designed to classify the severity of this disease has not yet been validated. This work proposes the implementation of an artificial neural network (ANN) in order to develop a severity scale for monitoring the disease progression in FAP patients. In order to achieve this goal, relevant symptoms and laboratory findings were collected from 98 Brazilian patients included in THAOS - the Transthyretin Amyloidosis Outcomes Survey. Ninety-three percent of them bore Val30Met, the most prevalent variant of TTR worldwide; 63 were symptomatic and 35 were asymptomatic. These data were numerically codified for the purpose of constructing a Self-Organizing Map (SOM), which maps data onto a grid of artificial neurons. Mapped data could be clustered by similarity into five groups, based on increasing FAP severity (from Groups 1 to 5). Most symptoms were virtually absent from patients who mapped to Group 1, which also includes the asymptomatic patients. Group 2 encompasses the patients bearing symptoms considered to be initial markers of FAP, such as first signs of walking disabilities and lack of sensitivity to temperature and pain. Interestingly, the patients with cardiac symptoms, which also carry cardiac-associated mutations of the TTR gene (such as Val112Ile and Ala19Asp), were concentrated in Group 3. Symptoms such as urinary and fecal incontinence and diarrhea characterized particularly Groups 4 and 5. Renal impairment was found almost exclusively in Group 5. Model validation was accomplished by considering the symptoms from a sample with 48 additional Brazilian patients. The severity scores proposed here not only identify the current stage of a patient's disease but also offer to the physician an easy-to-read, 2D map that makes it possible to track disease progression.
Databáze: MEDLINE
Externí odkaz:
Nepřihlášeným uživatelům se plný text nezobrazuje