Fingerprints as Predictors of Schizophrenia: A Deep Learning Study.

BACKGROUND AND HYPOTHESIS: The existing developmental bond between fingerprint generation and growth of the central nervous system points to a potential use of fingerprints as risk markers in schizophrenia. However, the high complexity of fingerprints geometrical patterns may require flexible algorithms capable of characterizing such complexity. STUDY DESIGN: Based on an initial sample of scanned fingerprints from 612 patients with a diagnosis of non-affective psychosis and 844 healthy subjects, we have built deep learning classification algorithms based on convolutional neural networks. Previously, the general architecture of the network was chosen from exploratory fittings carried out with an independent fingerprint dataset from the National Institute of Standards and Technology. The network architecture was then applied for building classification algorithms (patients vs controls) based on single fingers and multi-input models. Unbiased estimates of classification accuracy were obtained by applying a 5-fold cross-validation scheme. STUDY RESULTS: The highest level of accuracy from networks based on single fingers was achieved by the right thumb network (weighted validation accuracy = 68%), while the highest accuracy from the multi-input models was attained by the model that simultaneously used images from the left thumb, index and middle fingers (weighted validation accuracy = 70%). CONCLUSION: Although fitted models were based on data from patients with a well established diagnosis, since fingerprints remain lifelong stable after birth, our results imply that fingerprints may be applied as early predictors of psychosis. Specially, if they are used in high prevalence subpopulations such as those of individuals at high risk for psychosis.

Huellas dactilares como predictores de la esquizofrenia: un estudio de aprendizaje profundo / Fingerprints as Predictors of Schizophrenia: A Deep Learning Study

El vínculo existente entre la formación de huellas dactilares y el crecimiento del sistema nervioso central apunta a un uso potencial delas huellas dactilares como marcadores de riesgo en la esquizofrenia.Sin embargo, la elevada complejidad de los patrones geométricos delas huellas dactilares requiere algoritmos flexibles capaces de caracterizar dicha complejidad. A partir de una muestra inicial de huellasdactilares escaneadas de 612 pacientes con diagnóstico de psicosisno afectiva y 844 sujetos sanos, hemos construido algoritmos declasificación de aprendizaje profundo basados en redes neuronalesconvolucionales. Previamente, se eligió la arquitectura general de lared a partir de ajustes exploratorios realizados con un conjunto dedatos independiente de huellas dactilares del National Institute ofStandards and Technology. A continuación, la arquitectura de la redse aplicó para construir algoritmos de clasificación (paciente frentea controles) basados en modelos de un solo dedo y en modelos devarios dedos. Se obtuvieron estimaciones de la precisión de la clasificación aplicando un esquema de validación cruzada quíntuple.El mayor nivel de precisión de las redes basadas en un solo dedo loalcanzó la red del pulgar derecho (precisión = 68%), mientras quela mayor precisión de los modelos multientrada la obtuvo el modeloque utilizó simultáneamente imágenes de los dedos pulgar, índice ycorazón izquierdos (precisión = 70%). Aunque los modelos ajustadosse basaron en datos de pacientes con un diagnóstico bien establecido, dado que las huellas dactilares permanecen estables durantetoda la vida después del nacimiento, nuestros resultados implicanque las huellas dactilares pueden aplicarse como predictores tempranos de psicosis. Especialmente, si se utilizan en subpoblacionescon alta prevalencia de esquizofrenia, como las de personas con alto riesgo de psicosis. (AU)

The link between fingerprint generation and central nervous system growth points to a potential use of fingerprints as risk markersin schizophrenia. However, the high complexity of fingerprint geometric patterns requires flexible algorithms capable of characterizing such complexity. From an initial sample of fingerprints scanned from 612 patients diagnosed with non-affective psychosis and844 healthy subjects, we have built deep learning classification algorithms based on convolutional neural networks. Previously, thegeneral network architecture was chosen from exploratory fittingsperformed on an independent fingerprint dataset from the National Institute of Standards and Technology. The network architecturewas then applied for buinding classification algorithms (patientsvs. controls) based on single-finger models and multi-finger models. Classification accuracy estimates were obtained by applyinga 5-fold cross-validation scheme. The highest level of accuracy ofthe single-finger-based networks was achieved by the right thumbnetwork (accuracy = 68%), whereas the highest accuracy of themulti-input models was obtained by the model that simultaneouslyused images of the left thumb, index and middle fingers (accuracy =70%). Although the fitted models were based on data from patientswith a well-established diagnosis, given that fingerprints remainstable throughout life after birth, our results imply that fingerprintscan be applied as early predictors of psychosis. Especially, if usedin subpopulations with high prevalence of schizophrenia, such as those at high risk for psychosis. (AU)