Observatorio de I+D+i UPM

Memorias de investigación
Thesis:
Image Processing Methods for Computer-Aided Screening for Disease
Year:2018
Research Areas
  • Medical equipment,
  • Electronic technology and of the communications
Information
Abstract
Screening for disease has been a widely accepted practice in healthcare since the twentieth century. Nowadays, in most developed countries we can find organized screening programs in different types of cancer (breast, cervical and colorectal are the most common) and for antenatal and newborns. Two clinical fields where the potential benefits of screening are very promising and have received worldwide attention are eye diseases, like diabetic retinopathy or glaucoma, and skin diseases, like melanoma and non-melanoma skin cancer. Screening programs have adopted a broad range of technological facilities. In this work we get an extensive insight in the role of image processing, telemedicine and Computer-Aided-Detection (CAD) systems and the benefits that can provide to aid in the fulfillment of the good practices and protocols established for a successful service. The use of medical images has become an indispensable diagnostic tool in many current screening programs. The great advance of medical imaging and image processing, their intensive use in the majority of the specialties to aid in the detection and follow-up of most conditions and their availability in health provider centers, have driven to be one of the tests typically performed in a screening campaign. Among all imaging modalities present in current screenings, we will focus on image processing methods for Color Fundus (CF) and Optical Coherence Tomography (OCT) images. While CF is basically used in retinal imaging, OCT can be used in retinal and skin imaging among many other fields. Telemedicine and e-health technology have played a significant role assisting in the provision of screening services. Associated to the most established screening services there have been remarkable telemedicine initiatives. The intensive use of imaging for the diagnosis and monitoring of eye diseases make telemedicine ideal for tele-screening, but it is still an open research field in pathologies like glaucoma that demands further developments to demonstrate its potential benefits. Glaucoma is one of the leading causes of global irreversible vision loss and it is estimated that affects 70 million people worldwide. This pathology represents a promising opportunity for screening due to the relevant percentage, up to 50% of population suffering the disease that is not aware of it and the benefits that an early diagnosis and treatment can represent to the patient. In this context, we have designed and implemented a new tele-screening tool with a multi-specialty and multi-disease approach to support some of the challenges identified in the screening services in the future. Besides, the tool was tested on more than 1000 patients in an ophthalmology screening campaign specially focused on glaucoma, with good results. This Thesis has also studied CAD systems which have demonstrated in many medical applications that can assist physicians in the interpretation of images for the detection and diagnosis of multiples pathologies. From all the steps involved in a typical CAD system we paid special attention to two of them: the pre-processing for image enhancement and the automatic classification of a specific disease. In the pre-processing we present a new imaging enhancement method for OCT images that can serve as the first step in CAD system for skin imaging as well as to improve emerging portable prototypes to be used in screening programs. We have tested the method successfully in several skin datasets, including an innovative qualitative assessment methodology to evaluate the clinical value of this type of image processing. Finally, for the automatic classification part, we present an extensive study of the application of deep convolutional neural networks to the challenging task of detecting glaucoma in color fundus images, which could serve as a valuable tool to be incorporated in CAD systems for large glaucoma screening campaigns.
International
Si
Type
Doctoral
Mark Rating
Sobresaliente cum laude
Date
27/09/2018
Participants
  • Autor: Juan Jose Gomez Valverde (UPM)
  • Director: Maria Jesus Ledesma Carbayo (UPM)
  • Director: Andres de Santos Lleo (UPM)
Research Group, Departaments and Institutes related
  • Creador: Grupo de Investigación: Tecnología de imágenes biomédicas
  • Centro o Instituto I+D+i: Centro de I+d+i en Procesado de la Información y Telecomunicaciones
  • Departamento: Ingeniería Electrónica
S2i 2020 Observatorio de investigación @ UPM con la colaboración del Consejo Social UPM
Cofinanciación del MINECO en el marco del Programa INNCIDE 2011 (OTR-2011-0236)
Cofinanciación del MINECO en el marco del Programa INNPACTO (IPT-020000-2010-22)