LNDetector: Automatic Detection, Segmentation and Classification of Pulmonary Nodules System in Computed Tomography Images (PTDC/EEI-SII/6599/2014)
PI: Aurélio Campilho (INESC TEC/FEUP)
Research teams: INESC TEC – C-BER, Faculty of Medicine, University of Porto
Starting date: June 1, 2016 Duration: 36 months
Keywords: Medical Image Analysis; Computer-aided diagnosis; Lung Nodules; Computed Tomography
Abstract:
Lung cancer is the world’s most deadliest type of cancer, in 2012 approximately 1.8 million new cases and 1.6 million related deaths were accounted all over the world. It represents approximately 20% of all medical cases with lung nodules. The main contributing factor to successful treatment is early diagnosis. For this purpose radiologists must be capable of performing detailed search throughout chest computerized tomography (CT) scans. Each CT scans generates between 300–500 slices making the diagnosis search a very time consuming and often physically demanding procedure and leading to errors. Possible omissions can lead to late diagnosis.
Computer-Aided Diagnosis (CAD) has become one of the most active research areas within medical image analysis and the automatic detection of pulmonary nodules in CT scans one of the most studied CAD applications.
The main research challenges in the development of CAD system are: (i) accurate segmentation of the lung fields; (ii) designing an efficient CAD system for nodule detection (CADe) able to detect nodules of different shapes, nodules attached to the lung borders, and small nodules; (iii) improving lung nodule segmentation techniques; (iv) nodule volumetric measurements of growth rate should take into account the global motion of the patients; (v) find methods and nodule descriptors for diagnosing of special types of lung nodules such as cavities and ground glass nodules (GGO); (vi) develop qualitative measures for describing the whole shape and appearance of the detected nodules; and (vii) provide larger databases for efficient validation.
Lung Nodule Detector (LNDetector) proposes a research line for the automatic detection, segmentation and classification of nodular regions in chest CT scans, by embedding radiological knowledge in nodule characterization and in CT volume search strategies. The main goal is to create a working prototype of a CAD system for lung nodule diagnosis to be introduced and evaluated in a radiology department. The prototype will include three main modules for nodule detection, nodule segmentation and nodule classification.