Active learning for an efficient training strategy of computer-aided diagnosis systems: application to diabetic retinopathy screening.

The performance of computer-aided diagnosis (CAD) systems can be highly influenced by the training strategy. CAD systems are traditionally trained using available labeled data, extracted from a specific data distribution or from public databases. Due to the wide variability of medical data, these databases might not be representative enough when the CAD system is applied to data extracted from a different clinical setting, diminishing the performance or requiring more labeled samples in order to get better data generalization. In this work, we propose the incorporation of an active learning approach in the training phase of CAD systems for reducing the number of required training samples while maximizing the system performance. The benefit of this approach has been evaluated using a specific CAD system for Diabetic Retinopathy screening. The results show that (1) using a training set obtained from a different data source results in a considerable reduction of the CAD performance; and (2) using active learning the selected training set can be reduced from 1000 to 200 samples while maintaining an area under the Receiver Operating Characteristic curve of 0.856.

Practical considerations for optic nerve location in telemedicine.

The projected increase in diabetes in the United States and worldwide has created a need for broad-based, inexpensive screening for diabetic retinopathy (DR), an eye disease which can lead to vision impairment. A telemedicine network with retina cameras and automated quality control, physiological feature location, and lesion / anomaly detection is a low-cost way of achieving broad-based screening. In this work we report on the effect of quality estimation on an optic nerve (ON) detection method with a confidence metric. We report on an improvement of the method using a data set from an ophthalmologist practice then show the results of the method as a function of image quality on a set of images from an on-line telemedicine network collected in Spring 2009 and another broad-based screening program. We show that the fusion method, combined with quality estimation processing, can improve detection performance and also provide a method for utilizing a physician-in-the-loop for images that may exceed the capabilities of automated processing.

Retina lesion and microaneurysm segmentation using morphological reconstruction methods with ground-truth data.

In this work we report on a method for lesion segmentation based on the morphological reconstruction methods of Sbeh et. al. We adapt the method to include segmentation of dark lesions with a given vasculature segmentation. The segmentation is performed at a variety of scales determined using ground-truth data. Since the method tends to over-segment imagery, ground-truth data was used to create post-processing filters to separate nuisance blobs from true lesions. A sensitivity and specificity of 90% of classification of blobs into nuisance and actual lesion was achieved on two data sets of 86 images and 1296 images.

Elliptical local vessel density: a fast and robust quality metric for retinal images.

A great effort of the research community is geared towards the creation of an automatic screening system able to promptly detect diabetic retinopathy with the use of fundus cameras. In addition, there are some documented approaches for automatically judging the image quality. We propose a new set of features independent of field of view or resolution to describe the morphology of the patient's vessels. Our initial results suggest that these features can be used to estimate the image quality in a time one order of magnitude shorter than previous techniques.

Automated localization of the optic disc and the fovea.

The detection of the position of the normal anatomy in color fundus photographs is an important step in the automated analysis of retinal images. An automatic system for the detection of the position of the optic disc and the fovea is presented. The method integrates the use of local vessel geometry and image intensity features to find the correct positions in the image. A kNN regressor is used to accomplish the integration. Evaluation was performed on a set of 250 digital color fundus photographs and the detection performance for the optic disc and the fovea were 99.2% and 96.4% respectively.

Patients with persistent pain after enucleation studied by MRI dynamic color mapping and histopathology.

PURPOSE: To study possible causes of persistent pain in patients who underwent enucleation of the globe and in whom all other noninvasively detectable causes of pain had been ruled out. METHODS: Twenty patients were studied, 10 with intractable pain (score >5 on a 0-to-9 self-reporting pain scale) persisting for more than 6 months after enucleation (for various reasons) and 10 without pain (score <4) at least 6 months after enucleation. Magnetic resonance imaging (MRI) with dynamic color mapping (MRI-DCM) was used to quantify the motion of the optic nerve in millimeters per degree of gaze, 2 to 3 mm behind the implant. Histopathologic study of biopsy specimens was used to verify imaging findings. RESULTS: The optic nerve was attached to the implant in almost all (19/20) patients. On average, the motion was significantly less in patients with persistent intractable pain (0.04 mm/deg) than in patients without pain (0.08 mm/deg; normal orbit, 0.13 mm/deg). A biopsy specimen was available in 5 of 10 patients with persistent pain, and in 4 of those 5, microscopic neuroma was found close to the optic nerve-implant junction. CONCLUSIONS: In the enucleated orbit, the optic nerve is usually attached to the implant and soft tissue motion is decreased. In patients who have persistent pain after enucleation, motion is decreased even more, and a high percentage of microscopic amputation neuromas are found. Increased stiffness of orbital soft tissue and optic nerve attachment after enucleation are detectable using MRI-DCM, and may play a role in susceptible patients in the development of microscopic amputation neuroma and pain.

Objective quantification of the motion of soft tissues in the orbit.

Orbital soft-tissue motion analysis aids in the localization and diagnosis of orbital disorders. A technique has been developed to objectively quantify and visualize motion in the orbit during gaze. T1-weighted MR volume sequences are acquired during gaze and soft-tissue motion is quantified using optical flow techniques. The flow field is visualized using color-coding: orientation of the flow vector is coded by hue and magnitude by saturation of the pixel. Current clinical circumstances limit MR image acquisition to short sequences and short acquisition times. The effect of these limitations on the performance of optical flow computation has been studied for four representative optical flow algorithms: on short (nine frames) and long (21 frames) simulated sequences of rotation of a magnetic resonance (MR) imaged object, on short measured MR sequences of controlled rotation of the same object and on short MR sequences of motion in the orbit. On the short simulated and motion-controlled sequences, the Lucas and Kanade algorithm showed the best performance with respect to both accuracy and robustness. These motion estimates were accurate to within 20%. Motion in the orbit ranged between 0.05 and 0.25 mm/degree gaze. Color-coding was found to be attractive as a visualization technique, because it shows both magnitude and orientation of all flow vectors without cluttering.

MRI dynamic color mapping: a new quantitative technique for imaging soft tissue motion in the orbit.

PURPOSE: To investigate both feasibility and clinical potential of magnetic resonance imaging-dynamic color mapping (MRI-DCM) in measuring the motion of soft tissues in the orbit and in the diagnosis of orbital disorders by detecting changes in motion. METHODS: Sequences of MRI scans were acquired (acquisition time, 5 seconds) in a shoot-stop manner, while the patient fixated at a sequence of 13 gaze positions (8 degrees intervals). Motion was quantified off-line (in millimeters per degree of gaze change) using an optical flow algorithm. The motion was displayed in a color-coded image in which color saturation of a pixel shows the displacement and the hue the displacement's orientation. Six healthy volunteers and four patients (two with an orbital mass and two with acrylic ball implant after enucleation) were studied. RESULTS: The technique was found to be clinically feasible. For a gaze change of 1 degrees, orbital tissues moved between 0.0 and 0.25 mm/deg, depending on the type of tissue and location in the orbit. In the patients with an orbital mass, motion of the mass was similar to that of the medial rectus muscle, suggesting disease of muscular origin. In the enucleated orbits, soft tissue motion was decreased. One eye showed attachment of the optic nerve to the implant, which could be verified by biopsy. CONCLUSIONS: MRI-DCM allows noninvasive and quantitative measurement of soft tissue motion and the changes in motion due to pathologic conditions. In cases in which the diagnosis of a tumor in the apex is in doubt, it may reduce the need for biopsy. In contrast to static computed tomographic (CT) scans and MRIs, it can differentiate between juxtaposition and continuity and may be a new and promising tool in the differential diagnosis of intraorbital lesions.