Melanoma and other skin lesion detection using smart handheld devices.

Smartphones of the latest generation featuring advanced multicore processors, dedicated microchips for graphics, high-resolution cameras, and innovative operating systems provide a portable platform for running sophisticated medical screening software and delivering point-of-care patient diagnostic services at a very low cost. In this chapter, we present a smartphone digital dermoscopy application that can analyze high-resolution images of skin lesions and provide the user with feedback about the likelihood of malignancy. The same basic procedure has been adapted to evaluate other skin lesions, such as the flesh-eating bacterial disease known as Buruli ulcer. When implemented on the iPhone, the accuracy and speed achieved by this application are comparable to that of a desktop computer, demonstrating that smartphone applications can combine portability and low cost with high performance. Thus, smartphone-based systems can be used as assistive devices by primary care physicians during routine office visits, and they can have a significant impact in underserved areas and in developing countries, where health-care infrastructure is limited.

SkinScan©: A PORTABLE LIBRARY FOR MELANOMA DETECTION ON HANDHELD DEVICES.

We have developed a portable library for automated detection of melanoma termed SkinScan© that can be used on smartphones and other handheld devices. Compared to desktop computers, embedded processors have limited processing speed, memory, and power, but they have the advantage of portability and low cost. In this study we explored the feasibility of running a sophisticated application for automated skin cancer detection on an Apple iPhone 4. Our results demonstrate that the proposed library with the advanced image processing and analysis algorithms has excellent performance on handheld and desktop computers. Therefore, deployment of smartphones as screening devices for skin cancer and other skin diseases can have a significant impact on health care delivery in underserved and remote areas.

Implementation of the 7-point checklist for melanoma detection on smart handheld devices.

In this paper we implement the 7-point checklist, a set of dermoscopic criteria widely used by clinicians for melanoma detection, on smart handheld devices, such as the Apple iPhone and iPad. The application developed is using sophisticated image processing and pattern recognition algorithms, yet it is light enough to run on a handheld device with limited memory and computational speed. When combined with a commercially available handheld dermoscope that provides proper lesion illumination, this application provides a truly self-contained handheld system for melanoma detection. Such a device can be used in a clinical setting for routine skin screening, or as an assistive diagnostic device in underserved areas and in developing countries with limited healthcare infrastructure.

EVALUATING SAMPLING STRATEGIES OF DERMOSCOPIC INTEREST POINTS.

Among the most critical components of a computerized system for automated melanoma detection is image sampling and pooling of the extracted features. In this paper, we propose a new method for sampling and pooling based on a combination of spatial pooling and graph theory features. The performance of the new method is evaluated using a dataset of more than 1,500 images representing pigmented skin lesions of known pathology. In our comparisons, we include several methods ranging from simple and multi-scale sampling on a regular grid to more sophisticated approaches, such as blob and curvilinear structure detectors. Our results show that, despite its simplicity, simple sampling on a regular grid provides highly competitive performance, compared to the more sophisticated approaches, while multi-scale sampling yields only trivial improvements. However, the proposed method provides significant performance improvement in terms of sensitivity and area under the receiver operating characteristic curve (95% t-test), and the best performance in terms of specificity compared to all other methods explored.

Assisting Main Task Learning by Heterogeneous Auxiliary Tasks with Applications to Skin Cancer Screening.

In typical classification problems, high level concept features provided by a domain expert are usually available during classifier training but not during its deployment. We address this problem from a multitask learning (MTL) perspective by treating these features as auxiliary learning tasks. Previous efforts in MTL have mostly assumed that all tasks have the same input space. However, auxiliary tasks can have different input spaces, since their learning targets are different. Thus, to handle cases with heterogeneous input, in this paper we present a newly developed model using heterogeneous auxiliary tasks to help main task learning. First, we formulate a convex optimization problem for the proposed model, and then, we analyze its hypothesis class and derive true risk bounds. Finally, we compare the proposed model with other relevant methods when applied to the problem of skin cancer screening and public datasets. Our results show that the performance of the proposed method is highly competitive compared to other relevant methods.

Functional connectivity networks in the autistic and healthy brain assessed using Granger causality.

In this study, we analyze brain connectivity based on Granger causality computed from magnetoencephalographic (MEG) activity obtained at the resting state in eight autistic and eight normal subjects along with measures of network connectivity derived from graph theory in an attempt to understand how communication in a human brain network is affected by autism. A connectivity matrix was computed for each subject individually and then group templates were estimated by averaging all matrices in each group. Furthermore, we performed classification of the subjects using support vector machines and Fisher's criterion to rank the features and identify the best subset for maximum separation of the groups. Our results show that a combined model based on connectivity matrices and graph theory measures can provide 87.5% accuracy in separating the two groups. These findings suggest that analysis of functional connectivity patterns may provide a valuable method for the early detection of autism.

Boosting instance prototypes to detect local dermoscopic features.

Local dermoscopic features are useful in many dermoscopic criteria for skin cancer detection. We address the problem of detecting local dermoscopic features from epiluminescence (ELM) microscopy skin lesion images. We formulate the recognition of local dermoscopic features as a multi-instance learning (MIL) problem. We employ the method of diverse density (DD) and evidence confidence (EC) function to convert MIL to a single-instance learning (SIL) problem. We apply Adaboost to improve the classification performance with support vector machines (SVMs) as the base classifier. We also propose to boost the selection of instance prototypes through changing the data weights in the DD function. We validate the methods on detecting ten local dermoscopic features from a dataset with 360 images. We compare the performance of the MIL approach, its boosting version, and a baseline method without using MIL. Our results show that boosting can provide performance improvement compared to the other two methods.

Modeling spatial relation in skin lesion images by the graph walk kernel.

Early skin cancer detection with the help of dermoscopic images is becoming more and more important. Previous methods generally ignored the spatial relation of the pixels or regions inside the lesion. We propose to employ a graph representation of the skin lesion to model the spatial relation. We then use the graph walk kernel, a similarity measure between two graphs, to build a classifier based on support vector machines for melanoma detection. In experiments, we compare the sensitivities and specificities of models with and without spatial information. Experimental results show that the model with spatial information performs the best in both sensitivity and specificity. Statistical test indicates that the improvement is significant.

Information communication networks in severe traumatic brain injury.

In this study we explored the use of coherence and Granger causality (GC) to separate patients in minimally conscious state (MCS) from patients with severe neurocognitive disorders (SND) that show signs of awareness. We studied 16 patients, 7 MCS and 9 SND with age between 18 and 49 years. Three minutes of ongoing electroencephalographic (EEG) activity was obtained at rest from 19 standard scalp locations, while subjects were alert but kept their eyes closed. GC was formulated in terms of linear autoregressive models that predict the evolution of several EEG time series, each representing the activity of one channel. The entire network of causally connected brain areas can be summarized as a graph of incompletely connected nodes. The 19 channels were grouped into five gross anatomical regions, frontal, left and right temporal, central, and parieto-occipital, while data analysis was performed separately in each of the five classical EEG frequency bands, namely delta, theta, alpha, beta, and gamma. Our results showed that the SND group consistently formed a larger number of connections compared to the MCS group in all frequency bands. Additionally, the number of connections in the delta band (0.1-4 Hz) between the left temporal and parieto-occipital areas was significantly different (P < 0.1%) in the two groups. Furthermore, in the beta band (12-18 Hz), the input to the frontal areas from all other cortical areas was also significantly different (P < 0.1%) in the two groups. Finally, classification of the subjects into distinct groups using as features the number of connections within and between regions in all frequency bands resulted in 100% classification accuracy of all subjects. The results of this study suggest that analysis of brain connectivity networks based on GC can be a highly accurate approach for classifying subjects affected by severe traumatic brain injury.

Coi-wiz: An interactive computer wizard for analyzing cardiac optical signals.

A number of revolutionary techniques have been developed for cardiac electrophysiology research to better study the various arrhythmia mechanisms that can enhance ablating strategies for cardiac arrhythmias. Once the three-dimensional high resolution cardiac optical imaging data is acquired, it is time consuming to manually go through them and try to identify the patterns associated with various arrhythmia symptoms. In this paper, we present an interactive computer wizard that helps cardiac electrophysiology researchers to visualize and analyze the high resolution cardiac optical imaging data. The wizard provides a file interface that accommodates different file formats. A series of analysis algorithms output waveforms, activation and action potential maps after spatial and temporal filtering, velocity field and heterogeneity measure. The interactive GUI allows the researcher to identify the region of interest in both the spatial and temporal domain, thus enabling them to study different heart chamber at their choice.

Malignant melanoma detection by Bag-of-Features classification.

In this paper, we apply a Bag-of-Features approach to malignant melanoma detection based on epiluminescence microscopy imaging. Each skin lesion is represented by a histogram of codewords or clusters identified from a training data set. Classification results using Naive Bayes classification and Support Vector Machines are reported. The best performance obtained is 82.21% on a dataset of 100 skin lesion images. Furthermore, since in melanoma screening false negative errors have a much higher impact and associated cost than false positive ones, we use the Neyman-Pearson score in our model selection scheme.