Kwok Pun Lee

Feature Subset Selection for Improving the Performance of False Positive Reduction in Lung Nodule CAD

In this paper, we propose a feature subset selection method based on genetic algorithms to improve the performance of false positive reduction in lung nodule CAD. It is coupled with a classifier based on support vector machines. The proposed approach determines automatically the optimal size of the feature set and chooses the most relevant features from a feature pool. Its performance was tested using a lung nodule database (66 true nodules and 123 false ones) acquired by multi-slice CT scans. From 23 features calculated for each detected structure, the suggested method determined 9 as the optimal feature subset size and selected the nine features. A support vector machine-based classifier trained with the optimal feature subset has resulted in 92.4% sensitivity and 85.4% specificity using leave-one-out cross validation. Experiments also showed significant improvement achieved by a system incorporating the proposed method over a system without it. It can be also applied to other machine learning problems: e.g. computer-aided diagnosis of lung nodules.

XML Schema Representation of DICOM Structured Reporting

OBJECTIVE The Digital Imaging and Communications in Medicine (DICOM) Structured Reporting (SR) standard improves the expressiveness, precision, and comparability of documentation about diagnostic images and waveforms. It supports the interchange of clinical reports in which critical features shown by images and waveforms can be denoted unambiguously by the observer, indexed, and retrieved selectively by subsequent reviewers. It is essential to provide access to clinical reports across the health care enterprise by using technologies that facilitate information exchange and processing by computers as well as provide support for robust and semantically rich standards, such as DICOM. This is supported by the current trend in the healthcare industry towards the use of Extensible Markup Language (XML) technologies for storage and exchange of medical information. The objective of the work reported here is to develop XML Schema for representing DICOM SR as XML documents. DESIGN We briefly describe the document type definition (DTD) for XML and its limitations, followed by XML Schema (the intended replacement for DTD) and its features. A framework for generating XML Schema for representing DICOM SR in XML is presented next. MEASUREMENTS None applicable. RESULTS A schema instance based on an SR example in the DICOM specification was created and validated against the schema. The schema is being used extensively in producing reports on Philips Medical Systems ultrasound equipment. CONCLUSION With the framework described it is feasible to generate XML Schema using the existing DICOM SR specification. It can also be applied to generate XML Schemas for other DICOM information objects.

Information object definition-based unified modeling language representation of DICOM structured reporting: a case study of transcoding DICOM to XML.

Supplement 23 to DICOM (Digital Imaging and Communications for Medicine), Structured Reporting, is a specification that supports a semantically rich representation of image and waveform content, enabling experts to share image and related patient information. DICOM SR supports the representation of textual and coded data linked to images and waveforms. Nevertheless, the medical information technology community needs models that work as bridges between the DICOM relational model and open object-oriented technologies. The authors assert that representations of the DICOM Structured Reporting standard, using object-oriented modeling languages such as the Unified Modeling Language, can provide a high-level reference view of the semantically rich framework of DICOM and its complex structures. They have produced an object-oriented model to represent the DICOM SR standard and have derived XML-exchangeable representations of this model using World Wide Web Consortium specifications. They expect the model to benefit developers and system architects who are interested in developing applications that are compliant with the DICOM SR specification.

Feature subset selection for improving the performance of false positive reduction in lung nodule CAD

We propose a feature subset selection method based on genetic algorithms to improve the performance of false positive reduction in lung nodule computer-aided detection (CAD). It is coupled with a classifier based on support vector machines. The proposed approach determines automatically the optimal size of the feature set, and chooses the most relevant features from a feature pool. Its performance was tested using a lung nodule database (52 true nodules and 443 false ones) acquired by multislice CT scans. From 23 features calculated for each detected structure, the suggested method determined ten to be the optimal feature subset size, and selected the most relevant ten features. A support vector machine classifier trained with the optimal feature subset resulted in 100% sensitivity and 56.4% specificity using an independent validation set. Experiments show significant improvement achieved by a system incorporating the proposed method over a system without it. This approach can be also applied to other machine learning problems; e.g. computer-aided diagnosis of lung nodules

Predicting respiratory instability in the ICU

Acute lung injury (ALI) and acute respiratory distress syndrome (ARDS) contribute to the morbidity and mortality of intensive care patients worldwide, and have large associated human and financial costs. We identified a reference data set of 624 mechanically-ventilated patients in the MIMIC-II intensive care database with and without low PaO2/FiO2 ratios (termed respiratory instability), and developed prediction algorithms for distinguishing these patients prior to the critical event. In the end, we had four rule sets using mean airway pressure, plateau pressure, total respiratory rate and oxygen saturation (SpO2), where the specificity/sensitivity rates were either 80%/60% or 90%/50%.

Heuristics to determine ventilation times of ICU patients from the MIMIC-II database

Mechanical ventilation is an important life support tool for patients in intensive care units (ICU). For various research purposes related to patient hemodynamic and cardiopulmonary monitoring, it is important to know when a patient is on a ventilator. Unfortunately, the widely used MIMIC-II database contains results from user charted data, where the user did not always store ventilation on and off times explicitly and accurately. The resulting ventilation-related data are subject to error. Therefore, there are no simple rules to define ventilation times retrospectively for this dataset. Hence, we designed a simple set of rules to determine the ventilation times using multiple sources of mechanical ventilator-related settings and physiological measurements by expert heuristics. The rules worked well in comparison with nursing notes regarding ventilation events. We conclude that our rule sets for determining ventilation times may be useful in assisting with MIMIC-II database analysis.