Alex Skovsbo Jørgensen

Semi-automatic vessel tracking and segmentation using epicardial ultrasound in bypass surgery

The purpose of intraoperative quality assessment of coronary artery bypass graft surgery is to confirm graft patency and disclose technical errors to reduce cardiac mortality, morbidity and improve clinical outcome for the patient. Epicardial ultrasound has been suggested as an alternative approach for quality assessment of anastomoses. To make a quantitative assessment of the anastomotic quality using ultrasound images, the vessel border has to be delineated to estimate the area of the vessel lumen. A tracking and segmentation algorithm was developed consisting of an active contour modeling approach and quality control of the segmentations. Evaluation of the tracking algorithm showed 91.96% of the segmentations were segmented correct with a mean error in height and width at 5.65% and 11.50% respectively.

Using cell nuclei features to detect colon cancer tissue in hematoxylin and eosin stained slides

Currently, diagnosis of colon cancer is based on manual examination of histopathological images by a pathologist. This can be time consuming and interpretation of the images is subject to inter‐ and intra‐observer variability. This may be improved by introducing a computer‐aided diagnosis (CAD) system for automatic detection of cancer tissue within whole slide hematoxylin and eosin (H&E) stains. Cancer disrupts the normal control mechanisms of cell proliferation and differentiation, affecting the structure and appearance of the cells. Therefore, extracting features from segmented cell nuclei structures may provide useful information to detect cancer tissue. A framework for automatic classification of regions of interest (ROI) containing either benign or cancerous colon tissue extracted from whole slide H&E stained images using cell nuclei features was proposed. A total of 1,596 ROIs were extracted from 87 whole slide H&E stains (44 benign and 43 cancer). A cell nuclei segmentation algorithm consisting of color deconvolution, k‐means clustering, local adaptive thresholding, and cell separation was performed within the ROIs to extract cell nuclei features. From the segmented cell nuclei structures a total of 750 texture and intensity‐based features were extracted for classification of the ROIs. The nine most discriminative cell nuclei features were used in a random forest classifier to determine if the ROIs contained benign or cancer tissue. The ROI classification obtained an area under the curve (AUC) of 0.96, sensitivity of 0.88, specificity of 0.92, and accuracy of 0.91 using an optimized threshold. The developed framework showed promising results in using cell nuclei features to classify ROIs into containing benign or cancer tissue in H&E stained tissue samples.

Whole sensory nerve recordings with spiral nerve cuff electrode

We have used a self-curling nerve cuff electrode to record sensory information from a cutaneous nerve. This type of cuffs has previously been used only for stimulation, but its mechanical properties could make it very suitable for recording also, since it can be fitted closer to the nerve than traditional cuffs without compromising the nerve. In this study we show that it is possible to record neural signals with at least the same quality as traditional cuffs.

Accurate Measurement of Airway Morphology on Chest CT Images

In recent years, the ability to accurately measuring and analyzing the morphology of small pulmonary structures on chest CT images, such as airways, is becoming of great interest in the scientific community. As an example, in COPD the smaller conducting airways are the primary site of increased resistance in COPD, while small changes in airway segments can identify early stages of bronchiectasis. To date, different methods have been proposed to measure airway wall thickness and airway lumen, but traditional algorithms are often limited due to resolution and artifacts in the CT image. In this work, we propose a Convolutional Neural Regressor (CNR) to perform cross-sectional measurements of airways, considering wall thickness and airway lumen at once. To train the networks, we developed a generative synthetic model of airways that we refined using a Simulated and Unsupervised Generative Adversarial Network (SimGAN). We evaluated the proposed method by first computing the relative error on a dataset of synthetic images refined with SimGAN, in comparison with other methods. Then, due to the high complexity to create an in-vivo ground-truth, we performed a validation on an airway phantom constructed to have airways of different sizes. Finally, we carried out an indirect validation analyzing the correlation between the percentage of the predicted forced expiratory volume in one second (FEV1%) and the value of the Pi10 parameter. As shown by the results, the proposed approach paves the way for the use of CNNs to precisely and accurately measure small lung airways with high accuracy.

Exploiting Multiple Color Representations to Improve Colon Cancer Detection in Whole Slide H&E Stains

Currently, colon cancer diagnosis is based on manual assessment of tissue samples stained with hematoxylin and eosin (H&E). This is a high volume, time consuming, and subjective task which could be aided by automatic cancer detection. We propose an algorithm for automatic cancer detection within WSI H&E stains using a multi class colon tissue classifier based on features extracted from 5 different color representations. Approx. 32000 tissue patches were extracted for the classifier from manual annotations of 9 representative colon tissue types from 74 WSI H&E stains. Colon tissue classifiers based on gray level or color features were trained using leave-one-out forward selection. The best colon tissue classifier was based on color texture features obtaining an average tissue precision-recall (PR) area under the curve (AUC) of 0.886 and a cancer PR-AUC of 0.950 on 20 validation WSI H&E stains.

Automatic coronary artery bypass anastomosis segmentation in epicardial ultrasound

Biopsy is commonly used to confirm cancer diagnosis when radiologically indicated. Given the ability of PET to localize malignancies in heterogeneous tumors and tumors that do not have a CT correlate, PET/CT guided biopsy may improve the diagnostic yield of biopsies. To facilitate PET/CT guided needle biopsy, we developed a workflow that allows us to bring PET image guidance into the interventional CT suite. In this abstract, we present SlicerPET, a user-friendly workflow based module developed using open source software libraries to guide needle biopsy in the interventional suite.

Detection and Segmentation of Anastomoses in Epicardial Ultrasound Images for Quality Assessment of Coronary Artery Bypass Graft Surgery

Up to 9% of coronary artery bypass graft surgery (CABG) anastomoses contain stenoses >50% post-surgery. This can cause post-operative morbidity and mortality for the patients. Intraoperative anastomosis quality assessment can be used to detect anastomotic errors to enable anastomosis revision during the primary surgery. Epicardial ultrasound (EUS) can be used to locate errors and quantify the stenotic rates within anastomoses to determine the anastomotic quality. Currently, the anastomotic quality is evaluated manually from EUS images as no objective methods are available. This can be time consuming and surgeons have to be trained in interpreting EUS images or use peer reviews by a radiologist. The aim of this thesis was to develop medical image analysis methods to enable automatic quantification of stenotic rates from in vivo EUS sequences of CABG anastomoses made on healthy porcine vessels. For this purpose methods were developed to automatically detect and extract of the vessel lumen area of anastomotic structures within in vivo EUS sequences. The anastomosis detection was used to locate anastomotic structures within EUS images to remove human interaction in analysis of EUS sequences. To extract the vessel lumen area of anastomotic structures from in vivo EUS sequences approaches for vessel lumen segmentation, inter-frame vessel motion correction, and segmentation quality control were developed. An area under the curve of 0.966 (95% CI: 0.951-0.984) and 0.989 (95% CI: 0.985-0.993, p < 0.001) of a precision-recall and receiver operator characteristic curve respectively, was obtained in detecting vessel regions extracted within EUS images in the anastomosis detection algorithm. The vessel lumen area of anastomotic structures was extracted with a mean Dice coefficient of 0.85 (± 0.13) and a mean absolute area difference of 20.62% (± 25.85) when compared to manual segmentations. The developed methods were able to automatically detect and track anastomotic structures within EUS sequences without user interaction. The proposed methods have the potential to extract the vessel lumen area from EUS sequences to quantify the stenotic rates of CABG anastomoses.