Subject2Vec: Generative-Discriminative Approach from a Set of Image Patches to a Vector.

We propose an attention-based method that aggregates local image features to a subject-level representation for predicting disease severity. In contrast to classical deep learning that requires a fixed dimensional input, our method operates on a set of image patches; hence it can accommodate variable length input image without image resizing. The model learns a clinically interpretable subject-level representation that is reflective of the disease severity. Our model consists of three mutually dependent modules which regulate each other: (1) a discriminative network that learns a fixed-length representation from local features and maps them to disease severity; (2) an attention mechanism that provides interpretability by focusing on the areas of the anatomy that contribute the most to the prediction task; and (3) a generative network that encourages the diversity of the local latent features. The generative term ensures that the attention weights are non-degenerate while maintaining the relevance of the local regions to the disease severity. We train our model end-to-end in the context of a large-scale lung CT study of Chronic Obstructive Pulmonary Disease (COPD). Our model gives state-of-the art performance in predicting clinical measures of severity for COPD.The distribution of the attention provides the regional relevance of lung tissue to the clinical measurements.

A Likelihood-Free Approach for Characterizing Heterogeneous Diseases in Large-Scale Studies.

We propose a non-parametric approach for characterizing heterogeneous diseases in large-scale studies. We target diseases where multiple types of pathology present simultaneously in each subject and a more severe disease manifests as a higher level of tissue destruction. For each subject, we model the collection of local image descriptors as samples generated by an unknown subject-specific probability density. Instead of approximating the probability density via a parametric family, we propose to side step the parametric inference by directly estimating the divergence between subject densities. Our method maps the collection of local image descriptors to a signature vector that is used to predict a clinical measurement. We are able to interpret the prediction of the clinical variable in the population and individual levels by carefully studying the divergences. We illustrate an application this method on simulated data as well as on a large-scale lung CT study of Chronic Obstructive Pulmonary Disease (COPD). Our approach outperforms classical methods on both simulated and COPD data and demonstrates the state-of-the-art prediction on an important physiologic measure of airflow (the forced respiratory volume in one second, FEV1).

Spherical Topic Models for Imaging Phenotype Discovery in Genetic Studies.

In this paper, we use Spherical Topic Models to discover the latent structure of lung disease. This method can be widely employed when a measurement for each subject is provided as a normalized histogram of relevant features. In this paper, the resulting descriptors are used as phenotypes to identify genetic markers associated with the Chronic Obstructive Pulmonary Disease (COPD). Features extracted from images capture the heterogeneity of the disease and therefore promise to improve detection of relevant genetic variants in Genome Wide Association Studies (GWAS). Our generative model is based on normalized histograms of image intensity of each subject and it can be readily extended to other forms of features as long as they are provided as normalized histograms. The resulting algorithm represents the intensity distribution as a combination of meaningful latent factors and mixing co-efficients that can be used for genetic association analysis. This approach is motivated by a clinical hypothesis that COPD symptoms are caused by multiple coexisting disease processes. Our experiments show that the new features enhance the previously detected signal on chromosome 15 with respect to standard respiratory and imaging measurements.

Prediction of MCI to AD conversion, via MRI, CSF biomarkers, and pattern classification.

Magnetic resonance imaging (MRI) patterns were examined together with cerebrospinal fluid (CSF) biomarkers in serial scans of Alzheimer's Disease Neuroimaging Initiative (ADNI) participants with mild cognitive impairment (MCI). The SPARE-AD score, summarizing brain atrophy patterns, was tested as a predictor of short-term conversion to Alzheimer's disease (AD). MCI individuals that converted to AD (MCI-C) had mostly positive baseline SPARE-AD (Spatial Pattern of Abnormalities for Recognition of Early AD) and atrophy in temporal lobe gray matter (GM) and white matter (WM), posterior cingulate/precuneous, and insula. MCI individuals that converted to AD had mostly AD-like baseline CSF biomarkers. MCI nonconverters (MCI-NC) had mixed baseline SPARE-AD and CSF values, suggesting that some MCI-NC subjects may later convert. Those MCI-NC with most negative baseline SPARE-AD scores (normal brain structure) had significantly higher baseline Mini Mental State Examination (MMSE) scores (28.67) than others, and relatively low annual rate of Mini Mental State Examination decrease (-0.25). MCI-NC with midlevel baseline SPARE-AD displayed faster annual rates of SPARE-AD increase (indicating progressing atrophy). SPARE-AD and CSF combination improved prediction over individual values. In summary, both SPARE-AD and CSF biomarkers showed high baseline sensitivity, however, many MCI-NC had abnormal baseline SPARE-AD and CSF biomarkers. Longer follow-up will elucidate the specificity of baseline measurements.

DISEASE CLASSIFICATION AND PREDICTION VIA SEMI-SUPERVISED DIMENSIONALITY REDUCTION.

We present a new semi-supervised algorithm for dimensionality reduction which exploits information of unlabeled data in order to improve the accuracy of image-based disease classification based on medical images. We perform dimensionality reduction by adopting the formalism of constrained matrix decomposition of [1] to semi-supervised learning. In addition, we add a new regularization term to the objective function to better captur the affinity between labeled and unlabeled data. We apply our method to a data set consisting of medical scans of subjects classified as Normal Control (CN) and Alzheimer (AD). The unlabeled data are scans of subjects diagnosed with Mild Cognitive Impairment (MCI), which are at high risk to develop AD in the future. We measure the accuracy of our algorithm in classifying scans as AD and NC. In addition, we use the classifier to predict which subjects with MCI will converge to AD and compare those results to the diagnosis given at later follow ups. The experiments highlight that unlabeled data greatly improves the accuracy of our classifier.

Anomalous origin and retropulmonary course of an atherosclerotic stenosed left circumflex coronary artery.

We here present the case of a rarely seen anomalous origin and retropulmonary course of the left circumflex artery from the proximal right coronary artery. The patient suffered from coronary ischemia due to stenotic lesions both in the aberrant circumflex coronary artery and in the first and second diagonal branches. Coronary bypass operation was performed.

Reevaluation of two techniques of harvesting the skeletonized internal mammary artery.

BACKGROUND: Selective skeletonization of the internal mammary artery (IMA) without adjacent vasculo-muscular structures reduces trauma to the chest wall, results in elongated grafts, makes ideal graft positioning possible, and eliminates the need to implant a dissected or hypoplastic graft with direct visual control of the vessel. We compared two techniques of skeletonizing the IMA in a prospective randomized trial. METHODS: 51 IMAs were randomly harvested and divided into two groups according to the technique of skeletonization. In group I (n = 31), IMAs were harvested in a skeletonized fashion with the Harmonic Ultrasonic scalpel, and in group II (n = 20) using scissors and hemostatic clips. We compared arterial wall histology, harvesting time, spasm frequency, and the use of hemostatic clips between the two groups. RESULTS: There were no significant morphological differences in the arterial wall in the two groups. Use of an ultrasonically-activated scalpel reduced the IMA's harvesting time (p < 0.001), the frequency of spasm (p = 0.01), and the use of hemostatic clips (p < 0.001). CONCLUSIONS: Ultrasonic harvesting of a skeletonized IMA is a non-traumatic preparatory technique that reduces the costs of surgical clips and that can be performed safely and quickly.

Bridge to operation with the GPIIb/IIIa inhibitor abciximab in high-risk coronary patients.

BACKGROUND: Glycoprotein-IIb/IIIa inhibitors are now frequently used in the cardiological treatment of high-risk coronary patients even if the patient is considered suitable for surgical intervention. However, there is no consensus whether GPIIb/IIIa inhibitors should be stopped before operation because of an increased risk of bleeding or if surgery should even be delayed until the anticoagulating effect subsides. METHODS: From June 2002 to August 2003 140 patients who had to undergo primary aorto-coronary bypass for ongoing myocardial ischemia were enrolled in the present study. The patients received either clopidogrel, aspirin and heparin or additionally abciximab until operation. RESULTS: Although the intraoperative need for blood products was higher in the abciximab group, there was no significant difference in postoperative blood loss. The hemodynamic situation of the abciximab patients after the operation was better compared to the other groups. 30-day mortality was not increased when compared to the elective control group (6.7 % vs. 6.1 %). CONCLUSION: The GPIIb/IIIa inhibitor abciximab can be safely used as a bridge to operation and results in a better hemodynamic outcome in high-risk coronary patients while reducing the incidence of major ischemic events.

[Computer-assisted reconstruction of coronary vessels]. / Rechnerunterstützte Rekonstruktion von Herzkranzgefässen.

End stage coronary artery disease with linear stenosis of the main vessels despite several coronary interventions is a current challenge for surgical treatment. As the long term results are mainly determined by the pathology of the coronary vessels a simple revascularisation with arterial or venous grafts provides no adequate solution of the problem. An exactly controlled 3D-reconstruction of the coronary vessels enabling selective thrombendarteriectomy (TEA) seems to be a new approach, that may be limited by neointimal hyperproliferation of the coronary vessels. Intraoperative brachytherapy may be a tool to inhibit this process.

Effects of trifluoperazine on the contraction kinetics of the isolated intact tracheal and pulmonary artery smooth muscle.

BACKGROUND: We studied the effect of the calmodulin antagonist trifluoperazine (TFP) on isolated intact rat tracheal and pulmonary artery smooth muscle contractile behaviour. METHODS: Experimental series: 1) TFP-dose-response curves for TFP's effect on force generation were constructed using rat tracheal smooth muscles and rat pulmonary artery preparations (n = 8). A concentration of 1 micromol/l TFP was chosen for the subsequent experimental series. 2) Tracheas and pulmonary arteries (n = 14) were dissected in three segments. One of them was used immediately for experiments ("native"), the other two were treated for 12 h in 4 degrees C Tyrode solution without ("12 h cold storage") or with 1 micromol/l TFP ("12 h cold storage + TFP"). These preparations contracted after supramaximal effective electrical field stimulation. The force-clamping technique was used to analyse kinetic and mechanical parameters of smooth-muscle contraction in both types of preparation (measurement conditions: resting tension 2 mN, 37 degrees C, modified Krebs-Henseleit solution). RESULTS: 1) TFP decreased developed force dose-dependently in pulmonary artery and tracheal smooth muscle. 2) During sustained tonic activation, the contraction kinetics become slower both with and without TFP treatment (p < 0.0001). 3) TFP caused a dramatic retardation of the velocity of force generation in both types of preparation for any given time interval during the course of a tonic activation (p < 0.0005). 4) The dramatic effects of TFP on the contraction kinetics were not associated with effects on the extent of force generation. CONCLUSIONS: These results support the assumption that tracheal and pulmonary artery smooth muscle cross-bridge rates are controlled by a calcium-calmodulin-dependent myosin light chain kinase. This finding suggests the involvement of a calmodulin-independent regulator process responsible for the changes observed in the cross-bridge cycling rates during sustained tonic activation. A direct intervention on the contractile apparatus level is a measure for reduction of smooth-muscle tone without negative inotropic side effects.

[Myocardial revascularization after pretreatment with the GPIIb/IIIa receptor blocker abciximab]. / Myocardrevaskularisation nach Vorbehandlung mit dem GPIIb/IIIa-Rezeptorblocker Abciximab.

AIM: The present prospective study analysis whether preoperative abciximab treatment 1) increases perioperative risks, 2) alters the quality of the surgical treatment, 3) whether secondary organ complications occur and 4) whether and how the perioperative surgical management has to be modified. METHODS: The perioperative results of patients operated upon after preoperative abciximab treatment (n = 31, 1/1994-4/2000) were compared with an optimal control group (elective patients and no preoperative intervention in the hemostaseological system) and with other patients (elective, urgent, emergency operations) receiving different strategies of preoperative hemostaseological interventions (n = 926, 1/1994-4/2000). All patients were operated upon with a well-defined strategy (one surgeon). 1100 perioperative items/patient were prospectively recorded and analyzed. The 30-day and 6 month follow-up was complete. RESULTS: The mean time interval between end of abciximab therapy and onset of operation was 4.3 h (range: 36-0.5 h, median: 2 h). As compared with the optimal control group (n = 81), perioperative blood loss and the need for substitution of blood and blood products was increased. However, as compared with those patients receiving any type of preoperative intervention in the hemostaseological system (the majority of the patients), no difference was observed regarding intraoperative parameters (duration of extracorporeal circulation and of the operation) and postoperative results (neurological, renal complications, myocardial infarction, psychiatric disturbances, etc.) when abciximab patients were compared with the reference populations. No patient of the abciximab group died within 30 days or during the 6 month follow-up period. CONCLUSION: The data show that coronary artery bypass grafting can be performed with similar results in abciximab-pretreated patients as in patients with other types of preoperative anticoagulation. Only slight modifications of the surgical management are recommended. As indicated by similar use IMA-grafts the quality of surgical treatment was not reduced.

[Preoperative administration of new GPIIb/IIIa inhibitors]. / Präoperativer Einsatz neuer GPIIb/IIIa Inhibitoren.

The influence of preoperative abciximab treatment on perioperative risks was analyzed. Six groups were formed in consecutive patients (n = 957, 1/1994-3/2000) operated upon for CABG according to the urgency of the intervention and the type of preoperative haemostaseological treatment. Thirty-one patients underwent urgent/emergency CABG with preceding abciximab treatment. After pretreatment with abciximab the surgical results did not differ from patients with other types of preoperative haemostaseological treatment. CABG can be performed with similar results in abciximab pretreated patients as in patients with other types of preoperative haemostaseological treatment.