Multimodal Prediction of Five-Year Breast Cancer Recurrence in Women Who Receive Neoadjuvant Chemotherapy.

In current clinical practice, it is difficult to predict whether a patient receiving neoadjuvant chemotherapy (NAC) for breast cancer is likely to encounter recurrence after treatment and have the cancer recur locally in the breast or in other areas of the body. We explore the use of clinical history, immunohistochemical markers, and multiparametric magnetic resonance imaging (DCE, ADC, Dixon) to predict the risk of post-treatment recurrence within five years. We performed a retrospective study on a cohort of 1738 patients from Institut Curie and analyzed the data using classical machine learning, image processing, and deep learning. Our results demonstrate the ability to predict recurrence prior to NAC treatment initiation using each modality alone, and the possible improvement achieved by combining the modalities. When evaluated on holdout data, the multimodal model achieved an AUC of 0.75 (CI: 0.70, 0.80) and 0.57 specificity at 0.90 sensitivity. We then stratified the data based on known prognostic biomarkers. We found that our models can provide accurate recurrence predictions (AUC > 0.89) for specific groups of women under 50 years old with poor prognoses. A version of our method won second place at the BMMR2 Challenge, with a very small margin from being first, and was a standout from the other challenge entries.

Development of a Non-invasive Device for Swallow Screening in Patients at Risk of Oropharyngeal Dysphagia: Results from a Prospective Exploratory Study.

Oropharyngeal dysphagia is prevalent in several at-risk populations, including post-stroke patients, patients in intensive care and the elderly. Dysphagia contributes to longer hospital stays and poor outcomes, including pneumonia. Early identification of dysphagia is recommended as part of the evaluation of at-risk patients, but available bedside screening tools perform inconsistently. In this study, we developed algorithms to detect swallowing impairment using a novel accelerometer-based dysphagia detection system (DDS). A sample of 344 individuals was enrolled across seven sites in the United States. Dual-axis accelerometry signals were collected prospectively with simultaneous videofluoroscopy (VFSS) during swallows of liquid barium stimuli in thin, mildly, moderately and extremely thick consistencies. Signal processing classifiers were trained using linear discriminant analysis and 10,000 random training-test data splits. The primary objective was to develop an algorithm to detect impaired swallowing safety with thin liquids with an area under receiver operating characteristic curve (AUC) > 80% compared to the VFSS reference standard. Impaired swallowing safety was identified in 7.2% of the thin liquid boluses collected. At least one unsafe thin liquid bolus was found in 19.7% of participants, but participants did not exhibit impaired safety consistently. The DDS classifier algorithms identified participants with impaired thin liquid swallowing safety with a mean AUC of 81.5%, (sensitivity 90.4%, specificity 60.0%). Thicker consistencies were effective for reducing the frequency of penetration-aspiration. This DDS reached targeted performance goals in detecting impaired swallowing safety with thin liquids. Simultaneous measures by DDS and VFSS, as performed here, will be used for future validation studies.

Improved classification of Alzheimer's disease data via removal of nuisance variability.

Diagnosis of Alzheimer's disease is based on the results of neuropsychological tests and available supporting biomarkers such as the results of imaging studies. The results of the tests and the values of biomarkers are dependent on the nuisance features, such as age and gender. In order to improve diagnostic power, the effects of the nuisance features have to be removed from the data. In this paper, four types of interactions between classification features and nuisance features were identified. Three methods were tested to remove these interactions from the classification data. In stratified analysis, a homogeneous subgroup was generated from a training set. Data correction method utilized linear regression model to remove the effects of nuisance features from data. The third method was a combination of these two methods. The methods were tested using all the baseline data from the Alzheimer's Disease Neuroimaging Initiative database in two classification studies: classifying control subjects from Alzheimer's disease patients and discriminating stable and progressive mild cognitive impairment subjects. The results show that both stratified analysis and data correction are able to statistically significantly improve the classification accuracy of several neuropsychological tests and imaging biomarkers. The improvements were especially large for the classification of stable and progressive mild cognitive impairment subjects, where the best improvements observed were 6% units. The data correction method gave better results for imaging biomarkers, whereas stratified analysis worked well with the neuropsychological tests. In conclusion, the study shows that the excess variability caused by nuisance features should be removed from the data to improve the classification accuracy, and therefore, the reliability of diagnosis making.

Error-corrected estimation of regional kinetic parameter histograms directly from pet projections.

Positron emission tomography (PET) is a unique method to investigate physiology in the living body. Kinetic models with kinetic rate constants describe the dynamic radioactive tracer uptake in living tissue. If the variation of the kinetic parameter values within a specific tissue region could be determined accurately, it would give valuable quantitative information about the tissue heterogeneity. In this study we developed a unique method to estimate the variation from the regional kinetic parameter histograms. To determine the kinetic parameter values, we chose non-penalized maximum likelihood (ML) estimation due to the specific statistical error properties of the ML estimates. The parameter values were estimated directly from the time series of PET projections. The choice of the estimation method enabled us to utilize the ML theory in error correction. We developed a Monte Carlo approach to determine the regional error distributions. The true variation of the kinetic parameters could then be revealed by correcting the regional ML estimate histograms with the estimated error distributions. The method was tested with simulated data. In simulations both the average and the deviation of the kinetic parameters were determined from the error-corrected histograms with good numerical accuracy for the selected region of interest.

Automatic FRAP analysis with inhomogeneous fluorescence distribution and movement compensation.

The analysis of fluorescence recovery after photobleaching (FRAP) data is complicated by the measurement noise, inhomogeneous fluorescence distribution, and image movement during experiment. Conventionally, these issues are tackled by data preprocessing and averaging, which causes loss of quantitative properties. In this study, we present a method which automatically estimates and compensates both the movement and inhomogeneous fluorescence distribution within the data analysis. The method is based on modeling the raw FRAP data with a parametric matrix and searching for maximum likelihood parameters between the model and the data. The developed method also automatically estimates also the bleach profile, immobile fraction, and noise variance. Suitable numerical computational method was developed and implemented in a computer grid. Simulated and experimental FRAP data was created and analyzed to evaluate the method.

Cholesterol substitution increases the structural heterogeneity of caveolae.

Caveolin-1 binds cholesterol and caveola formation involves caveolin-1 oligomerization and cholesterol association. The role of cholesterol in caveolae has so far been addressed by methods that compromise membrane integrity and abolish caveolar invaginations. To study the importance of sterol specificity for the structure and function of caveolae, we replaced cholesterol in mammalian cells with its immediate precursor desmosterol by inhibiting 24-dehydrocholesterol reductase. Desmosterol could substitute for cholesterol in maintaining cell growth, membrane integrity, and preserving caveolar invaginations. However, in desmosterol cells the affinity of caveolin-1 for sterol and the stability of caveolin oligomers were decreased. Moreover, caveolar invaginations became more heterogeneous in dimensions and in the number of caveolin-1 molecules per caveola. Despite the altered caveolar structure, caveolar ligand uptake was only moderately inhibited. We found that in desmosterol cells, Src kinase phosphorylated Cav1 at Tyr(14) more avidly than in cholesterol cells. Taken the role of Cav1 Tyr(14) phosphorylation in caveolar endocytosis, this may help to preserve caveolar uptake in desmosterol cells. We conclude that a sterol C24 double bond interferes with caveolin-sterol interaction and perturbs caveolar morphology but facilitates Cav1 Src phosphorylation and allows caveolar endocytosis. More generally, substitution of cholesterol by a structurally closely related sterol provides a method to selectively modify membrane protein-sterol affinity, structure and function of cholesterol-dependent domains without compromising membrane integrity.