On the feasibility of Vis-NIR spectroscopy and machine learning for real time SARS-CoV-2 detection.

The pandemic caused by Covid-19 is still present around the world. Despite advances in combating the disease, such as vaccine development, identifying infected individuals is still essential to optimize the control of human-to-human transmission of the virus. The main technique for detecting the virus is the RT-PCR method, which, despite its high relative cost, has a high accuracy in detecting the coronavirus. Given this, a method capable of performing the identification quickly, accurately, and inexpensively is necessary. Thus, this work aimed to analyze the feasibility of a new technique for identifying SARS-CoV-2 through the use of optical spectroscopy in the visible and near-infrared range (Vis-NIR) combined with machine learning algorithms. Spectral signals were obtained from nasopharyngeal swab samples previously analyzed using the RT-PCR method. The specimens were provided by the Molecular Diagnosis Laboratory of Covid-19 at Univasf. A total of 314 samples were analyzed, comprising 42 testing positive and 272 testing negative for Covid-19. Digital signal processing techniques, such as Savitzky-Golay filters and statistical methods were used to eliminate spurious elements from the original data and extract relevant features. Supervised machine learning algorithms such as SVM, Random Forest, and Naive Bayes classifiers were used to perform automatic sample identification. To evaluate the performance of the models, a 5-fold cross-validation technique was applied. With the proposed methodology, it was possible to achieve an accuracy of 75%, a sensitivity of 80%, and a specificity of 70%, in addition to an area under the ROC curve of 0.81, in the identification of nasopharyngeal swab samples from previously diagnosed individuals. From these results, it was possible to conclude that Vis-NIR spectroscopy is a promising, fast and relatively low cost technique to identify the SARS-CoV-2.

Prediction of dyskinesia in Parkinson's disease patients using machine learning algorithms.

Dyskinesias are non preventable abnormal involuntary movements that represent the main challenge of the long term treatment of Parkinson's disease (PD) with the gold standard dopamine precursor levodopa. Applying machine learning techniques on the data extracted from the Parkinson's Progression Marker Initiative (PPMI, Michael J. Fox Foundation), this study was aimed to identify PD patients who are at high risk of developing dyskinesias. Data regarding clinical, behavioral and neurological features from 697 PD patients were included in our study. Our results show that the Random Forest was the classifier with the best and most consistent performance, reaching an area under the receiver operating characteristic (ROC) curve of up to 91.8% with only seven features. Information regarding the severity of the symptoms, the semantic verbal fluency, and the levodopa treatment were the most important for the prediction, and were further used to create a Decision Tree, whose rules may guide the pharmacological management of PD symptoms. Our results contribute to the identification of PD patients who are prone to develop dyskinesia, and may be considered in future clinical trials aiming at developing new therapeutic approaches for PD.

Automatic segmentation of brain tumors in magnetic resonance imaging.

OBJECTIVE: To develop a computational algorithm applied to magnetic resonance imaging for automatic segmentation of brain tumors. METHODS: A total of 130 magnetic resonance images were used in the T1c, T2 and FSPRG T1C sequences and in the axial, sagittal and coronal planes of patients with brain cancer. The algorithms employed contrast correction, histogram normalization and binarization techniques to disconnect adjacent structures from the brain and enhance the region of interest. Automatic segmentation was performed through detection by coordinates and arithmetic mean of the area. Morphological operators were used to eliminate undesirable elements and reconstruct the shape and texture of the tumor. The results were compared with manual segmentations by two radiologists to determine the efficacy of the algorithms implemented. RESULTS: The correlated correspondence between the segmentation obtained and the gold standard was 89.23%. CONCLUSION: It is possible to locate and define the tumor region automatically with no the need for user interaction, based on two innovative methods to detect brain extreme sites and exclude non-tumor tissues on magnetic resonance images.

Automatic segmentation of brain tumors in magnetic resonance imaging / Segmentação automática de tumores cerebrais em imagens de ressonância magnética

ABSTRACT Objective To develop a computational algorithm applied to magnetic resonance imaging for automatic segmentation of brain tumors. Methods A total of 130 magnetic resonance images were used in the T1c, T2 and FSPRG T1C sequences and in the axial, sagittal and coronal planes of patients with brain cancer. The algorithms employed contrast correction, histogram normalization and binarization techniques to disconnect adjacent structures from the brain and enhance the region of interest. Automatic segmentation was performed through detection by coordinates and arithmetic mean of the area. Morphological operators were used to eliminate undesirable elements and reconstruct the shape and texture of the tumor. The results were compared with manual segmentations by two radiologists to determine the efficacy of the algorithms implemented. Results The correlated correspondence between the segmentation obtained and the gold standard was 89.23%. Conclusion It is possible to locate and define the tumor region automatically with no the need for user interaction, based on two innovative methods to detect brain extreme sites and exclude non-tumor tissues on magnetic resonance images.

RESUMO Objetivo Desenvolver um algoritmo computacional aplicado a imagens de ressonância magnética, para segmentação automática de tumores cerebrais. Métodos Foram utilizadas 130 imagens de ressonância magnética nas sequências T1c, T2 e FSPRG T1c e nos planos axial, sagital e coronal de pacientes acometidos com câncer cerebral. Os algoritmos empregaram técnicas de correção de contraste, normalização de histograma e binarização, para desconectar estruturas adjacentes do cérebro e realçar a região de interesse. A segmentação automática foi realizada por meio da detecção por coordenadas e por média aritmética da área. Operadores morfológicos foram utilizados para eliminar elementos indesejáveis e reconstruir a forma e a textura do tumor. Os resultados foram comparados com as segmentações manuais de dois médicos radiologistas, para determinar a eficácia dos algoritmos implementados. Resultados Os acertos foram de 89,23% na correspondência entre a segmentação obtida e o padrão-ouro. Conclusão É possível localizar e delimitar a região tumoral de forma automática, sem necessidade de interação com o usuário baseado em dois métodos inovadores de detecção dos extremos do cérebro e de exclusão dos tecidos não tumorais em imagens de ressonância magnética.

Segmentation and detection of breast cancer in mammograms combining wavelet analysis and genetic algorithm.

In Brazil, the National Cancer Institute (INCA) reports more than 50,000 new cases of the disease, with risk of 51 cases per 100,000 women. Radiographic images obtained from mammography equipments are one of the most frequently used techniques for helping in early diagnosis. Due to factors related to cost and professional experience, in the last two decades computer systems to support detection (Computer-Aided Detection - CADe) and diagnosis (Computer-Aided Diagnosis - CADx) have been developed in order to assist experts in detection of abnormalities in their initial stages. Despite the large number of researches on CADe and CADx systems, there is still a need for improved computerized methods. Nowadays, there is a growing concern with the sensitivity and reliability of abnormalities diagnosis in both views of breast mammographic images, namely cranio-caudal (CC) and medio-lateral oblique (MLO). This paper presents a set of computational tools to aid segmentation and detection of mammograms that contained mass or masses in CC and MLO views. An artifact removal algorithm is first implemented followed by an image denoising and gray-level enhancement method based on wavelet transform and Wiener filter. Finally, a method for detection and segmentation of masses using multiple thresholding, wavelet transform and genetic algorithm is employed in mammograms which were randomly selected from the Digital Database for Screening Mammography (DDSM). The developed computer method was quantitatively evaluated using the area overlap metric (AOM). The mean ± standard deviation value of AOM for the proposed method was 79.2 ± 8%. The experiments demonstrate that the proposed method has a strong potential to be used as the basis for mammogram mass segmentation in CC and MLO views. Another important aspect is that the method overcomes the limitation of analyzing only CC and MLO views.