A simple graphical quantitative analysis of ultrasonography images to decide when to perform fine needle aspiration biopsy in diagnosing malignancy in solid thyroid nodules? A two centres prospective study.

OBJECTIVE: Since the prevalence of thyroid nodules is high and ultrasonography (USG)-guided fine-needle aspiration biopsy (FNAB) as a diagnostic means cannot be performed in all cases, we aimed to evaluate the feasibility and applicability of simple graphical analysis of USG two-dimensional images, to identify patients with suspicious thyroid nodules who would benefit from FNAB. SUBJECTS AND METHODS: We studied prospectively 211 consecutive patients with thyroid nodules: 122 from the University Clinical Centre (UCC) of Maribor and 89 from the University Medical Centre (UMC) of Ljubljana who underwent USGguided FNAB from January 2011 to October 2013. The cytology report was categorized as benign or suspicious/malignant. Blind to cytology reports, we later performed graphical analysis of USG images using ImageJ (version 1.48r) which is a public domain Java image processing and analysis programme. We compared the average gray value and standard deviation (SD) of the gray values used to generate the mean gray within the selection, with cytology reports. RESULTS: According to cytology reports, 24 thyroid nodules were suspicious/malignant (14/10) and 187 benign. Graphical analysis of USG images performed with ImageJ demonstrated significantly higher values of SD of the gray values used to generate the mean gray value in suspicious/malignant thyroid nodules as compared to unsuspicious nodules in both UCC Maribor and in UMC Ljubljana (P<0.001 and P=0.002, respectively). A higher value of the SD of gray value used to generate the mean gray value meant variation or dispersion from the average value and was correlated by the presence of micro-calcifications. By applying a cut-off level of the quotient between the SD value of an examined thyroid nodule and the SD value of normal/reference thyroid tissue of 1.20, we found that 21/24 nodules were classified as true positive and 114/187 as true negative. CONCLUSION: Our results showed that our graphical quantitative analysis of USG images had a negative predictive value of more than 90% and was able to suggest which thyroid nodules were potentially malignant and needed further investigation.

Interleukins and inflammatory markers are useful in predicting the severity of acute pancreatitis.

Acute pancreatitis (AP) is a disease with significant morbidity and mortality. The aim of this study was to evaluate the predictive role of inflammatory markers, particularly interleukins (ILs), in the course of AP and to determine the frequency of etiologic factors of AP. We included patients with AP who were treated at our institution from May 1, 2012 to January 31, 2015. Different laboratory parameters, including ILs, and the severity scoring systems Ranson's criteria and Bedside Index of Severity in Acute Pancreatitis (BISAP) were analyzed. AP was classified into mild and severe, and independent parameters were compared between these groups. The predictive performance of each parameter was evaluated using receiver operating characteristic (ROC) curves and the area under the ROC curve (AUC). A binomial logistic regression was performed to evaluate Ranson's criteria and IL6, IL8, and IL10 (at admission and after 48 hours) in the course of AP. Overall, 96 patients were treated, 59 (61.5%) males and 37 (38.5%) females, average age 62.5 ± 16.8 years (range 22-91 years). The best predictor for the severity of AP was IL6, measured 48 hours after admission (AUC = 0.84). Other useful predictors of the severity of AP were lactate dehydrogenase (p < 0.001), serum glucose (p < 0.006), and difference in the platelet count (p < 0.001) between admission and after 48 hours (p < 0.001), hemoglobin (p < 0.027) and erythrocytes (p < 0.029). The major causes of AP were gallstones and alcohol consumption. According to our results, IL6 and Ranson score are important predictors of the severity of AP.

Explanatory approach for evaluation of machine learning-induced knowledge.

Progress in biomedical research has resulted in an explosive growth of data. Use of the world wide web for sharing data has opened up possibilities for exhaustive data mining analysis. Symbolic machine learning approaches used in data mining, especially ensemble approaches, produce large sets of patterns that need to be evaluated. Manual evaluation of all patterns by a human expert is almost impossible. We propose a new approach to the evaluation of machine learning-induced knowledge by introducing a pre-evaluation step. Pre-evaluation is the automatic evaluation of patterns obtained from the data mining phase, using text mining techniques and sentiment analysis. It is used as a filter for patterns according to the support found in online resources, such as publicly-available repositories of scientific papers and reports related to the problem. The domain expert can then more easily distinguish between patterns or rules that are potential candidates for new knowledge.

Classification of follicular lymphoma images: a holistic approach with symbol-based machine learning methods.

It is not very often to see a symbol-based machine learning approach to be used for the purpose of image classification and recognition. In this paper we will present such an approach, which we first used on the follicular lymphoma images. Lymphoma is a broad term encompassing a variety of cancers of the lymphatic system. Lymphoma is differentiated by the type of cell that multiplies and how the cancer presents itself. It is very important to get an exact diagnosis regarding lymphoma and to determine the treatments that will be most effective for the patient's condition. Our work was focused on the identification of lymphomas by finding follicles in microscopy images provided by the Laboratory of Pathology in the University Hospital of Tenerife, Spain. We divided our work in two stages: in the first stage we did image pre-processing and feature extraction, and in the second stage we used different symbolic machine learning approaches for pixel classification. Symbolic machine learning approaches are often neglected when looking for image analysis tools. They are not only known for a very appropriate knowledge representation, but also claimed to lack computational power. The results we got are very promising and show that symbolic approaches can be successful in image analysis applications.

Does size really matter--using a decision tree approach for comparison of three different databases from the medical field of acute appendicitis.

Decision trees have been successfully used for years in many medical decision making applications. Transparent representation of acquired knowledge and fast algorithms made decision trees one of the most often used symbolic machine learning approaches. This paper concentrates on the problem of separating acute appendicitis, which is a special problem of acute abdominal pain, from other diseases that cause acute abdominal pain by use of an decision tree approach. Early and accurate diagnosing of acute appendicitis is still a difficult and challenging problem in everyday clinical routine. An important factor in the error rate is poor discrimination between acute appendicitis and other diseases that cause acute abdominal pain. This error rate is still high, despite considerable improvements in history-taking and clinical examination, computer-aided decision-support, and special investigation such as ultrasound. We investigated three databases of different size with cases of acute abdominal pain to complete this task as successful as possible. The results show that the size of the database does not necessary directly influence the success of the decision tree built on it. Surprisingly we got the best results from the decision trees built on the smallest and the biggest database, where the database with medium size (relative to the other two) was not so successful. Despite this we were able to produce decision tree classifiers that were capable of producing correct decisions on test data sets with accuracy up to 84%, sensitivity to acute appendicitis up to 90%, and specificity up to 80% on the same test set.