In this paper, we present a method for preprocessing images from laryngeal high-speed videoendoscopy (LHSV). We developed image processing procedures to better prepare the images for automated computer analysis. Namely, we removed shifts between consecutive LHSV images and glares distorting the images, detected the region of interest, rotated the images, and finally detected vocal fold regions. We tested the developed algorithms on 13 LHSV recordings made for healthy patients and present example results of the conducted study.
The temporomandibular joint (TMJ) is an even joint in the human head that allows for a movable connection between the skull and the mandible and performs complex movements. According to statistical data, it is estimated that disorders within the masticatory apparatus affect 60 to 80
Prediction models that rely on time series data are often affected by diminished predictive accuracy. This occurs from the causal relationships of the data that shift over time. Thus, the changing weights that are used to create prediction models lose their informational value. One way to correct this change is by using concept drift information. That is exactly what prediction models in biomedical applications need. Currently, metabolomics is at the forefront in modeling analysis for phenotype prediction, making it one of the most interesting candidates for biomedical prediction diagnosis. However, metabolomics datasets include dynamic information that can harm prediction modeling. The study presents concept drift correction methods to account for dynamic changes that occur in metabolomics data for better prediction outcomes of phenotypes in a biomedical setting.
The paper presents a new approach to preventive examinations of employees in relation to the low reliability of the appraisals made due to experiencing stress that does not allow employees to fully present their potential. This approach is illustrated using the example of a case study of a voice professional. The study provides a detailed, quantitative view of the emotions experienced, revealed during the voice recording and audiometry procedure. Differences were found as a result of the study in terms of the intensity of emotions. EDA and HRV measurement values were highest when questionnaires were being completed concerning the emotions experienced during voice recording and audiometry. The discussion focuses on the possibilities of analysing emotions using psychophysiological measurements and on the benefits of combining research methods (physical examinations, psychological examinations, and psychophysiological measurements) in the context of employee appraisal to predict the effects of the impact of work on subjective well-being and health of individuals.
The aim of the article is to compare the morphological features of language between people suffering from anorexia and healthy people. The research was conducted in cooperation with the Medical University of Silesia on the basis of a pilot group of people: 41 from anorexia patients and 55 from healthy young females. The study focuses on the statistics for the following parts of speech: personal pronoun ‘I’, possessive adjective ‘my’ verbs, adjectives. Several significant differences were detected, including abnormal usage of the personal pronoun ‘I’ and mostly verb ‘to be’, and action verbs by anorectic people. In the future, it is planned to conduct the research on a larger group of patients and include the descriptions of the drawings in the re-search.
The aim of this research was to find the differentially expressed genes in the groups of patients with different stages of bladder cancer. Proper analysis could help to find biomarkers responsible for the occurrence of the most invasive forms of bladder cancer. The microarray data (series GSE31684) was used and the obtained genes were characterized and described. The data from 93 bladder cancer patients managed by radical cystectomy was used. The research also examined the impact of various parameters on survival after surgery. The results of the analysis showed that some genes can allow to diagnose various types of cancer grade and stage. The research also presented how smoking, lymph node metastases, cancer stage and grade can affect survival.
Breast cancer has the highest prevalence in women globally. The classification of breast cancer histopathological images has always been a hot spot of clinical concern. In Computer-Aided Diagnosis (CAD), traditional models mostly use a single network to extract features, which has significant limitations. Besides, many networks are trained and optimized on patient-level datasets, ignoring the lower-level labels. This paper proposes a deep ensemble model based on image-level labels for the binary classification of benign and malignant breast histopathological images. First, the BreakHis dataset is randomly divided into training, validation and test sets. Then, data augmentation techniques are used to balance the number of benign and malignant samples. Thirdly, VGG-16, Xception, Resnet-50, Densenet-201 are selected as the base classifiers resulting from their complementarity. Finally, with accuracy as the ensemble weight, an accuracy of 98.90% is achieved. In order to verify the capabilities of our method, the latest Transformer and Multilayer Perception (MLP) models have been experimentally compared on the same dataset. Our model wins with a 5–20% advantage, emphasizing the ensemble model’s far-reaching significance in classification tasks.
Artifact detection in the electroencephalogram (EEG) analysis is one of the preprocessing phases that improves the final recognition result. At this research step, a preprocessing function able to recognize selected types of artifacts is developed. Detected artifacts are labeled for further processing. As part of the proprietary concept, the effectiveness of the proposed software solution is compared with the expert outcome. This study is part of an international research project Tele-BRAIN, scientifically oriented at supporting the early diagnosis of Parkinson’s disease, with particular emphasis on cognitive disorders. The obtained results in the form of preprocessed EEG signals are then subjected to the analysis based on artificial intelligence methods, presented as a separate study.
Monitoring of uterine contractions is a routine procedure in obstetrical wards. The new method called an electrohysterography has many advantages in compering to the external tocography. However, there is no clinical standard for measuring and parameterization of the bioelectrical activity of a pregnant uterus. Particularly, factors affecting the electrohysterographical signals (EHG) were not studied. This paper shows that contractions of abdominal muscles can change the commonly used parameters of electrohysterographical signals in the same way like the uterine contractions. Therefore, we postulate to parametrize an envelope of the EHG signals instead of the raw EHG. Moreover, this paper initiates a terminological discusiosn asking whether EHG really measures bioelectrical or biomechanical activities of myometrium.
In this work we propose a method for numerical finding of a function representing the time-dependent virus transmission intensity coefficient in the exemplary SEIR model of infectious disease. Our method is based on gradient minimization of a predefined functional and uses a gradient obtained from adjoint sensitivity analysis. To apply this method to the exemplary SEIR model we used publicly available infection data concerning the COVID-19 cumulative cases in Poland.
The emergence of new information technologies affects various aspects of life. The use of mobile communication devices in healthcare is known as mobile health or mHealth. In our study, we present an app for measuring the heart rate in real time based on seismocardiography. The heartbeats were detected with a modified version of Pan-Tompkins algorithm. The results prove the feasibility of the designed app for real-time measurement of heart rate using only an accelerometer.
Closed-loop controllers for insulin pumps have been on the market for some time. It has been shown that modified PID or MPC control algorithms are best suited for artificial pancreas. However, due to nonnegative control values only and relatively slow dynamics of the response to insulin input, they are not well equipped to deal with hypoglycemia induced by a physical effort. This paper is focused on that aspect of blood glucose control. Two alternative solutions are proposed and compared. The first one is based on feedforward, with additional information about future physical effort entered by the user. The second approach uses an additional control in the form of glucagon. Simulation is run for a fixed scenario of three meals and additional physical effort that affects the insulin-glucose system for a cohort of virtual patients, for whom model parameters were sampled. Performance of control systems is evaluated with several quality indicators.
The focus of this study was to test a contactless photoplethysmography based method to calculate pulse of a patient from a video recording of their face. For this purpose deep convolution neural network was used for detection the region of interest skin area of face and then analyzing of the variability of the image values was processed as a signal in frequency domain for pulse reconstruction. The method was tested on three video sets with different video resolutions: 1920 × 1080 px, 960 × 540 px, and 640 × 580 px. The best results came from a set with a resolution of 960 × 540 px, with a relative error of 10.6
This article presents method for recognizing activity from data acquired from the accelerator, magnetometer, gyroscope and motion sensors. The experiments providing data were conducted in July 2021 in Katowice, Poland, as a part of the System for Monitoring Activity and Training Rationalization (SMART) project, financed by the polish National Centre for Research and Development. A variety of classifiers were tested in two approaches – using all available variables and using features selected with the Joint Mutual Information method. Separate models were built for each activity as well as models for selecting one activity out of 8 possible. The best obtained results exceeded 98
The influence of LIPUS on the properties of the biodegradable polymer coatings containing an active substance on the anodized titanium alloy was investigated. PLGA polymer coatings with ciprofloxacin were applied using the dipping method. The samples were stimulated by ultrasound in Ringer’s solution or were exposed only to solution for 1, 3 or 4 weeks. The influence of ultrasound stimulation was determined by microscopic observation, wettability, metallic ion release and drug release. The polymer coating used in the study is characterized by hydrophilic properties and barrier properties limiting the release of substrate degradation products into solution. Application of LIPUS caused a decrease of coatings’ barrier properties and an increase of hydrophilic properties. However, the results show the usefulness of polymer coatings in bone fracture stabilizers. Moreover, the application of biodegradable polymer coatings enables drug delivery to the bone fracture, which release can be controlled by different parameters of LIPUS therapy.
The selection of the appropriate voice signal recording equipment, including a microphone, and the selection of appropriate conditions for the recording site is crucial for obtaining reliable and authoritative values of acoustic parameters in both medical and biological research. The aim of this study was to compare selected acoustic measures of two microphone types – dynamic and condenser. The study involved 80 adults (including 48 women) for whom the values of voice parameters, i.e. intensity, fundamental and formant frequencies, jitter, shimmer and noise-to-harmonic ratio (NHR) were determined on the basis of 5 vowels recorded simultaneously with both microphones. The existence of significant differences in the values of selected acoustic parameters between the two types of microphones, despite the high correlation coefficient of these measures, was demonstrated. The results of this study may prove important for voice researchers when selecting the appropriate recording equipment.
Diagnostic instruments are nowadays an integral medicine part. Instruments try to make work of doctors easier. The measured data are often poorly understood by the layperson, so it is important that the physician sufficiently explains the information obtained from the biosignal. However, sometimes this information is not necessarily important. When measuring electromyographic signals during rehabilitation or training of athletes, rapid feedback is essential. This paper deals exclusively with the creation prototype of an electromyograph measurement chain with a quick and simple presentation of the electromyographic signal for the layperson. Signal pre-processing is discussed and many presentation variant of electromyographic signal. Such as acoustic output, lighting of LEDs by EMG, visualization of the EMG signal intensity on a cascade of LEDs and playback of the selected sound when the set intensity of the electromyographic signal is exceeded. The device has the ability to adjust the level of difficulty to monitore progress.
The article presents the results of a preliminary study analy-sing the physiological parameters obtained during exercises that teach the patient’s correct body posture while sitting. Electrodermal activity (EDA), blood volume pulse (BVP), and electromyographic (EMG) signals were recorded and analysed during the training process for position shaping. A music preference and musicality questionnaire was carried out before the study. The JAWS questionnaire was completed twice by the respondent, before and after exercises. The physiotherapists provided instructions with respect to the stimulation of the autonomic nervous system, observed in EDA, heart rate and the subsequent motor units. While performing the exercises, the subjects felt positive emotions, which can be perceived as a positive experience for the probands and suggests their willingness to learn and maintain correct body posture while sitting. The sonification of the therapist’s commands and their sonic emotional content is further researched.
Determining and tracking the location of specific points of real objects in space is not an easy task. Nowadays, this task is performed by artificial deep neural networks. Various methods and techniques have been competing with each other in recent years. Most of them do it effectively and with a satisfactory result. The success of such solutions is associated with long-term learning and a large amount of training material. The following article should answer the question whether and how the selected tool affects the results obtained. Several approaches to solving the problem using different technologies are presented. The material was verified for a selected group of photos of children in the first weeks of life.
Pain feeling assessment is crucial for a safe and efficient course of physiotherapy. Especially onset of severe pain stands for specific tissue guard and protects it from damage. In this study, an approach for automatic pain level recognition is described. Biomedical signals (EMG, BVP, EDA) and video data of a head pose are analyzed in patients undergoing fascial therapy. The impact of video data and their fusion with biomedical data is tested for the system’s performance. Decision trees and random forest are applied for classification, yielding an accuracy of 0.85. The energy of the EMG signal turned out to be a highly discriminative feature that dominated the weak classifier. Video features impact the classification results in ensembled methods.
