| 1. |
- Ali, Muhaddisa Barat, 1986, et al.
(author)
-
Prediction of glioma‑subtypes: comparison of performance on a DL classifier using bounding box areas versus annotated tumors
- 2022
-
In: BioMedical Engineering Online. - : Springer Science and Business Media LLC. - 1475-925X .- 2524-4426. ; 4
-
Journal article (peer-reviewed)abstract
- Background: For brain tumors, identifying the molecular subtypes from magnetic resonance imaging (MRI) isdesirable, but remains a challenging task. Recent machine learning and deep learning (DL) approaches may help theclassification/prediction of tumor subtypes through MRIs. However, most of these methods require annotated datawith ground truth (GT) tumor areas manually drawn by medical experts. The manual annotation is a time consumingprocess with high demand on medical personnel. As an alternative automatic segmentation is often used. However, itdoes not guarantee the quality and could lead to improper or failed segmented boundaries due to differences in MRIacquisition parameters across imaging centers, as segmentation is an ill‑defined problem. Analogous to visual objecttracking and classification, this paper shifts the paradigm by training a classifier using tumor bounding box areas inMR images. The aim of our study is to see whether it is possible to replace GT tumor areas by tumor bounding boxareas (e.g. ellipse shaped boxes) for classification without a significant drop in performance.Method: In patients with diffuse gliomas, training a deep learning classifier for subtype prediction by employ‑ing tumor regions of interest (ROIs) using ellipse bounding box versus manual annotated data. Experiments wereconducted on two datasets (US and TCGA) consisting of multi‑modality MRI scans where the US dataset containedpatients with diffuse low‑grade gliomas (dLGG) exclusively.Results: Prediction rates were obtained on 2 test datasets: 69.86% for 1p/19q codeletion status on US dataset and79.50% for IDH mutation/wild‑type on TCGA dataset. Comparisons with that of using annotated GT tumor data fortraining showed an average of 3.0% degradation (2.92% for 1p/19q codeletion status and 3.23% for IDH genotype).Conclusion: Using tumor ROIs, i.e., ellipse bounding box tumor areas to replace annotated GT tumor areas for train‑ing a deep learning scheme, cause only a modest decline in performance in terms of subtype prediction. With moredata that can be made available, this may be a reasonable trade‑off where decline in performance may be counter‑acted with more data.
|
|
| 2. |
- Alt Murphy, Margit, 1970, et al.
(author)
-
An upper body garment with integrated sensors for people with neurological disorders – early development and evaluation
- 2019
-
In: BMC Biomedical Engineering. - : Springer Science and Business Media LLC. - 2524-4426. ; 1:3
-
Journal article (peer-reviewed)abstract
- Background: To develop a novel wearable garment with integrated sensors for continuous monitoring of physiological and movement related variables to evaluate progression, tailor treatments and improve diagnosis in epilepsy, Parkinson’s disease and stroke. Methods: An iterative development process and evaluation of an upper body garment with integrated sensors included: identification of user needs, specification of technical and garment requirements, garment development and production as well as evaluation of garment design, functionality and usability. The project is a multidisciplinary collaboration with experts from medical, engineering, textile, and material science within the wearITmed consortium. The work was organized in regular meetings, task groups and hands-on workshops. User needs were identified using results from a mixed-methods systematic review, a focus group study and expert groups. Usability was evaluated in 19 individuals (13 controls, 6 patients with Parkinson’s disease) using semi-structured interviews and qualitative content analysis. Results: A prototype designed to monitor movements and heart rate was developed. The garment was well accepted by the users regarding design and comfort, although the users were cautious about the technology and suggested improvements. All electronic components passed a washability test. The most robust data was obtained from accelerometer and gyroscope sensors while the electrodes for heart rate registration were sensitive to motion. artefacts. The algorithm development within the wearITmed consortium has shown promising results. Conclusions: The prototype was accepted by the users. Technical improvements are needed, but preliminary data indicate that the garment has potential to be used as a tool for diagnosis and treatment selection and could provide added value for monitoring seizures in epilepsy, fluctuations in PD and activity levels in stroke. Future work aims to improve the prototype further, develop algorithms, and evaluate the functionality and usability in targeted patient groups. The potential of incorporating blood pressure and heart-rate variability monitoring will also be explored.
|
|
| 3. |
- Fridolfsson, Jonatan, 1992, et al.
(author)
-
Workplace activity classification from shoebased movement sensors
- 2020
-
In: BMC Biomedical Engineering. - : Springer Science and Business Media LLC. - 2524-4426. ; 2
-
Journal article (peer-reviewed)abstract
- Background: High occupational physical activity is associated with lower health. Shoe-based movement sensors can provide an objective measurement of occupational physical activity in a lab setting but the performance of such methods in a free-living environment have not been investigated. The aim of the current study was to investigate the feasibility and accuracy of shoe sensor-based activity classification in an industrial work setting. Results: An initial calibration part was performed with 35 subjects who performed different workplace activities in a structured lab setting while the movement was measured by a shoe-sensor. Three different machine-learning models (random forest (RF), support vector machine and k-nearest neighbour) were trained to classify activities using the collected lab data. In a second validation part, 29 industry workers were followed at work while an observer noted their activities and the movement was captured with a shoe-based movement sensor. The performance of the trained classification models were validated using the free-living workplace data. The RF classifier consistently outperformed the other models with a substantial difference in in the free-living validation. The accuracy of the initial RF classifier was 83% in the lab setting and 43% in the free-living validation. After combining activities that was difficult to discriminate the accuracy increased to 96 and 71% in the lab and freeliving setting respectively. In the free-living part, 99% of the collected samples either consisted of stationary activities or walking. Conclusions: Walking and stationary activities can be classified with high accuracy from a shoe-based movement sensor in a free-living occupational setting. The distribution of activities at the workplace should be considered when validating activity classification models in a free-living setting.
|
|
| 4. |
- Holm, Ivar, et al.
(author)
-
Fourteen days free-living evaluation of an open-source algorithm for counting steps in healthy adults with a large variation in physical activity level.
- 2023
-
In: BMC biomedical engineering. - 2524-4426. ; 5
-
Journal article (peer-reviewed)abstract
- The number of steps by an individual, has traditionally been assessed with a pedometer, but increasingly with an accelerometer. The ActiLife software (AL) is the most common way to process accelerometer data to steps, but it is not open source which could aid understanding of measurement errors. The aim of this study was to compare assessment of steps from the open-source algorithm part of the GGIR package and two closed algorithms, AL normal (n) and low frequency extension (lfe) algorithms to Yamax pedometer, as reference. Free-living in healthy adults with a wide range of activity level was studied.A total 46 participants divided by activity level into a low-medium active group and a high active group, wore both an accelerometer and a pedometer for 14days. In total 614 complete days were analyzed. A significant correlation between Yamax and all three algorithms was shown but all comparisons were significantly different with paired t-tests except for ALn vs Yamax. The mean bias shows that ALn slightly overestimated steps in the low-medium active group and slightly underestimated steps in high active group. The mean percentage error (MAPE) was 17% and 9% respectively. The ALlfe overestimated steps by approximately 6700/day in both groups and the MAPE was 88% in the low-medium active group and 43% in the high active group. The open-source algorithm underestimated steps with a systematic error related to activity level. The MAPE was 28% in the low-medium active group and 48% in the high active group.The open-source algorithm captures steps fairly well in low-medium active individuals when comparing with Yamax pedometer, but did not show satisfactory results in more active individuals, indicating that it must be modified before implemented in population research. The AL algorithm without the low frequency extension measures similar number of steps as Yamax in free-living and is a useful alternative before a valid open-source algorithm is available.
|
|
| 5. |
|
|
