Development and evaluation of machine learning models based on X-ray radiomics for the classification and differentiation of malignant and benign bone tumors.

OBJECTIVES: To develop and validate machine learning models to distinguish between benign and malignant bone lesions and compare the performance to radiologists. METHODS: In 880 patients (age 33.1 ± 19.4 years, 395 women) diagnosed with malignant (n = 213, 24.2%) or benign (n = 667, 75.8%) primary bone tumors, preoperative radiographs were obtained, and the diagnosis was established using histopathology. Data was split 70%/15%/15% for training, validation, and internal testing. Additionally, 96 patients from another institution were obtained for external testing. Machine learning models were developed and validated using radiomic features and demographic information. The performance of each model was evaluated on the test sets for accuracy, area under the curve (AUC) from receiver operating characteristics, sensitivity, and specificity. For comparison, the external test set was evaluated by two radiology residents and two radiologists who specialized in musculoskeletal tumor imaging. RESULTS: The best machine learning model was based on an artificial neural network (ANN) combining both radiomic and demographic information achieving 80% and 75% accuracy at 75% and 90% sensitivity with 0.79 and 0.90 AUC on the internal and external test set, respectively. In comparison, the radiology residents achieved 71% and 65% accuracy at 61% and 35% sensitivity while the radiologists specialized in musculoskeletal tumor imaging achieved an 84% and 83% accuracy at 90% and 81% sensitivity, respectively. CONCLUSIONS: An ANN combining radiomic features and demographic information showed the best performance in distinguishing between benign and malignant bone lesions. The model showed lower accuracy compared to specialized radiologists, while accuracy was higher or similar compared to residents. KEY POINTS: • The developed machine learning model could differentiate benign from malignant bone tumors using radiography with an AUC of 0.90 on the external test set. • Machine learning models that used radiomic features or demographic information alone performed worse than those that used both radiomic features and demographic information as input, highlighting the importance of building comprehensive machine learning models. • An artificial neural network that combined both radiomic and demographic information achieved the best performance and its performance was compared to radiology readers on an external test set.

The values of web archives.

This article considers how the development, promotion and adoption of a set of core values for web archives, linked to principles of "good governance", will help them to tackle the challenges of sustainability, accountability and inclusiveness that are central to their long-term societal and cultural worth. It outlines the work that has already been done to address these questions, as web archiving begins to move out of its establishment phase, and then discusses seven key principles of good governance that might be adapted by and embedded within web archives: participation, consensus, accountability, transparency, effectiveness and efficiency, inclusivity and legality. The article concludes with a call to action for researchers and archivists to co-create the core values for web archives that will be required if they are to remain a vital part of our cultural heritage infrastructure.

Multitask Deep Learning for Segmentation and Classification of Primary Bone Tumors on Radiographs.

Background An artificial intelligence model that assesses primary bone tumors on radiographs may assist in the diagnostic workflow. Purpose To develop a multitask deep learning (DL) model for simultaneous bounding box placement, segmentation, and classification of primary bone tumors on radiographs. Materials and Methods This retrospective study analyzed bone tumors on radiographs acquired prior to treatment and obtained from patient data from January 2000 to June 2020. Benign or malignant bone tumors were diagnosed in all patients by using the histopathologic findings as the reference standard. By using split-sample validation, 70% of the patients were assigned to the training set, 15% were assigned to the validation set, and 15% were assigned to the test set. The final performance was evaluated on an external test set by using geographic validation, with accuracy, sensitivity, specificity, and 95% CIs being used for classification, the intersection over union (IoU) being used for bounding box placements, and the Dice score being used for segmentations. Results Radiographs from 934 patients (mean age, 33 years ± 19 [standard deviation]; 419 women) were evaluated in the internal data set, which included 667 benign bone tumors and 267 malignant bone tumors. Six hundred fifty-four patients were in the training set, 140 were in the validation set, and 140 were in the test set. One hundred eleven patients were in the external test set. The multitask DL model achieved 80.2% (89 of 111; 95% CI: 72.8, 87.6) accuracy, 62.9% (22 of 35; 95% CI: 47, 79) sensitivity, and 88.2% (67 of 76; CI: 81, 96) specificity in the classification of bone tumors as malignant or benign. The model achieved an IoU of 0.52 ± 0.34 for bounding box placements and a mean Dice score of 0.60 ± 0.37 for segmentations. The model accuracy was higher than that of two radiologic residents (71.2% and 64.9%; P = .002 and P < .001, respectively) and was comparable with that of two musculoskeletal fellowship-trained radiologists (83.8% and 82.9%; P = .13 and P = .25, respectively) in classifying a tumor as malignant or benign. Conclusion The developed multitask deep learning model allowed for accurate and simultaneous bounding box placement, segmentation, and classification of primary bone tumors on radiographs. © RSNA, 2021 Online supplemental material is available for this article. See also the editorial by Carrino in this issue.

Daily stress, presleep arousal, and sleep in healthy young women: a daily life computerized sleep diary and actigraphy study.

OBJECTIVE: Our study aimed to further elucidate the mediating role of presleep arousal in the relationship between daily stress and sleep by investigating subjective sleep quality and actigraphy-assessed sleep efficiency (SE) on both within- and between-participant levels in a sample of healthy young women. METHODS: Multilevel modeling was applied on electronically assessed data comprising 14 consecutive nights in 145 healthy young women to assess the relationship between daily stress, presleep (somatic and cognitive) arousal, and sleep on both levels between participants and within participants across days. RESULTS: Higher levels of daily stress were consistently and significantly associated with higher levels of somatic and cognitive arousal. Somatic arousal mediated the relationship between daily stress and worsened subjective sleep quality on the between-participant level, while cognitive arousal mediated the relationship between daily stress and worsened subjective sleep quality on the within-participants level. Unexpectedly, healthy young women showed higher SE following days with above-average stress with somatic arousal mediating this relationship. CONCLUSIONS: Our data corroborate the role of presleep arousal mediating the relationship between daily stress and subjective sleep quality. Interestingly this effect was restricted to somatic arousal being relevant on interindividual levels and cognitive arousal on intraindividual levels. For young and healthy individuals who experience high stress and arousal, well-established cognitive-behavioral techniques could be useful to regulate arousal and prevent worse subjective sleep quality.

Heightened beta EEG activity during nonrapid eye movement sleep in primary insomnia patients with reports of childhood maltreatment.

The present study explores the relationship between childhood maltreatment experiences and spectral power in high-frequency EEG activity during sleep in a sample of adults experiencing primary insomnia. Forty-five nontreated patients with primary insomnia spent three consecutive nights in the sleep laboratory, during which polysomnographic recordings were carried out. Nonrapid eye movement and rapid eye movement EEG data were analyzed using spectral analysis. In addition, each participant completed several self-report questionnaires assessing maltreatment in childhood and adolescence, current level of stress, and current depressivity. Insomnia patients with self-reported history of moderate to severe childhood maltreatment (MAL group; n = 25), as measured by the Childhood Trauma Questionnaire, were compared with insomnia patients without such a history (non-MAL group; n = 20). The MAL group exhibited more absolute and relative beta 1 and beta 2 power in nonrapid eye movement sleep and more absolute beta 1 and beta 2 activity in rapid eye movement sleep than the non-MAL group. Contrary to hypothesis, no group differences were found in gamma frequency band. The results suggest an association between history of childhood maltreatment and increased beta EEG activity particularly during nonrapid eye movement sleep in adult insomnia, what may reflect heightened psychophysiologic arousal during sleep.

Relationship between early-life stress load and sleep in psychiatric outpatients: a sleep diary and actigraphy study.

The present study aimed to investigate whether stress experienced early in life is associated with actigraphic and subjective sleep measures in a sample of adult psychiatric outpatients. A total of 48 psychiatric outpatients completed self-report questionnaires assessing current depression, current anxiety symptoms and stress load during childhood (before the age of 13 years), adolescence (between the age of 13 and 18 years) and adulthood (between the age of 19 and current age). Sleep-related activity was measured using 24-h wrist actigraphy over a 7-day period at home, during which participants also kept a sleep diary. High stress load in childhood, but not in adolescence, was associated with shortened actigraphically assessed total sleep time, prolonged sleep onset latency, decreased sleep efficiency and an increased number of body movements in sleep, even after accounting for the effects of later occurring stress and psychopathological symptoms such as depression and anxiety scores. Unexpectedly, no significant associations between early-life stress load and subjective sleep measures were found. Results are consistent with findings from previous studies indicating an association between childhood adversities and higher levels of nocturnal activity. The findings suggest that high stress load during childhood might be a vulnerability factor for sleep continuity problems in adulthood.

Adverse childhood experiences associated with sleep in primary insomnia.

The objectives were to explore the association between self-reported adverse childhood experiences (ACE) and sleep in adults suffering from primary insomnia and to examine the impact of presleep stress on this relationship. Fifty-nine patients with primary insomnia, aged 21-55 years, were administered the Childhood Trauma Questionnaire (CTQ) and then divided into two groups according to the achieved scores: with moderate/severe or low/no reports of ACE. The participants spent three consecutive nights in the sleep laboratory in order to record polysomnographic and actigraphic sleep parameters. A stress induction technique was administered by activating negative autobiographical memories immediately before sleep in the second or third night. Results show that 46% of the insomniac patients reported moderate to severe ACE. This group exhibited a significantly greater number of awakenings and more movement arousals compared to patients with low or no reports of ACE. Actigraphic data also indicated more disturbed sleep and increased nocturnal activity for the high-ACE group. On the other hand, no specific group differences were found with regard to stress condition. The results support the assumption that it is possible to identify a subgroup among patients with primary insomnia who has experienced severe maltreatment in childhood and adolescence. This subgroup appears to differ in several sleep parameters, indicating a more disturbed sleep compared to primary insomniacs with low or no reports of ACE. With regard to sleep-disturbing nightly patterns of arousal, parallels between individuals with high ACE and trauma victims as well as post-traumatic stress disorder-patients suggest themselves.

Increased nocturnal activity associated with adverse childhood experiences in patients with primary insomnia.

The present study examined to what extent adverse childhood experiences (ACE), in addition to demographic characteristics, current level of stress, depression, and arousability predisposition, are associated with sleep measures in adult insomnia. Thirty-nine adults suffering from primary insomnia completed self-report questionnaires assessing ACE, current level of stress, predisposition towards increased arousability, and depression. They were monitored for 7 consecutive nights at home with wrist actigraphs to evaluate objective sleep-related activity. Blockwise multiple regression analyses were performed to determine which variables were the most important predictors of sleep measures. ACE proved to be important predictors of actigraphically assessed sleep onset latency, sleep efficiency, number of body movements, and moving time, whereas the set of the remaining variables had no significant impact on these sleep measures. These findings suggest that there is an association between childhood maltreatment history and sleep in patients with primary insomnia. We presume that sleep-related nightly activity can be regarded as an aftereffect of long-lasting stressful experiences in childhood.