Generalizability of machine learning for classification of schizophrenia based on resting‐state functional MRI data

Machine learning has increasingly been applied to classification of schizophrenia in neuroimaging research. However, direct replication studies and studies seeking to investigate generalizability are scarce. To address these issues, we assessed within‐site and between‐site generalizability of a machine learning classification framework which achieved excellent performance in a previous study using two independent resting‐state functional magnetic resonance imaging data sets collected from different sites and scanners. We established within‐site generalizability of the classification framework in the main data set using cross‐validation. Then, we trained a model in the main data set and investigated between‐site generalization in the validated data set using external validation. Finally, recognizing the poor between‐site generalization performance, we updated the unsupervised algorithm to investigate if transfer learning using additional unlabeled data were able to improve between‐site classification performance. Cross‐validation showed that the published classification procedure achieved an accuracy of 0.73 using majority voting across all selected components. External validation found a classification accuracy of 0.55 (not significant) and 0.70 (significant) using the direct and transfer learning procedures, respectively. The failure of direct generalization from one site to another demonstrates the limitation of within‐site cross‐validation and points toward the need to incorporate efforts to facilitate application of machine learning across multiple data sets. The improvement in performance with transfer learning highlights the importance of taking into account the properties of data when constructing predictive models across samples and sites. Our findings suggest that machine learning classification result based on a single study should be interpreted cautiously.     

Non-response for health-related quality of life outcomes in ICU patients: A systematic review of the reporting in randomised trials

Background

Health-related quality of life (HRQoL) is frequently assessed in randomised clinical trials (RCTs) in the intensive care unit (ICU), but data are limited regarding the proportions of patients without responses or not surviving to HRQoL follow-up and the handling of this. We aimed to describe the extent and pattern of missing HRQoL data in intensive care trials and describe how these data and deaths were handled statistically.

Methods

We conducted a systematic review and meta-analysis following a published protocol. We searched PubMed, EMBASE, CINAHL and Cochrane Library for RCTs involving adult ICU patients reporting HRQoL as an outcome and excluded RCTs unobtainable in full text. We performed risk of bias assessment independently and in duplicate.

Results

We included 196 outcomes from 88 RCTs published in the years 2002–2022; the numbers of patients alive and eligible to respond HRQoL were reported in 76% of trials. At follow-up, median 27% (interquartile range 14%–39%) of patients had died, and median 20% (9%–38%) of survivors did not respond across outcomes. Analyses of 80% of outcomes were restricted to complete cases only. The handling of non-survivors in analyses were reported for 46% of outcomes, with 26% of all outcomes reported as including non-survivors (using the value zero or the worst possible score).

Conclusion

For HRQoL outcomes in ICU trials, we found that mortality at time of follow-up was high and non-response among survivors frequent. The reporting and statistical handling of these issues were insufficient, which may have biased results.     

Bioimpedance as a measure of fluids in patients with septic shock. A prospective observational study

Background

Septic shock is often treated with aggressive fluid resuscitation leading to profound fluid overload. The assessment of fluid status relies on suboptimal measures making treatment difficult. Bioelectrical impedance analysis is an alternative but the validity is unclear. The aim of this study was to determine the validity of bioelectrical impedance analysis for fluid measures in patients with septic shock.

Methods

Single-center, prospective observational cohort study. We included adult ICU patients with septic shock. We evaluated the agreement between measures on the left and right side of the patient and measures 1 h apart by two bioelectrical impedance devices. Results are presented as Bland Altman plots with 95% Limits of Agreements (LoA) and as correlations between bioelectrical impedance analysis results and clinical markers of fluids.

Results

Forty-nine patients were included. The agreement between measures on the left and the right side of the patient and after 1 h was overall without bias, but with wide LoA's. Fluid overload 1 h apart showed the most narrow 95% LoA (−2.4–2.9 L). The same wide limits of agreements were observed when comparing devices. For example, total body water with 95% LoA of −14.8 –16.7 L. Correlations between bioelectrical impedance analysis and clinical measures were low but statistically significant.

Conclusions

In patients with septic shock bioelectrical impedance analysis had no systematic errors or bias, but wide limits of agreement, indicating that the devices have a large and uncorrectable random error. Fluid status by bioelectrical impedance analysis is not sufficiently accurate to guide treatment in this group of patients.     

Kidney‐Induced Cardiac Allograft Tolerance in Miniature Swine is Dependent on MHC‐Matching of Donor Cardiac and Renal Parenchyma

Kidney allografts possess the ability to enable a short course of immunosuppression to induce tolerance of themselves and of cardiac allografts across a full‐MHC barrier in miniature swine. However, the renal element(s) responsible for kidney‐induced cardiac allograft tolerance (KICAT) are unknown. Here we investigated whether MHC disparities between parenchyma versus hematopoietic‐derived “passenger” cells of the heart and kidney allografts affected KICAT. Heart and kidney allografts were co‐transplanted into MHC‐mismatched recipients treated with high‐dose tacrolimus for 12 days. Group 1 animals (n = 3) received kidney and heart allografts fully MHC‐mismatched to each other and to the recipient. Group 2 animals (n = 3) received kidney and heart allografts MHC‐matched to each other but MHC‐mismatched to the recipient. Group 3 animals (n = 3) received chimeric kidney allografts whose parenchyma was MHC‐mismatched to the donor heart. Group 4 animals (n = 3) received chimeric kidney allografts whose passenger leukocytes were MHC‐mismatched to the donor heart. Five of six heart allografts in Groups 1 and 3 rejected <40 days. In contrast, heart allografts in Groups 2 and 4 survived >150 days without rejection (p < 0.05). These data demonstrate that KICAT requires MHC‐matching between kidney allograft parenchyma and heart allografts, suggesting that cells intrinsic to the kidney enable cardiac allograft tolerance.     

A novel suture method to place and adjust peripheral nerve catheters

We have developed a peripheral nerve catheter, attached to a needle, which works like an adjustable suture. We used in‐plane ultrasound guidance to place 45 catheters close to the femoral, saphenous, sciatic and distal tibial nerves in cadaver legs. We displaced catheters after their initial placement and then attempted to return them to their original positions. We used ultrasound to evaluate the initial and secondary catheter placements and the spread of injectate around the nerves. In 10 cases, we confirmed catheter position by magnetic resonance imaging. We judged 43/45 initial placements successful and 42/43 secondary placements successful by ultrasound, confirmed in 10/10 cases by magnetic resonance imaging.     

The effect of the bite plane splint on the electromyographic silent period duration

An electromyographic study of the masseter and anterior part of the temporalis muscles was performed on ten patients presenting temporomandibular joint dysfunction symptoms. The EMG silent periods (SP) produced in the open-close-clench cycle and jaw-jerk reflex were compared for duration before and after treatment with an occlusal bite splint. Following use of the splint, there was a shortening of SP indicating the possible use of the duration of SP as a diagnostic measurement, and also as an indication of treatment effectiveness.