Modality redundancy for MRI-based glioblastoma segmentation.

PURPOSE: Automated glioblastoma segmentation from magnetic resonance imaging is generally performed on a four-modality input, including T1, contrast T1, T2 and FLAIR. We hypothesize that information redundancy is present within these image combinations, which can possibly reduce a model's performance. Moreover, for clinical applications, the risk of encountering missing data rises as the number of required input modalities increases. Therefore, this study aimed to explore the relevance and influence of the different modalities used for MRI-based glioblastoma segmentation. METHODS: After the training of multiple segmentation models based on nnU-Net and SwinUNETR architectures, differing only in their amount and combinations of input modalities, each model was evaluated with regard to segmentation accuracy and epistemic uncertainty. RESULTS: Results show that T1CE-based segmentation (for enhanced tumor and tumor core) and T1CE-FLAIR-based segmentation (for whole tumor and overall segmentation) can reach segmentation accuracies comparable to the full-input version. Notably, the highest segmentation accuracy for nnU-Net was found for a three-input configuration of T1CE-FLAIR-T1, suggesting the confounding effect of redundant input modalities. The SwinUNETR architecture appears to suffer less from this, where said three-input and the full-input model yielded statistically equal results. CONCLUSION: The T1CE-FLAIR-based model can therefore be considered as a minimal-input alternative to the full-input configuration. Addition of modalities beyond this does not statistically improve and can even deteriorate accuracy, but does lower the segmentation uncertainty.

Determination of rotation center and diameter of femoral heads using off-the-shelf augmented reality hardware for navigation.

In total hip arthroplasty (THA), determining the center of rotation (COR) and diameter of the hip joint (acetabulum and femoral head) is essential to restore patient biomechanics. This study investigates on-the-fly determination of hip COR and size, using off-the-shelf augmented reality (AR) hardware. An AR head-mounted device (HMD) was configured with inside-out infrared tracking enabling the determination of surface coordinates using a handheld stylus. Two investigators examined 10 prosthetic femoral heads and cups, and 10 human femurs. The HMD calculated the diameter and COR through sphere fitting. Results were compared to data obtained from either verified prosthetic geometry or post-hoc CT analysis. Repeated single-observer measurements showed a mean diameter error of 0.63 mm ± 0.48 mm for the prosthetic heads and 0.54 mm ± 0.39 mm for the cups. Inter-observer comparison yielded mean diameter errors of 0.28 mm ± 0.71 mm and 1.82 mm ± 1.42 mm for the heads and cups, respectively. Cadaver testing found a mean COR error of 3.09 mm ± 1.18 mm and a diameter error of 1.10 mm ± 0.90 mm. Intra- and inter-observer reliability averaged below 2 mm. AR-based surface mapping using HMD proved accurate and reliable in determining the diameter of THA components with promise in identifying COR and diameter of osteoarthritic femoral heads.

Structure-function in smokers: when a small airways test really reflects the small airways.

If multiple-breath washout (MBW)-derived acinar ventilation heterogeneity (Sacin) really represents peripheral units, the N2 phase-III of the first MBW exhalation should be curvilinear. This is essentially due to the superposed effect of gas diffusion and convection resulting in an equilibration of N2 concentrations between neighboring lung units throughout exhalation. We investigated this in smokers with computed tomography (CT)-proven functional small airway disease. Instantaneous N2-slopes were computed over 40-ms intervals throughout phase-III and normalized by mean phase-III N2 concentration. N2 phase-III (concave) curvilinearity was quantified as the rate at which the instantaneous N2-slope decreases past the phase-II peak over a 1-s interval; for a linear N2 phase-III unaffected by diffusion, this rate would amount to 0 L-1/s. N2 phase-III curvilinearity was obtained on the experimental curves and on existing model simulations of N2 curves from a normal peripheral lung model and one with missing terminal bronchioles (either 50% or 30% TB left). In 46 smokers [66 (±8) yr; 49 (±26) pack·yr] with CT-based evidence of peripheral lung destruction, instantaneous N2-slope decrease was compared between those with (fSAD+fEmphys) > 20% [-0.26 ± 0.14 (SD) L-1/s; n = 24] and those with (fSAD+fEmphys) < 20% [-0.16 ± 0.12 (SD) L-1/s; n = 22] (P = 0.014). Experimental values fell in the range predicted by a realistic peripheral lung model with progressive reduction of terminal bronchioles: values of instantaneous N2-slope decrease obtained from model simulations were -0.09 L-1/s (normal lung; 100% TB left), -0.17 L-1/s (normal lung 50% TB left), and -0.29 L-1/s (30% TB left). In smokers with CT-based evidence of functional small airway alterations, it is possible to demonstrate that Sacin really does represent the most peripheral airspaces.NEW & NOTEWORTHY In smokers with computed tomography-based evidence of functional small airway alterations by parametric response mapping, it is possible to demonstrate that the multiple-breath washout-derived Sacin, an index of acinar ventilation heterogeneity, actually does represent the most peripheral airspaces. This is done by verifying on experimental N2 washout curves of the first breath, N2 phase-III concavity predicted by the diffusion-convection interdependence model.

The STOIC2021 COVID-19 AI challenge: Applying reusable training methodologies to private data.

Challenges drive the state-of-the-art of automated medical image analysis. The quantity of public training data that they provide can limit the performance of their solutions. Public access to the training methodology for these solutions remains absent. This study implements the Type Three (T3) challenge format, which allows for training solutions on private data and guarantees reusable training methodologies. With T3, challenge organizers train a codebase provided by the participants on sequestered training data. T3 was implemented in the STOIC2021 challenge, with the goal of predicting from a computed tomography (CT) scan whether subjects had a severe COVID-19 infection, defined as intubation or death within one month. STOIC2021 consisted of a Qualification phase, where participants developed challenge solutions using 2000 publicly available CT scans, and a Final phase, where participants submitted their training methodologies with which solutions were trained on CT scans of 9724 subjects. The organizers successfully trained six of the eight Final phase submissions. The submitted codebases for training and running inference were released publicly. The winning solution obtained an area under the receiver operating characteristic curve for discerning between severe and non-severe COVID-19 of 0.815. The Final phase solutions of all finalists improved upon their Qualification phase solutions.