LoViT: Long Video Transformer for surgical phase recognition.

Online surgical phase recognition plays a significant role towards building contextual tools that could quantify performance and oversee the execution of surgical workflows. Current approaches are limited since they train spatial feature extractors using frame-level supervision that could lead to incorrect predictions due to similar frames appearing at different phases, and poorly fuse local and global features due to computational constraints which can affect the analysis of long videos commonly encountered in surgical interventions. In this paper, we present a two-stage method, called Long Video Transformer (LoViT), emphasizing the development of a temporally-rich spatial feature extractor and a phase transition map. The temporally-rich spatial feature extractor is designed to capture critical temporal information within the surgical video frames. The phase transition map provides essential insights into the dynamic transitions between different surgical phases. LoViT combines these innovations with a multiscale temporal aggregator consisting of two cascaded L-Trans modules based on self-attention, followed by a G-Informer module based on ProbSparse self-attention for processing global temporal information. The multi-scale temporal head then leverages the temporally-rich spatial features and phase transition map to classify surgical phases using phase transition-aware supervision. Our approach outperforms state-of-the-art methods on the Cholec80 and AutoLaparo datasets consistently. Compared to Trans-SVNet, LoViT achieves a 2.4 pp (percentage point) improvement in video-level accuracy on Cholec80 and a 3.1 pp improvement on AutoLaparo. Our results demonstrate the effectiveness of our approach in achieving state-of-the-art performance of surgical phase recognition on two datasets of different surgical procedures and temporal sequencing characteristics. The project page is available at https://github.com/MRUIL/LoViT.

Lazy Resampling: Fast and information preserving preprocessing for deep learning.

BACKGROUND AND OBJECTIVE: Preprocessing of data is a vital step for almost all deep learning workflows. In computer vision, manipulation of data intensity and spatial properties can improve network stability and can provide an important source of generalisation for deep neural networks. Models are frequently trained with preprocessing pipelines composed of many stages, but these pipelines come with a drawback; each stage that resamples the data costs time, degrades image quality, and adds bias to the output. Long pipelines can also be complex to design, especially in medical imaging, where cropping data early can cause significant artifacts. METHODS: We present Lazy Resampling, a software that rephrases spatial preprocessing operations as a graphics pipeline. Rather than each transform individually modifying the data, the transforms generate transform descriptions that are composited together into a single resample operation wherever possible. This reduces pipeline execution time and, most importantly, limits signal degradation. It enables simpler pipeline design as crops and other operations become non-destructive. Lazy Resampling is designed in such a way that it provides the maximum benefit to users without requiring them to understand the underlying concepts or change the way that they build pipelines. RESULTS: We evaluate Lazy Resampling by comparing traditional pipelines and the corresponding lazy resampling pipeline for the following tasks on Medical Segmentation Decathlon datasets. We demonstrate lower information loss in lazy pipelines vs. traditional pipelines. We demonstrate that Lazy Resampling can avoid catastrophic loss of semantic segmentation label accuracy occurring in traditional pipelines when passing labels through a pipeline and then back through the inverted pipeline. Finally, we demonstrate statistically significant improvements when training UNets for semantic segmentation. CONCLUSION: Lazy Resampling reduces the loss of information that occurs when running processing pipelines that traditionally have multiple resampling steps and enables researchers to build simpler pipelines by making operations such as rotation and cropping effectively non-destructive. It makes it possible to invert labels back through a pipeline without catastrophic loss of accuracy. A reference implementation for Lazy Resampling can be found at https://github.com/KCL-BMEIS/LazyResampling. Lazy Resampling is being implemented as a core feature in MONAI, an open source python-based deep learning library for medical imaging, with a roadmap for a full integration.

Qualitative exploration of the lived experiences of loneliness in later life to inform technology development.

PURPOSE: Loneliness is a negative emotional state which is common in later life. The accumulative effects of loneliness have a significant impact on the physical and mental health of older adults. We aim to qualitatively explore the experiences of loneliness in later life and identify relevant behaviours and indicators which will inform novel methods of loneliness detection and intervention. METHODS: We conducted 60 semi-structured interviews with people aged 65 and over between September 2022 and August 2023. Data were analysed using a reflective thematic approach with early theme development on NVIVO software. RESULTS: Three themes were identified from the experiences of loneliness in older adults. 1) Unique responses to loneliness, including crying, increased eating or drinking and sleep difficulties, 2) Age-related losses, such as networks, roles, and abilities to engage in activities reducing over time and 3) Individual differences in overcoming loneliness, where strategies such as keeping busy and adopting a positive mindset were impacted by motivation and mood of older adults. CONCLUSION: Distinct signs and relevant factors to loneliness in later life have been identified which can be detected by future sensing technologies. Findings of this in-depth qualitative study highlight that loneliness is a subjective experience requiring a holistic and person-centred approach to detection and intervention.

Blood Harmonisation of Endoscopic Transsphenoidal Surgical Video Frames on Phantom Models.

Physical phantom models have been integral to surgical training, yet they lack realism and are unable to replicate the presence of blood resulting from surgical actions. Existing domain transfer methods aim to enhance realism, but none facilitate blood simulation. This study investigates the overlay of blood on images acquired during endoscopic transsphenoidal pituitary surgery on phantom models. The process involves employing manual techniques using the GIMP image manipulation application and automated methods using pythons Blend Modes module. We then approach this as an image harmonisation task to assess its practicality and feasibility. Our evaluation uses Structural Similarity Index Measure and Laplacian metrics. The results we obtained emphasize the significance of image harmonisation, offering substantial insights within the surgical field. Our work is a step towards investigating data-driven models that can simulate blood for increased realism during surgical training on phantom models.

Brain reserve and physical disability in secondary progressive multiple sclerosis.

Background: The brain reserve hypothesis posits that larger maximal lifetime brain growth (MLBG) may confer protection against physical disability in multiple sclerosis (MS). Larger MLBG as a proxy for brain reserve, has been associated with reduced progression of physical disability in patients with early MS; however, it is unknown whether this association remains once in the secondary progressive phase of MS (SPMS). Our aim was to assess whether larger MLBG is associated with decreased physical disability progression in SPMS. Methods: We conducted a post hoc analysis of participants in the MS-Secondary Progressive Multi-Arm Randomisation Trial (NCT01910259), a multicentre randomised placebo-controlled trial of the neuroprotective potential of three agents in SPMS. Physical disability was measured by Expanded Disability Status Scale (EDSS), 9-hole peg test (9HPT) and 25-foot timed walk test (T25FW) at baseline, 48 and 96 weeks. MLBG was estimated by baseline intracranial volume (ICV). Multivariable time-varying Cox regression models were used to investigate the association between MLBG and physical disability progression. Results: 383 participants (mean age 54.5 years, 298 female) were followed up over 96 weeks. Median baseline EDSS was 6.0 (range 4.0-6.5). Adjusted for covariates, larger MLBG was associated with a reduced risk of EDSS progression (HR 0.84,95% CI:0.72 to 0.99;p=0.04). MLBG was not independently associated with time to progression as measured by 9HPT or T25FW. Conclusion: Larger MLBG is independently associated with physical disability progression over 96 weeks as measured by EDSS in SPMS. This suggests that MLBG as a proxy for brain reserve may continue to confer protection against disability when in the secondary progression phase of MS. Trail registration number: NCT01910259.

Reduction of surgical complications via 3D models during robotic assisted radical prostatectomy: review of current evidence and meta-analysis.

The use of 3-dimensional (3D) technology has become increasingly popular across different surgical specialities to improve surgical outcomes. 3D technology has the potential to be applied to robotic assisted radical prostatectomy to visualise the patient's prostate anatomy to be used as a preoperative and peri operative surgical guide. This literature review aims to analyse all relevant pre-existing research on this topic. Following PRISMA guidelines, a search was carried out on PubMed, Medline, and Scopus. A total of seven studies were included in this literature review; two of which used printed-3D models and the remaining five using virtual augmented reality (AR) 3D models. Results displayed variation with select studies presenting that the use of 3D models enhances surgical outcomes and reduces complications whilst others displayed conflicting evidence. The use of 3D modelling within surgery has potential to improve various areas. This includes the potential surgical outcomes, including complication rates, due to improved planning and education.

Realistic morphology-preserving generative modelling of the brain.

Medical imaging research is often limited by data scarcity and availability. Governance, privacy concerns and the cost of acquisition all restrict access to medical imaging data, which, compounded by the data-hungry nature of deep learning algorithms, limits progress in the field of healthcare AI. Generative models have recently been used to synthesize photorealistic natural images, presenting a potential solution to the data scarcity problem. But are current generative models synthesizing morphologically correct samples? In this work we present a three-dimensional generative model of the human brain that is trained at the necessary scale to generate diverse, realistic-looking, high-resolution and morphologically preserving samples and conditioned on patient characteristics (for example, age and pathology). We show that the synthetic samples generated by the model preserve biological and disease phenotypes and are realistic enough to permit use downstream in well-established image analysis tools. While the proposed model has broad future applicability, such as anomaly detection and learning under limited data, its generative capabilities can be used to directly mitigate data scarcity, limited data availability and algorithmic fairness.

Computer-aided prognosis of tuberculous meningitis combining imaging and non-imaging data.

Tuberculous meningitis (TBM) is the most lethal form of tuberculosis. Clinical features, such as coma, can predict death, but they are insufficient for the accurate prognosis of other outcomes, especially when impacted by co-morbidities such as HIV infection. Brain magnetic resonance imaging (MRI) characterises the extent and severity of disease and may enable more accurate prediction of complications and poor outcomes. We analysed clinical and brain MRI data from a prospective longitudinal study of 216 adults with TBM; 73 (34%) were HIV-positive, a factor highly correlated with mortality. We implemented an end-to-end framework to model clinical and imaging features to predict disease progression. Our model used state-of-the-art machine learning models for automatic imaging feature encoding, and time-series models for forecasting, to predict TBM progression. The proposed approach is designed to be robust to missing data via a novel tailored model optimisation framework. Our model achieved a 60% balanced accuracy in predicting the prognosis of TBM patients over the six different classes. HIV status did not alter the performance of the models. Furthermore, our approach identified brain morphological lesions caused by TBM in both HIV and non-HIV-infected, associating lesions to the disease staging with an overall accuracy of 96%. These results suggest that the lesions caused by TBM are analogous in both populations, regardless of the severity of the disease. Lastly, our models correctly identified changes in disease symptomatology and severity in 80% of the cases. Our approach is the first attempt at predicting the prognosis of TBM by combining imaging and clinical data, via a machine learning model. The approach has the potential to accurately predict disease progression and enable timely clinical intervention.

Characterizing T1 in the fetal brain and placenta over gestational age at 0.55T.

PURPOSE: T1 mapping and T1-weighted contrasts have a complimentary but currently under utilized role in fetal MRI. Emerging clinical low field scanners are ideally suited for fetal T1 mapping. The advantages are lower T1 values which results in higher efficiency and reduced field inhomogeneities resulting in a decreased requirement for specialist tools. In addition the increased bore size associated with low field scanners provides improved patient comfort and accessibility. This study aims to demonstrate the feasibility of fetal brain T1 mapping at 0.55T. METHODS: An efficient slice-shuffling inversion-recovery echo-planar imaging (EPI)-based T1-mapping and postprocessing was demonstrated for the fetal brain at 0.55T in a cohort of 38 fetal MRI scans. Robustness analysis was performed and placental measurements were taken for validation. RESULTS: High-quality T1 maps allowing the investigation of subregions in the brain were obtained and significant correlation with gestational age was demonstrated for fetal brain T1 maps ( p < 0 . 05 $$ p<0.05 $$ ) as well as regions-of-interest in the deep gray matter and white matter. CONCLUSIONS: Efficient, quantitative T1 mapping in the fetal brain was demonstrated on a clinical 0.55T MRI scanner, providing foundations for both future research and clinical applications including low-field specific T1-weighted acquisitions.

MONAI Label: A framework for AI-assisted interactive labeling of 3D medical images.

The lack of annotated datasets is a major bottleneck for training new task-specific supervised machine learning models, considering that manual annotation is extremely expensive and time-consuming. To address this problem, we present MONAI Label, a free and open-source framework that facilitates the development of applications based on artificial intelligence (AI) models that aim at reducing the time required to annotate radiology datasets. Through MONAI Label, researchers can develop AI annotation applications focusing on their domain of expertise. It allows researchers to readily deploy their apps as services, which can be made available to clinicians via their preferred user interface. Currently, MONAI Label readily supports locally installed (3D Slicer) and web-based (OHIF) frontends and offers two active learning strategies to facilitate and speed up the training of segmentation algorithms. MONAI Label allows researchers to make incremental improvements to their AI-based annotation application by making them available to other researchers and clinicians alike. Additionally, MONAI Label provides sample AI-based interactive and non-interactive labeling applications, that can be used directly off the shelf, as plug-and-play to any given dataset. Significant reduced annotation times using the interactive model can be observed on two public datasets.

Deep learning for automatic segmentation of vestibular schwannoma: a retrospective study from multi-center routine MRI.

Automatic segmentation of vestibular schwannoma (VS) from routine clinical MRI has potential to improve clinical workflow, facilitate treatment decisions, and assist patient management. Previous work demonstrated reliable automatic segmentation performance on datasets of standardized MRI images acquired for stereotactic surgery planning. However, diagnostic clinical datasets are generally more diverse and pose a larger challenge to automatic segmentation algorithms, especially when post-operative images are included. In this work, we show for the first time that automatic segmentation of VS on routine MRI datasets is also possible with high accuracy. We acquired and publicly release a curated multi-center routine clinical (MC-RC) dataset of 160 patients with a single sporadic VS. For each patient up to three longitudinal MRI exams with contrast-enhanced T1-weighted (ceT1w) (n = 124) and T2-weighted (T2w) (n = 363) images were included and the VS manually annotated. Segmentations were produced and verified in an iterative process: (1) initial segmentations by a specialized company; (2) review by one of three trained radiologists; and (3) validation by an expert team. Inter- and intra-observer reliability experiments were performed on a subset of the dataset. A state-of-the-art deep learning framework was used to train segmentation models for VS. Model performance was evaluated on a MC-RC hold-out testing set, another public VS datasets, and a partially public dataset. The generalizability and robustness of the VS deep learning segmentation models increased significantly when trained on the MC-RC dataset. Dice similarity coefficients (DSC) achieved by our model are comparable to those achieved by trained radiologists in the inter-observer experiment. On the MC-RC testing set, median DSCs were 86.2(9.5) for ceT1w, 89.4(7.0) for T2w, and 86.4(8.6) for combined ceT1w+T2w input images. On another public dataset acquired for Gamma Knife stereotactic radiosurgery our model achieved median DSCs of 95.3(2.9), 92.8(3.8), and 95.5(3.3), respectively. In contrast, models trained on the Gamma Knife dataset did not generalize well as illustrated by significant underperformance on the MC-RC routine MRI dataset, highlighting the importance of data variability in the development of robust VS segmentation models. The MC-RC dataset and all trained deep learning models were made available online.

Symptoms before and after COVID-19: a population and case-control study using prospective data.

BACKGROUND: Some individuals experience prolonged illness after acute coronavirus disease 2019 (COVID-19). We assessed whether pre-infection symptoms affected post-acute COVID illness duration. METHODS: Survival analysis was performed in adults (n=23 452) with community-managed severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection prospectively self-logging data through the ZOE COVID Symptom Study app, at least weekly, from 8 weeks before to 12 weeks after COVID-19 onset, conditioned on presence versus absence of baseline symptoms (4-8 weeks before COVID-19). A case-control study was performed in 1350 individuals with long illness (≥8 weeks, including 906 individuals (67.1%) with illness ≥12 weeks), matched 1:1 (for age, sex, body mass index, testing week, prior infection, vaccination, smoking, index of multiple deprivation) with 1350 individuals with short illness (<4 weeks). Baseline symptoms were compared between the two groups, and against post-COVID symptoms. RESULTS: Individuals reporting baseline symptoms had longer COVID-related symptom duration (median 15 days versus 10 days for individuals without baseline symptoms) with baseline fatigue nearly doubling duration. Two-thirds (910 (67.4%) of 1350) of individuals with long illness were asymptomatic beforehand. However, 440 (32.6%) had baseline symptoms, versus 255 (18.9%) of 1350 individuals with short illness (p<0.0001). Baseline symptoms doubled the odds ratio for long illness (2.14, 95% CI 1.78-2.57). Prior comorbidities were more common in individuals with long versus short illness. In individuals with long illness, baseline symptomatic (versus asymptomatic) individuals were more likely to be female, younger, and have prior comorbidities; and baseline and post-acute symptoms, and symptom burden, correlated strongly. CONCLUSIONS: Individuals experiencing symptoms before COVID-19 had longer illness duration and increased odds of long illness. However, many individuals with long illness were well before SARS-CoV-2 infection.

Optimising brain age estimation through transfer learning: A suite of pre-trained foundation models for improved performance and generalisability in a clinical setting.

Estimated age from brain MRI data has emerged as a promising biomarker of neurological health. However, the absence of large, diverse, and clinically representative training datasets, along with the complexity of managing heterogeneous MRI data, presents significant barriers to the development of accurate and generalisable models appropriate for clinical use. Here, we present a deep learning framework trained on routine clinical data (N up to 18,890, age range 18-96 years). We trained five separate models for accurate brain age prediction (all with mean absolute error ≤4.0 years, R2 ≥ .86) across five different MRI sequences (T2 -weighted, T2 -FLAIR, T1 -weighted, diffusion-weighted, and gradient-recalled echo T2 *-weighted). Our trained models offer dual functionality. First, they have the potential to be directly employed on clinical data. Second, they can be used as foundation models for further refinement to accommodate a range of other MRI sequences (and therefore a range of clinical scenarios which employ such sequences). This adaptation process, enabled by transfer learning, proved effective in our study across a range of MRI sequences and scan orientations, including those which differed considerably from the original training datasets. Crucially, our findings suggest that this approach remains viable even with limited data availability (as low as N = 25 for fine-tuning), thus broadening the application of brain age estimation to more diverse clinical contexts and patient populations. By making these models publicly available, we aim to provide the scientific community with a versatile toolkit, promoting further research in brain age prediction and related areas.

New robotic platform for transoral robotic surgery: an IDEAL stage 0 study.

Objectives: This study aims to assess the feasibility to perform transoral robotic surgery (TORS) with a new robotic platform, the Versius Surgical System (CMR Surgical, UK) in a preclinical cadaveric setting in accordance to stage 0 of the IDEAL-D framework. Design: IDEAL stage 0 preclinical assessment of the Versius Robotic System in TORS in human cadavers. Setting: All procedures were performed in a simulated operating theatre environment at a UK surgical training centre. Participants: 11 consultant head and neck surgeons from the UK, mainland Europe and the USA took part in TORS procedures on six human cadavers. Interventions: 3 key index procedures were assessed that represent the core surgical workload of TORS: lateral oropharyngectomy, tongue base resection and partial supraglottic laryngectomy. Main outcome measures: The primary outcome was the successful completion of each surgical procedure. Secondary outcomes included the optimisation of system setup, instrumentation and surgeon-reported outcomes for feasibility of each component procedural step. Results: 33 cadaveric procedures were performed and 32 were successfully completed. One supraglottic laryngectomy was not fully completed due to issues dividing the epiglottic cartilage with available instrumentation. Surgeon-reported outcomes met the minimal level of feasibility in all procedures and a consensus that it is feasible to perform TORS with Versius was reached. Available instrumentation was not representative of other robotic platforms used in TORS and further instrument optimisation is recommended before wider dissemination. Conclusions: It is feasible to perform TORS with the Versius Surgical System (CMR Surgical) within a pre-clinical cadaveric setting. Clinical evaluation is needed and appropriate with the system. Further instrument development and optimisation is desirable.

Prospective development study of the Versius Surgical System for use in transoral robotic surgery: an IDEAL stage 1/2a first in human and initial case series experience.

PURPOSE: Transoral robotic surgery is well established in the treatment paradigm of oropharyngeal pathology. The Versius Surgical System (CMR Surgical) is a robotic platform in clinical use in multiple specialities but is currently untested in the head and neck. This study utilises the IDEAL framework of surgical innovation to prospectively evaluate and report a first in human clinical experience and single centre case series of transoral robotic surgery (TORS) with Versius. METHODS: Following IDEAL framework stages 1 and 2a, the study evaluated Versius to perform first in human TORS before transitioning from benign to malignant cases. Iterative adjustments were made to system setup, instrumentation, and technique, recorded in accordance with IDEAL recommendations. Evaluation criteria included successful procedure completion, setup time, operative time, complications, and subjective impressions. Further evaluation of the system to perform four-arm surgery was conducted. RESULTS: 30 TORS procedures were successfully completed (15 benign, 15 malignant) without intraoperative complication or conversion to open surgery. Setup time significantly decreased over the study period. Instrumentation challenges were identified, urging the need for TORS-specific instruments. The study introduced four-arm surgery, showcasing Versius' unique capabilities, although limitations in distal access were observed. CONCLUSIONS: TORS is feasible with the Versius Surgical System. The development of TORS-specific instruments would benefit performance and wider adoption of the system. 4-arm surgery is possible however further evaluation is required. Multicentre evaluation (IDEAL stage 2b) is recommended.

Unpaired intra-operative OCT (iOCT) video super-resolution with contrastive learning.

Regenerative therapies show promise in reversing sight loss caused by degenerative eye diseases. Their precise subretinal delivery can be facilitated by robotic systems alongside with Intra-operative Optical Coherence Tomography (iOCT). However, iOCT's real-time retinal layer information is compromised by inferior image quality. To address this limitation, we introduce an unpaired video super-resolution methodology for iOCT quality enhancement. A recurrent network is proposed to leverage temporal information from iOCT sequences, and spatial information from pre-operatively acquired OCT images. Additionally, a patchwise contrastive loss enables unpaired super-resolution. Extensive quantitative analysis demonstrates that our approach outperforms existing state-of-the-art iOCT super-resolution models. Furthermore, ablation studies showcase the importance of temporal aggregation and contrastive loss in elevating iOCT quality. A qualitative study involving expert clinicians also confirms this improvement. The comprehensive evaluation demonstrates our method's potential to enhance the iOCT image quality, thereby facilitating successful guidance for regenerative therapies.

Assessing the optimal MRI descriptors to diagnose Ménière's disease and the added value of analysing the vestibular aqueduct.

OBJECTIVES: To evaluate the diagnostic performance and reliability of MRI descriptors used for the detection of Ménière's disease (MD) on delayed post-gadolinium MRI. To determine which combination of descriptors should be optimally applied and whether analysis of the vestibular aqueduct (VA) contributes to the diagnosis. MATERIALS AND METHODS: This retrospective single centre case-control study evaluated delayed post-gadolinium MRI of patients with Ménièriform symptoms examined consecutively between Dec 2017 and March 2023. Two observers evaluated 17 MRI descriptors of MD and quantified perilymphatic enhancement (PLE) in the cochlea. Definite MD ears according to the 2015 Barany Society criteria were compared to control ears. Cohen's kappa and diagnostic odds ratio (DORs) were calculated for each descriptor. Forward stepwise logistic regression determined which combination of MRI descriptors would best predict MD ears, and the area under the receiver operating characteristic curve for this model was measured. RESULTS: A total of 227 patients (mean age 48.3 ± 14.6, 99 men) with 96 definite MD and 78 control ears were evaluated. The presence of saccular abnormality (absent, as large as or confluent with the utricle) performed best with a DOR of 292.6 (95% confidence interval (CI), 38.305-2235.058). All VA descriptors demonstrated excellent reliability and with DORs of 7.761 (95% CI, 3.517-17.125) to 18.1 (95% CI, 8.445-39.170). Combining these saccular abnormalities with asymmetric cochlear PLE and an incompletely visualised VA correctly classified 90.2% of cases (sensitivity 84.4%, specificity 97.4%, AUC 0.938). CONCLUSION: Either absent, enlarged or confluent saccules are the best predictors of MD. Incomplete visualisation of the VA adds value to the diagnosis. CLINICAL RELEVANCE STATEMENT: A number of different MRI descriptors have been proposed for the diagnosis of Ménière's disease, but by establishing the optimally performing MRI features and highlighting new useful descriptors, there is an opportunity to improve the diagnostic performance of Ménière's disease imaging. KEY POINTS: • A comprehensive range of existing and novel vestibular aqueduct delayed post-gadolinium MRI descriptors were compared for their diagnostic performance in Ménière's disease. • Saccular abnormality (absent, confluent with or larger than the utricle) is a reliable descriptor and is the optimal individual MRI predictor of Ménière's disease. • The presence of this saccule descriptor or asymmetric perilymphatic enhancement and incomplete vestibular aqueduct visualisation will optimise the MRI diagnosis of Ménière's disease.

Glioblastoma and radiotherapy: A multicenter AI study for Survival Predictions from MRI (GRASP study).

BACKGROUND: The aim was to predict survival of glioblastoma at 8 months after radiotherapy (a period allowing for completing a typical course of adjuvant temozolomide), by applying deep learning to the first brain MRI after radiotherapy completion. METHODS: Retrospective and prospective data were collected from 206 consecutive glioblastoma, isocitrate dehydrogenase -wildtype patients diagnosed between March 2014 and February 2022 across 11 UK centers. Models were trained on 158 retrospective patients from 3 centers. Holdout test sets were retrospective (n = 19; internal validation), and prospective (n = 29; external validation from 8 distinct centers). Neural network branches for T2-weighted and contrast-enhanced T1-weighted inputs were concatenated to predict survival. A nonimaging branch (demographics/MGMT/treatment data) was also combined with the imaging model. We investigated the influence of individual MR sequences; nonimaging features; and weighted dense blocks pretrained for abnormality detection. RESULTS: The imaging model outperformed the nonimaging model in all test sets (area under the receiver-operating characteristic curve, AUC P = .038) and performed similarly to a combined imaging/nonimaging model (P > .05). Imaging, nonimaging, and combined models applied to amalgamated test sets gave AUCs of 0.93, 0.79, and 0.91. Initializing the imaging model with pretrained weights from 10 000s of brain MRIs improved performance considerably (amalgamated test sets without pretraining 0.64; P = .003). CONCLUSIONS: A deep learning model using MRI images after radiotherapy reliably and accurately determined survival of glioblastoma. The model serves as a prognostic biomarker identifying patients who will not survive beyond a typical course of adjuvant temozolomide, thereby stratifying patients into those who might require early second-line or clinical trial treatment.