Designing clinically translatable artificial intelligence systems for high-dimensional medical imaging

Shad, Rohan; Cunningham, John P.; Ashley, Euan A.; Langlotz, Curtis P.; Hiesinger, William

doi:10.1038/s42256-021-00399-8

Cited by 46 publications

(28 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“… 3 , 4 Certain methods, such as saliency maps or feature attribution attempt to deduce how these learning algorithms detect complex features. 99 However, just 2.1% ( n = 5) of studies reported such methods, hindering model interpretation. This highlights the importance of future work reporting DL interpretation to improve comprehension and transparency of algorithmic predictions.…”

Section: Discussionmentioning

confidence: 99%

Automated brain tumor identification using magnetic resonance imaging: A systematic review and meta-analysis

Kouli

Hassane

Badran

et al. 2022

Neuro-Oncology Advances

View full text Add to dashboard Cite

Background Automated brain tumor identification facilitates diagnosis and treatment planning. We evaluate the performance of traditional machine learning (TML) and deep learning (DL) in brain tumor detection and segmentation, using MRI. Methods A systematic literature search from January 2000 to May 8, 2021 was conducted. Study quality was assessed using the Checklist for Artificial Intelligence in Medical Imaging (CLAIM). Detection meta-analysis was performed using a unified hierarchical model. Segmentation studies were evaluated using a random effects model. Sensitivity analysis was performed for externally validated studies. Results Of 224 studies included in the systematic review, 46 segmentation and 38 detection studies were eligible for meta-analysis. In detection, DL achieved a lower false positive rate compared to TML; 0.018 (95% CI, 0.011 to 0.028) and 0.048 (0.032 to 0.072) (P < .001), respectively. In segmentation, DL had a higher dice similarity coefficient (DSC), particularly for tumor core (TC); 0.80 (0.77 to 0.83) and 0.63 (0.56 to 0.71) (P < .001), persisting on sensitivity analysis. Both manual and automated whole tumor (WT) segmentation had “good” (DSC ≥ 0.70) performance. Manual TC segmentation was superior to automated; 0.78 (0.69 to 0.86) and 0.64 (0.53 to 0.74) (P = .014), respectively. Only 30% of studies reported external validation. Conclusions The comparable performance of automated to manual WT segmentation supports its integration into clinical practice. However, manual outperformance for sub-compartmental segmentation highlights the need for further development of automated methods in this area. Compared to TML, DL provided superior performance for detection and sub-compartmental segmentation. Improvements in the quality and design of studies, including external validation, are required for the interpretability and generalizability of automated models.

show abstract

Section: Discussionmentioning

confidence: 99%

Automated brain tumor identification using magnetic resonance imaging: A systematic review and meta-analysis

Kouli

Hassane

Badran

et al. 2022

Neuro-Oncology Advances

View full text Add to dashboard Cite

show abstract

“…In the vein of personalized medicine, this process is solely dependent on data from the patient in question, which avoids the necessity for consideration of confounders introduced by different patients 33, 50 . This constitutes an extremely relevant feature in the face of increasing evidence that many research studies overestimate the performance of DL algorithms due to poor selection of test data relative to the training data 51, 52 . Furthermore, algorithms trained on data from a specific camera may not necessarily generalize to slightly adapted conditions 33, 53 .…”

Section: Discussionmentioning

confidence: 99%

Spectral imaging enables contrast agent-free real-time ischemia monitoring in laparoscopic surgery

Ayala

Adler

Seidlitz

et al. 2022

Preprint

View full text Add to dashboard Cite

Laparoscopic surgery has evolved as a key technique for cancer diagnosis and therapy. While characterization of the tissue perfusion is crucial in various procedures, such as partial nephrectomy, doing so by means of visual inspection remains highly challenging. Spectral imaging takes advantage of the fact that different tissue components have unique optical properties to recover relevant information on tissue function such as ischemia. However, clinical success stories for advancing laparoscopic surgery with spectral imaging are lacking to date. To address this bottleneck, we developed the first laparoscopic real-time multispectral imaging (MSI) system featuring a compact and lightweight multispectral camera and the possibility to complement the conventional RGB (Red, Green, and Blue) surgical view of the patient with functional information at a video rate of 25 Hz. To account for the high inter-patient variability of human tissue, we phrase the problem of ischemia detection as an out-of-distribution (OoD) detection problem that does not rely on data from any other patient. Using an ensemble of invertible neural networks (INNs) as a core component, our algorithm computes the likelihood of ischemia based on a short (several seconds) video sequence acquired at the beginning of each surgery. A first-in-human trial performed on 10 patients undergoing partial nephrectomy demonstrates the feasibility of our approach for fully-automatic live ischemia monitoring during laparoscopic surgery. Compared to the clinical state-of-the-art approach based on indocyanine green (ICG) fluorescence, the proposed MSI-based method does not require the injection of a contrast agent and is repeatable if the wrong segment has been clamped. Spectral imaging combined with advanced deep learning-based analysis tools could thus evolve as an important tool for fast, efficient, reliable and safe functional imaging in minimally invasive surgery.

show abstract

“…In prospective applications, even a robust deep learning model could still be vulnerable to new or unknown confounding factors or new data that fall outside the purview of the original training data (e.g., images acquired with a new MRI scanner). Uncertainty estimation to detect out-of-distribution samples is a recent area of interest in deep learning [ 5 , 15 ], which has seen the development of sophisticated means of measuring uncertainty [ 16 – 18 ]), usually formulated as methods for quantifying and detecting the “distance” of an input datapoint from the training set. A simpler approach is to simply train an ensemble of independent base learners and average their output, quantifying uncertainty by measuring how consistently they agree with one another [ 19 , 20 ].…”

Section: Introductionmentioning

confidence: 99%

Adversarial confound regression and uncertainty measurements to classify heterogeneous clinical MRI in Mass General Brigham

Leming

Das

2023

PLoS ONE

View full text Add to dashboard Cite

In this work, we introduce a novel deep learning architecture, MUCRAN (Multi-Confound Regression Adversarial Network), to train a deep learning model on clinical brain MRI while regressing demographic and technical confounding factors. We trained MUCRAN using 17,076 clinical T1 Axial brain MRIs collected from Massachusetts General Hospital before 2019 and demonstrated that MUCRAN could successfully regress major confounding factors in the vast clinical dataset. We also applied a method for quantifying uncertainty across an ensemble of these models to automatically exclude out-of-distribution data in AD detection. By combining MUCRAN and the uncertainty quantification method, we showed consistent and significant increases in the AD detection accuracy for newly collected MGH data (post-2019; 84.6% with MUCRAN vs. 72.5% without MUCRAN) and for data from other hospitals (90.3% from Brigham and Women’s Hospital and 81.0% from other hospitals). MUCRAN offers a generalizable approach for deep-learning-based disease detection in heterogenous clinical data.

show abstract

Designing clinically translatable artificial intelligence systems for high-dimensional medical imaging

Cited by 46 publications

References 52 publications

Automated brain tumor identification using magnetic resonance imaging: A systematic review and meta-analysis

Automated brain tumor identification using magnetic resonance imaging: A systematic review and meta-analysis

Spectral imaging enables contrast agent-free real-time ischemia monitoring in laparoscopic surgery

Adversarial confound regression and uncertainty measurements to classify heterogeneous clinical MRI in Mass General Brigham

Contact Info

Product

Resources

About