Shortcut learning in medical AI hinders generalization: method for estimating AI model generalization without external data.

Healthcare datasets are becoming larger and more complex, necessitating the development of accurate and generalizable AI models for medical applications. Unstructured datasets, including medical imaging, electrocardiograms, and natural language data, are gaining attention with advancements in deep convolutional neural networks and large language models. However, estimating the generalizability of these models to new healthcare settings without extensive validation on external data remains challenging. In experiments across 13 datasets including X-rays, CTs, ECGs, clinical discharge summaries, and lung auscultation data, our results demonstrate that model performance is frequently overestimated by up to 20% on average due to shortcut learning of hidden data acquisition biases (DAB). Shortcut learning refers to a phenomenon in which an AI model learns to solve a task based on spurious correlations present in the data as opposed to features directly related to the task itself. We propose an open source, bias-corrected external accuracy estimate, PEst, that better estimates external accuracy to within 4% on average by measuring and calibrating for DAB-induced shortcut learning.

A Prospective Study of Machine Learning-Assisted Radiation Therapy Planning for Patients Receiving 54 Gy to the Brain.

PURPOSE: The capacity for machine learning (ML) to facilitate radiation therapy (RT) planning for primary brain tumors has not been described. We evaluated ML-assisted RT planning with regard to clinical acceptability, dosimetric outcomes, and planning efficiency for adults and children with primary brain tumors. METHODS AND MATERIALS: In this prospective study, children and adults receiving 54 Gy fractionated RT for a primary brain tumor were enrolled. For each patient, one ML-assisted RT plan was created and compared with 1 or 2 plans created using standard ("manual") planning procedures. Plans were evaluated by the treating oncologist, who was blinded to the method of plan creation. The primary endpoint was the proportion of ML plans that were clinically acceptable for treatment. Secondary endpoints included the frequency with which ML plans were selected as preferable for treatment, and dosimetric differences between ML and manual plans. RESULTS: A total of 116 manual plans and 61 ML plans were evaluated across 61 patients. Ninety-four percent of ML plans and 93% of manual plans were judged to be clinically acceptable (P = 1.0). Overall, the quality of ML plans was similar to manual plans. ML plans comprised 34.5% of all plans evaluated and were selected for treatment in 36.1% of cases (P = .82). Similar tumor target coverage was achieved between both planning methods. Normal brain (brain minus planning target volume) received an average of 1 Gy less mean dose with ML plans (compared with manual plans, P < .001). ML plans required an average of 45.8 minutes less time to create, compared with manual plans (P < .001). CONCLUSIONS: ML-assisted automated planning creates high-quality plans for patients with brain tumors, including children. Plans created with ML assistance delivered slightly less dose to normal brain tissues and can be designed in less time.

Lipidomic-Based Approach to 10 s Classification of Major Pediatric Brain Cancer Types with Picosecond Infrared Laser Mass Spectrometry.

Picosecond infrared laser mass spectrometry (PIRL-MS) is shown, through a retrospective patient tissue study, to differentiate medulloblastoma cancers from pilocytic astrocytoma and two molecular subtypes of ependymoma (PF-EPN-A, ST-EPN-RELA) using laser-extracted lipids profiled with PIRL-MS in 10 s of sampling and analysis time. The average sensitivity and specificity values for this classification, taking genomic profiling data as standard, were 96.41 and 99.54%, and this classification used many molecular features resolvable in 10 s PIRL-MS spectra. Data analysis and liquid chromatography coupled with tandem high-resolution mass spectrometry (LC-MS/MS) further allowed us to reduce the molecular feature list to only 18 metabolic lipid markers most strongly involved in this classification. The identified 'metabolite array' was comprised of a variety of phosphatidic and fatty acids, ceramides, and phosphatidylcholine/ethanolamine and could mediate the above-mentioned classification with average sensitivity and specificity values of 94.39 and 98.78%, respectively, at a 95% confidence in prediction probability threshold. Therefore, a rapid and accurate pathology classification of select pediatric brain cancer types from 10 s PIRL-MS analysis using known metabolic biomarkers can now be available to the neurosurgeon. Based on retrospective mining of 'survival' versus 'extent-of-resection' data, we further identified pediatric cancer types that may benefit from actionable 10 s PIRL-MS pathology feedback. In such cases, aggressiveness of the surgical resection can be optimized in a manner that is expected to benefit the patient's overall or progression-free survival. PIRL-MS is a promising tool to drive such personalized decision-making in the operating theater.

Incremental value of machine learning for risk prediction in tetralogy of Fallot.

OBJECTIVE: Machine learning (ML) can facilitate prediction of major adverse cardiovascular events (MACEs) in repaired tetralogy of Fallot (rTOF). We sought to determine the incremental value of ML above expert clinical judgement for risk prediction in rTOF. METHODS: Adult congenital heart disease (ACHD) clinicians (≥10 years of experience) participated (one cardiac surgeon and four cardiologists (two paediatric and two adult cardiology trained) with expertise in heart failure (HF), electrophysiology, imaging and intervention). Clinicians identified 10 high-yield variables for 5-year MACE prediction (defined as a composite of mortality, resuscitated sudden death, sustained ventricular tachycardia and HF). Risk for MACE (low, moderate or high) was assigned by clinicians blinded to outcome for adults with rTOF identified from an institutional database (n=25 patient reviews conducted by five independent observers). A validated ML model identified 10 variables for risk prediction in the same population. RESULTS: Prediction by ML was similar to the aggregate score of all experts (area under the curve (AUC) 0.85 (95% CI 0.58 to 0.96) vs 0.92 (0.72 to 0.98), p=0.315). Experts with ≥20 years of experience had superior discriminative capacity compared with <20 years (AUC 0.98 (95% CI 0.86 to 0.99) vs 0.80 (0.56 to 0.93), p=0.027). In those with <20 years of experience, ML provided incremental value such that the combined (clinical+ML) AUC approached ≥20 years (AUC 0.85 (95% CI 0.61 to 0.95), p=0.055). CONCLUSIONS: Robust prediction of 5-year MACE in rTOF was achieved using either ML or a multidisciplinary team of ACHD experts. Risk prediction of some clinicians was enhanced by incorporation of ML suggesting that there may be incremental value for ML in select circumstances.

Multi-institutional Prognostic Modeling in Head and Neck Cancer: Evaluating Impact and Generalizability of Deep Learning and Radiomics.

Artificial intelligence (AI) and machine learning (ML) are becoming critical in developing and deploying personalized medicine and targeted clinical trials. Recent advances in ML have enabled the integration of wider ranges of data including both medical records and imaging (radiomics). However, the development of prognostic models is complex as no modeling strategy is universally superior to others and validation of developed models requires large and diverse datasets to demonstrate that prognostic models developed (regardless of method) from one dataset are applicable to other datasets both internally and externally. Using a retrospective dataset of 2,552 patients from a single institution and a strict evaluation framework that included external validation on three external patient cohorts (873 patients), we crowdsourced the development of ML models to predict overall survival in head and neck cancer (HNC) using electronic medical records (EMR) and pretreatment radiological images. To assess the relative contributions of radiomics in predicting HNC prognosis, we compared 12 different models using imaging and/or EMR data. The model with the highest accuracy used multitask learning on clinical data and tumor volume, achieving high prognostic accuracy for 2-year and lifetime survival prediction, outperforming models relying on clinical data only, engineered radiomics, or complex deep neural network architecture. However, when we attempted to extend the best performing models from this large training dataset to other institutions, we observed significant reductions in the performance of the model in those datasets, highlighting the importance of detailed population-based reporting for AI/ML model utility and stronger validation frameworks. We have developed highly prognostic models for overall survival in HNC using EMRs and pretreatment radiological images based on a large, retrospective dataset of 2,552 patients from our institution.Diverse ML approaches were used by independent investigators. The model with the highest accuracy used multitask learning on clinical data and tumor volume.External validation of the top three performing models on three datasets (873 patients) with significant differences in the distributions of clinical and demographic variables demonstrated significant decreases in model performance. Significance: ML combined with simple prognostic factors outperformed multiple advanced CT radiomics and deep learning methods. ML models provided diverse solutions for prognosis of patients with HNC but their prognostic value is affected by differences in patient populations and require extensive validation.

Machine Learning for Prediction of Adverse Cardiovascular Events in Adults With Repaired Tetralogy of Fallot Using Clinical and Cardiovascular Magnetic Resonance Imaging Variables.

BACKGROUND: Existing models for prediction of major adverse cardiovascular events (MACE) after repair of tetralogy of Fallot have been limited by modest predictive capacity and limited applicability to routine clinical practice. We hypothesized that an artificial intelligence model using an array of parameters would enhance 5-year MACE prediction in adults with repaired tetralogy of Fallot. METHODS: A machine learning algorithm was applied to 2 nonoverlapping, institutional databases of adults with repaired tetralogy of Fallot: (1) for model development, a prospectively constructed clinical and cardiovascular magnetic resonance registry; (2) for model validation, a retrospective database comprised of variables extracted from the electronic health record. The MACE composite outcome included mortality, resuscitated sudden death, sustained ventricular tachycardia and heart failure. Analysis was restricted to individuals with MACE or followed ≥5 years. A random forest model was trained using machine learning (n=57 variables). Repeated random sub-sampling validation was sequentially applied to the development dataset followed by application to the validation dataset. RESULTS: We identified 804 individuals (n=312 for development and n=492 for validation). Model prediction (area under the curve [95% CI]) for MACE in the validation dataset was strong (0.82 [0.74-0.89]) with superior performance to a conventional Cox multivariable model (0.63 [0.51-0.75]; P=0.003). Model performance did not change significantly with input restricted to the 10 strongest features (decreasing order of strength: right ventricular end-systolic volume indexed, right ventricular ejection fraction, age at cardiovascular magnetic resonance imaging, age at repair, absolute ventilatory anaerobic threshold, right ventricular end-diastolic volume indexed, ventilatory anaerobic threshold % predicted, peak aerobic capacity, left ventricular ejection fraction, and pulmonary regurgitation fraction; 0.81 [0.72-0.89]; P=0.232). Removing exercise parameters resulted in inferior model performance (0.75 [0.65-0.84]; P=0.002). CONCLUSIONS: In this single-center study, a machine learning-based prediction model comprised of readily available clinical and cardiovascular magnetic resonance imaging variables performed well in an independent validation cohort. Further study will determine the value of this model for risk stratification in adults with repared tetralogy of Fallot.

Comprehensive Cardiovascular Magnetic Resonance Tissue Characterization and Cardiotoxicity in Women With Breast Cancer.

Importance: There is a growing interest in understanding whether cardiovascular magnetic resonance (CMR) myocardial tissue characterization helps identify risk of cancer therapy-related cardiac dysfunction (CTRCD). Objective: To describe changes in CMR tissue biomarkers during breast cancer therapy and their association with CTRCD. Design, Setting, and Participants: This was a prospective, multicenter, cohort study of women with ERBB2 (formerly HER2)-positive breast cancer (stages I-III) who were scheduled to receive anthracycline and trastuzumab therapy with/without adjuvant radiotherapy and surgery. From November 7, 2013, to January 16, 2019, participants were recruited from 3 University of Toronto-affiliated hospitals. Data were analyzed from July 2021 to June 2022. Exposures: Sequential therapy with anthracyclines, trastuzumab, and radiation. Main Outcomes and Measures: CMR, high-sensitivity cardiac troponin I (hs-cTnI), and B-type natriuretic peptide (BNP) measurements were performed before anthracycline treatment, after anthracycline and before trastuzumab treatment, and at 3-month intervals during trastuzumab therapy. CMR included left ventricular (LV) volumes, LV ejection fraction (EF), myocardial strain, early gadolinium enhancement imaging to assess hyperemia (inflammation marker), native/postcontrast T1 mapping (with extracellular volume fraction [ECV]) to assess edema and/or fibrosis, T2 mapping to assess edema, and late gadolinium enhancement (LGE) to assess replacement fibrosis. CTRCD was defined using the Cardiac Review and Evaluation Committee criteria. Fixed-effects models or generalized estimating equations were used in analyses. Results: Of 136 women (mean [SD] age, 51.1 [9.2] years) recruited from 2013 to 2019, 37 (27%) developed CTRCD. Compared with baseline, tissue biomarkers of myocardial hyperemia and edema peaked after anthracycline therapy or 3 months after trastuzumab initiation as demonstrated by an increase in mean (SD) relative myocardial enhancement (baseline, 46.3% [16.8%] to peak, 56.2% [18.6%]), native T1 (1012 [26] milliseconds to 1035 [28] milliseconds), T2 (51.4 [2.2] milliseconds to 52.6 [2.2] milliseconds), and ECV (25.2% [2.4%] to 26.8% [2.7%]), with P <.001 for the entire follow-up. The observed values were mostly within the normal range, and the changes were small and recovered during follow-up. No new replacement fibrosis developed. Increase in T1, T2, and/or ECV was associated with increased ventricular volumes and BNP but not hs-cTnI level. None of the CMR tissue biomarkers were associated with changes in LVEF or myocardial strain. Change in ECV was associated with concurrent and subsequent CTRCD, but there was significant overlap between patients with and without CTRCD. Conclusions and Relevance: In women with ERBB2-positive breast cancer receiving sequential anthracycline and trastuzumab therapy, CMR tissue biomarkers suggest inflammation and edema peaking early during therapy and were associated with ventricular remodeling and BNP elevation. However, the increases in CMR biomarkers were transient, were not associated with LVEF or myocardial strain, and were not useful in identifying traditional CTRCD risk.

Domain adaptation of automated treatment planning from computed tomography to magnetic resonance.

Objective.Machine learning (ML) based radiation treatment planning addresses the iterative and time-consuming nature of conventional inverse planning. Given the rising importance of magnetic resonance (MR) only treatment planning workflows, we sought to determine if an ML based treatment planning model, trained on computed tomography (CT) imaging, could be applied to MR through domain adaptation.Methods.In this study, MR and CT imaging was collected from 55 prostate cancer patients treated on an MR linear accelerator. ML based plans were generated for each patient on both CT and MR imaging using a commercially available model in RayStation 8B. The dose distributions and acceptance rates of MR and CT based plans were compared using institutional dose-volume evaluation criteria. The dosimetric differences between MR and CT plans were further decomposed into setup, cohort, and imaging domain components.Results.MR plans were highly acceptable, meeting 93.1% of all evaluation criteria compared to 96.3% of CT plans, with dose equivalence for all evaluation criteria except for the bladder wall, penile bulb, small and large bowel, and one rectum wall criteria (p< 0.05). Changing the input imaging modality (domain component) only accounted for about half of the dosimetric differences observed between MR and CT plans. Anatomical differences between the ML training set and the MR linac cohort (cohort component) were also a significant contributor.Significance.We were able to create highly acceptable MR based treatment plans using a CT-trained ML model for treatment planning, although clinically significant dose deviations from the CT based plans were observed. Future work should focus on combining this framework with atlas selection metrics to create an interpretable quality assurance QA framework for ML based treatment planning.

Cardiovascular signatures of COVID-19 predict mortality and identify barrier stabilizing therapies.

BACKGROUND: Endothelial cell (EC) activation, endotheliitis, vascular permeability, and thrombosis have been observed in patients with severe coronavirus disease 2019 (COVID-19), indicating that the vasculature is affected during the acute stages of SARS-CoV-2 infection. It remains unknown whether circulating vascular markers are sufficient to predict clinical outcomes, are unique to COVID-19, and if vascular permeability can be therapeutically targeted. METHODS: Prospectively evaluating the prevalence of circulating inflammatory, cardiac, and EC activation markers as well as developing a microRNA atlas in 241 unvaccinated patients with suspected SARS-CoV-2 infection allowed for prognostic value assessment using a Random Forest model machine learning approach. Subsequent ex vivo experiments assessed EC permeability responses to patient plasma and were used to uncover modulated gene regulatory networks from which rational therapeutic design was inferred. FINDINGS: Multiple inflammatory and EC activation biomarkers were associated with mortality in COVID-19 patients and in severity-matched SARS-CoV-2-negative patients, while dysregulation of specific microRNAs at presentation was specific for poor COVID-19-related outcomes and revealed disease-relevant pathways. Integrating the datasets using a machine learning approach further enhanced clinical risk prediction for in-hospital mortality. Exposure of ECs to COVID-19 patient plasma resulted in severity-specific gene expression responses and EC barrier dysfunction, which was ameliorated using angiopoietin-1 mimetic or recombinant Slit2-N. INTERPRETATION: Integration of multi-omics data identified microRNA and vascular biomarkers prognostic of in-hospital mortality in COVID-19 patients and revealed that vascular stabilizing therapies should be explored as a treatment for endothelial dysfunction in COVID-19, and other severe diseases where endothelial dysfunction has a central role in pathogenesis. FUNDING: This work was directly supported by grant funding from the Ted Rogers Center for Heart Research, Toronto, Ontario, Canada and the Peter Munk Cardiac Center, Toronto, Ontario, Canada.

SuPART: supervised projective adapted resonance theory for automatic quality assurance approval of radiotherapy treatment plans.

Radiotherapy is a common treatment modality for the treatment of cancer, where treatments must be carefully designed to deliver appropriate dose to targets while avoiding healthy organs. The comprehensive multi-disciplinary quality assurance (QA) process in radiotherapy is designed to ensure safe and effective treatment plans are delivered to patients. However, the plan QA process is expensive, often time-intensive, and requires review of large quantities of complex data, potentially leading to human error in QA assessment. We therefore develop an automated machine learning algorithm to identify 'acceptable' plans (plans that are similar to historically approved plans) and 'unacceptable' plans (plans that are dissimilar to historically approved plans). This algorithm is a supervised extension of projective adaptive resonance theory, called SuPART, that learns a set of distinctive features, and considers deviations from them indications of unacceptable plans. We test SuPART on breast and prostate radiotherapy datasets from our institution, and find that SuPART outperforms common classification algorithms in several measures of accuracy. When no falsely approved plans are allowed, SuPART can correctly auto-approve 34% of the acceptable breast and 32% of the acceptable prostate plans, and can also correctly reject 53% of the unacceptable breast and 56% of the unacceptable prostate plans. Thus, usage of SuPART to aid in QA could potentially yield significant time savings.

A pilot study of machine-learning based automated planning for primary brain tumours.

PURPOSE: High-quality radiotherapy (RT) planning for children and young adults with primary brain tumours is essential to minimize the risk of late treatment effects. The feasibility of using automated machine-learning (ML) to aid RT planning in this population has not previously been studied. METHODS AND MATERIALS: We developed a ML model that identifies learned relationships between image features and expected dose in a training set of 95 patients with a primary brain tumour treated with focal radiotherapy to a dose of 54 Gy in 30 fractions. This ML method was then used to create predicted dose distributions for 15 previously-treated brain tumour patients across two institutions, as a testing set. Dosimetry to target volumes and organs-at-risk (OARs) were compared between the clinically-delivered (human-generated) plans versus the ML plans. RESULTS: The ML method was able to create deliverable plans in all 15 patients in the testing set. All ML plans were generated within 30 min of initiating planning. Planning target volume coverage with 95% of the prescription dose was attained in all plans. OAR doses were similar across most structures evaluated; mean doses to brain and left temporal lobe were lower in ML plans than manual plans (mean difference to left temporal, - 2.3 Gy, p = 0.006; mean differences to brain, - 1.3 Gy, p = 0.017), whereas mean doses to right cochlea and lenses were higher in ML plans (+ 1.6-2.2 Gy, p < 0.05 for each). CONCLUSIONS: Use of an automated ML method to aid RT planning for children and young adults with primary brain tumours is dosimetrically feasible and can be successfully used to create high-quality 54 Gy RT plans. Further evaluation after clinical implementation is planned.

Understanding machine learning classifier decisions in automated radiotherapy quality assurance.

The complexity of generating radiotherapy treatments demands a rigorous quality assurance (QA) process to ensure patient safety and to avoid clinically significant errors. Machine learning classifiers have been explored to augment the scope and efficiency of the traditional radiotherapy treatment planning QA process. However, one important gap in relying on classifiers for QA of radiotherapy treatment plans is the lack of understanding behind a specific classifier prediction. We develop explanation methods to understand the decisions of two automated QA classifiers: (1) a region of interest (ROI) segmentation/labeling classifier, and (2) a treatment plan acceptance classifier. For each classifier, a local interpretable model-agnostic explanation (LIME) framework and a novel adaption of team-based Shapley values framework are constructed. We test these methods in datasets for two radiotherapy treatment sites (prostate and breast), and demonstrate the importance of evaluating QA classifiers using interpretable machine learning approaches. We additionally develop a notion of explanation consistency to assess classifier performance. Our explanation method allows for easy visualization and human expert assessment of classifier decisions in radiotherapy QA. Notably, we find that our team-based Shapley approach is more consistent than LIME. The ability to explain and validate automated decision-making is critical in medical treatments. This analysis allows us to conclude that both QA classifiers are moderately trustworthy and can be used to confirm expert decisions, though the current QA classifiers should not be viewed as a replacement for the human QA process.

A Qualitative Investigation Into Faith and LGBTQ Identities in a Nonclinical Sample of Readers of the Dear Abby Column: The Dear Abby Project.

ABSTRACT: To better understand the relationship between faith and LGBTQ+ identity, we conducted a qualitative analysis of 86 respondents to a general question posed through the Dear Abby column. Responses were anonymized and analyzed using a grounded theory approach. Analysis revealed six themes, reflecting a diversity of lived experience from community rejection to acceptance, and self-rejection to feelings of acceptance by God. Despite frequent media portrayals of conflict between faith and LGBTQ+ identity, the reality is more complex, and faith and LGBTQ+ identity development can be complementary.

Clinical integration of machine learning for curative-intent radiation treatment of patients with prostate cancer.

Machine learning (ML) holds great promise for impacting healthcare delivery; however, to date most methods are tested in 'simulated' environments that cannot recapitulate factors influencing real-world clinical practice. We prospectively deployed and evaluated a random forest algorithm for therapeutic curative-intent radiation therapy (RT) treatment planning for prostate cancer in a blinded, head-to-head study with full integration into the clinical workflow. ML- and human-generated RT treatment plans were directly compared in a retrospective simulation with retesting (n = 50) and a prospective clinical deployment (n = 50) phase. Consistently throughout the study phases, treating physicians assessed ML- and human-generated RT treatment plans in a blinded manner following a priori defined standardized criteria and peer review processes, with the selected RT plan in the prospective phase delivered for patient treatment. Overall, 89% of ML-generated RT plans were considered clinically acceptable and 72% were selected over human-generated RT plans in head-to-head comparisons. RT planning using ML reduced the median time required for the entire RT planning process by 60.1% (118 to 47 h). While ML RT plan acceptability remained stable between the simulation and deployment phases (92 versus 86%), the number of ML RT plans selected for treatment was significantly reduced (83 versus 61%, respectively). These findings highlight that retrospective or simulated evaluation of ML methods, even under expert blinded review, may not be representative of algorithm acceptance in a real-world clinical setting when patient care is at stake.

Performance stability evaluation of atlas-based machine learning radiation therapy treatment planning in prostate cancer.

Atlas-based machine learning (ML) for radiation therapy (RT) treatment planning is effective at tailoring dose distributions to account for unique patient anatomies by selecting the most appropriate patients from the training database (atlases) to inform dose prediction for new patients. However, variations in clinical practice between the training dataset and a new patient to be planned may impact ML performance by confounding atlas selection. In this study, we simulated various contouring practices in prostate cancer RT to investigate the impact of changing input data on atlas-based ML treatment planning. We generated 225 ML plans for nine bespoke contouring protocol scenarios (reduced target margins, modified organ-at-risk (OAR) definitions, and inclusion of optional OARs less represented in the training database) on 25 patient datasets by applying a single, previously trained and validated ML model for prostate cancer followed by dose mimicking to create a final deliverable plan. ML treatment plans for each scenario were compared to base ML treatment plans that followed a contouring protocol consistent with the model training data. ML performance was evaluated based on atlas distance metrics that are calculated during ML dose prediction. There were significant changes between atlases selected for the base ML treatment plans and treatment plans when planning target volume margins were reduced and/or optional OARs were included. The deliverability of ML predicted dose distributions based on gamma analysis between predicted and mimicked final deliverable dose showed significant differences for seven out of eight scenarios compared with the base ML treatment plans. Overall, there were small but statistically significant dosimetric changes in predicted and mimicked dose with addition of optional OAR contours. This work presents a framework for benchmarking and performance monitoring of ML treatment planning algorithms in the context of evolving clinical practices.

Mass Spectrometry Imaging Reveals a Gradient of Cancer-like Metabolic States in the Vicinity of Cancer Not Seen in Morphometric Margins from Microscopy.

Spatially resolved ambient mass spectrometry imaging methods have gained popularity to characterize cancer sites and their borders using molecular changes in the lipidome. This utility, however, is predicated on metabolic homogeneity at the border, which would create a sharp molecular transition at the morphometric borders. We subjected murine models of human medulloblastoma brain cancer to mass spectrometry imaging, a technique that provides a direct readout of tissue molecular content in a spatially resolved manner. We discovered a distance-dependent gradient of cancer-like lipid molecule profiles in the brain tissue within 1.2 mm of the cancer border, suggesting that a cancer-like state progresses beyond the histologic border, into the healthy tissue. The results were further corroborated using orthogonal liquid chromatography and mass spectrometry (LC-MS) analysis of selected tissue regions subjected to laser capture microdissection. LC-MS/MS analysis for robust identification of the affected molecules implied changes in a number of different lipid classes, some of which are metabolized from the essential docosahexaenoic fatty acid (DHA) present in the interstitial fluid. Metabolic molecular borders are thus not as sharp as morphometric borders, and mass spectrometry imaging can reveal molecular nuances not observed with microscopy. Caution must be exercised in interpreting multimodal imaging results stipulated on a coincidental relationship between metabolic and morphometric borders of cancer, at least within animal models used in preclinical research.

User-controlled pipelines for feature integration and head and neck radiation therapy outcome predictions.

PURPOSE: Precision cancer medicine is dependent on accurate prediction of disease and treatment outcome, requiring integration of clinical, imaging and interventional knowledge. User controlled pipelines are capable of feature integration with varied levels of human interaction. In this work we present two pipelines designed to combine clinical, radiomic (quantified imaging), and RTx-omic (quantified radiation therapy (RT) plan) information for prediction of locoregional failure (LRF) in head and neck cancer (H&N). METHODS: Pipelines were designed to extract information and model patient outcomes based on clinical features, computed tomography (CT) imaging, and planned RT dose volumes. We predict H&N LRF using: 1) a highly user-driven pipeline that leverages modular design and machine learning for feature extraction and model development; and 2) a pipeline with minimal user input that utilizes deep learning convolutional neural networks to extract and combine CT imaging, RT dose and clinical features for model development. RESULTS: Clinical features with logistic regression in our highly user-driven pipeline had the highest precision recall area under the curve (PR-AUC) of 0.66 (0.33-0.93), where a PR-AUC = 0.11 is considered random. CONCLUSIONS: Our work demonstrates the potential to aggregate features from multiple specialties for conditional-outcome predictions using pipelines with varied levels of human interaction. Most importantly, our results provide insights into the importance of data curation and quality, as well as user, data and methodology bias awareness as it pertains to result interpretation in user controlled pipelines.

External validation and transfer learning of convolutional neural networks for computed tomography dental artifact classification.

Quality assurance of data prior to use in automated pipelines and image analysis would assist in safeguarding against biases and incorrect interpretation of results. Automation of quality assurance steps would further improve robustness and efficiency of these methods, motivating widespread adoption of techniques. Previous work by our group demonstrated the ability of convolutional neural networks (CNN) to efficiently classify head and neck (H&N) computed-tomography (CT) images for the presence of dental artifacts (DA) that obscure visualization of structures and the accuracy of Hounsfield units. In this work we demonstrate the generalizability of our previous methodology by validating CNNs on six external datasets, and the potential benefits of transfer learning with fine-tuning on CNN performance. 2112 H&N CT images from seven institutions were scored as DA positive or negative. 1538 images from a single institution were used to train three CNNs with resampling grid sizes of 643, 1283 and 2563. The remaining six external datasets were used in five-fold cross-validation with a data split of 20% training/fine-tuning and 80% validation. The three pre-trained models were each validated using the five-folds of the six external datasets. The pre-trained models also underwent transfer learning with fine-tuning using the 20% training/fine-tuning data, and validated using the corresponding validation datasets. The highest micro-averaged AUC for our pre-trained models across all external datasets occurred with a resampling grid of 2563 (AUC = 0.91 ± 0.01). Transfer learning with fine-tuning improved generalizability when utilizing a resampling grid of 2563 to a micro-averaged AUC of 0.92 ± 0.01. Despite these promising results, transfer learning did not improve AUC when utilizing small resampling grids or small datasets. Our work demonstrates the potential of our previously developed automated quality assurance methods to generalize to external datasets. Additionally, we showed that transfer learning with fine-tuning using small portions of external datasets can be used to fine-tune models for improved performance when large variations in images are present.