Remote, tablet-based assessment of gaze following: a nationwide infant twin study.

Introduction: Much of our understanding of infant psychological development relies on an in-person, laboratory-based assessment. This limits research generalizability, scalability, and equity in access. One solution is the development of new, remotely deployed assessment tools that do not require real-time experimenter supervision. Methods: The current nationwide (Sweden) infant twin study assessed participants remotely via their caregiver's tablets (N = 104, ages 3 to 17 months). To anchor our findings in previous research, we used a gaze-following task where experimental and age effects are well established. Results: Closely mimicking results from conventional eye tracking, we found that a full head movement elicited more gaze following than isolated eye movements. Furthermore, predictably, we found that older infants followed gaze more frequently than younger infants. Finally, while we found no indication of genetic contributions to gaze-following accuracy, the latency to disengage from the gaze cue and orient toward a target was significantly more similar in monozygotic twins than in dizygotic twins, an indicative of heritability. Discussion: Together, these results highlight the potential of remote assessment of infants' psychological development, which can improve generalizability, inclusion, and scalability in developmental research.

The Selective Social Attention task in children with autism spectrum disorder: Results from the Autism Biomarkers Consortium for Clinical Trials (ABC-CT) feasibility study.

The Selective Social Attention (SSA) task is a brief eye-tracking task involving experimental conditions varying along socio-communicative axes. Traditionally the SSA has been used to probe socially-specific attentional patterns in infants and toddlers who develop autism spectrum disorder (ASD). This current work extends these findings to preschool and school-age children. Children 4- to 12-years-old with ASD (N = 23) and a typically-developing comparison group (TD; N = 25) completed the SSA task as well as standardized clinical assessments. Linear mixed models examined group and condition effects on two outcome variables: percent of time spent looking at the scene relative to scene presentation time (%Valid), and percent of time looking at the face relative to time spent looking at the scene (%Face). Age and IQ were included as covariates. Outcome variables' relationships to clinical data were assessed via correlation analysis. The ASD group, compared to the TD group, looked less at the scene and focused less on the actress' face during the most socially-engaging experimental conditions. Additionally, within the ASD group, %Face negatively correlated with SRS total T-scores with a particularly strong negative correlation with the Autistic Mannerism subscale T-score. These results highlight the extensibility of the SSA to older children with ASD, including replication of between-group differences previously seen in infants and toddlers, as well as its ability to capture meaningful clinical variation within the autism spectrum across a wide developmental span inclusive of preschool and school-aged children. The properties suggest that the SSA may have broad potential as a biomarker for ASD.

Memory characteristics in mesial temporal lobe epilepsy: Insights from an eye tracking memory game and neuropsychological assessments.

AIMS: To compare different patterns of memory impairment in patients with two subtypes of mesial temporal lobe epilepsy (MTLE) and healthy controls. METHODS: Thirty-five healthy controls and 41 patients with MTLE were recruited, of which 25 patients were diagnosed as hippocampal sclerosis (HS-MTLE), and the rest 16 patients were lesion-negative (MRI-neg MTLE). Participants completed the Wechsler memory assessment and a short-term memory game on an automated computer-based memory assessment platform with an eye tracker. RESULTS: Both the MRI-neg MTLE and HS-MTLE groups took longer time to complete the short-term memory game than healthy controls (p < 0.001, Cohen's d = 1.087; p = 0.047, Cohen's d = 0.787). During the memory encoding phase, the MRI-neg MTLE group spent significantly shorter time than healthy controls on the difficult levels with three (p = 0.004, Cohen's d = 0.993) and four targets (p = 0.016, Cohen's d = 0.858). During the memory decoding phase, the HS-MTLE group spent less time looking on the targets compared to controls when recalling and finding four targets (p = 0.004, Cohen's d = -0.793), while the MRI-neg MTLE group spent significantly longer time on the distractors and shorter time on the region of interests (ROIs) for all difficulty levels (all p < 0.05) than controls. Furthermore, the eye tracking data were correlated with the scores of the Wechsler Memory Scale after Bonferroni correction (p < 0.05). CONCLUSION: Patients with MRI-neg MTLE demonstrate impaired memory mostly due to attention deficits, while those with HS-MTLE show memory impairment with relative sparing of attention. Eye tracking technology has the potential of facilitating the investigation of the mechanism of memory defect in MTLE and can serve as a supplementary neuropsychological tool for clinical diagnosis and long-term monitoring.

A functional model for studying common trends across trial time in eye tracking experiments.

Eye tracking (ET) experiments commonly record the continuous trajectory of a subject's gaze on a two-dimensional screen throughout repeated presentations of stimuli (referred to as trials). Even though the continuous path of gaze is recorded during each trial, commonly derived outcomes for analysis collapse the data into simple summaries, such as looking times in regions of interest, latency to looking at stimuli, number of stimuli viewed, number of fixations or fixation length. In order to retain information in trial time, we utilize functional data analysis (FDA) for the first time in literature in the analysis of ET data. More specifically, novel functional outcomes for ET data, referred to as viewing profiles, are introduced that capture the common gazing trends across trial time which are lost in traditional data summaries. Mean and variation of the proposed functional outcomes across subjects are then modeled using functional principal components analysis. Applications to data from a visual exploration paradigm conducted by the Autism Biomarkers Consortium for Clinical Trials showcase the novel insights gained from the proposed FDA approach, including significant group differences between children diagnosed with autism and their typically developing peers in their consistency of looking at faces early on in trial time.

VSGD-Net: Virtual Staining Guided Melanocyte Detection on Histopathological Images.

Detection of melanocytes serves as a critical prerequisite in assessing melanocytic growth patterns when diagnosing melanoma and its precursor lesions on skin biopsy specimens. However, this detection is challenging due to the visual similarity of melanocytes to other cells in routine Hematoxylin and Eosin (H&E) stained images, leading to the failure of current nuclei detection methods. Stains such as Sox10 can mark melanocytes, but they require an additional step and expense and thus are not regularly used in clinical practice. To address these limitations, we introduce VSGD-Net, a novel detection network that learns melanocyte identification through virtual staining from H&E to Sox10. The method takes only routine H&E images during inference, resulting in a promising approach to support pathologists in the diagnosis of melanoma. To the best of our knowledge, this is the first study that investigates the detection problem using image synthesis features between two distinct pathology stainings. Extensive experimental results show that our proposed model outperforms state-of-the-art nuclei detection methods for melanocyte detection. The source code and pre-trained model are available at: https://github.com/kechunl/VSGD-Net.

Improving the Diagnosis of Skin Biopsies Using Tissue Segmentation.

Invasive melanoma, a common type of skin cancer, is considered one of the deadliest. Pathologists routinely evaluate melanocytic lesions to determine the amount of atypia, and if the lesion represents an invasive melanoma, its stage. However, due to the complicated nature of these assessments, inter- and intra-observer variability among pathologists in their interpretation are very common. Machine-learning techniques have shown impressive and robust performance on various tasks including healthcare. In this work, we study the potential of including semantic segmentation of clinically important tissue structure in improving the diagnosis of skin biopsy images. Our experimental results show a 6% improvement in F-score when using whole slide images along with epidermal nests and cancerous dermal nest segmentation masks compared to using whole-slide images alone in training and testing the diagnosis pipeline.

Automated quantitative assessment of amorphous calcifications: Towards improved malignancy risk stratification.

BACKGROUND: Amorphous calcifications noted on mammograms (i.e., small and indistinct calcifications that are difficult to characterize) are associated with high diagnostic uncertainty, often leading to biopsies. Yet, only 20% of biopsied amorphous calcifications are cancer. We present a quantitative approach for distinguishing between benign and actionable (high-risk and malignant) amorphous calcifications using a combination of local textures, global spatial relationships, and interpretable handcrafted expert features. METHOD: Our approach was trained and validated on a set of 168 2D full-field digital mammography exams (248 images) from 168 patients. Within these 248 images, we identified 276 image regions with segmented amorphous calcifications and a biopsy-confirmed diagnosis. A set of local (radiomic and region measurements) and global features (distribution and expert-defined) were extracted from each image. Local features were grouped using an unsupervised k-means clustering algorithm. All global features were concatenated with clustered local features and used to train a LightGBM classifier to distinguish benign from actionable cases. RESULTS: On the held-out test set of 60 images, our approach achieved a sensitivity of 100%, specificity of 35%, and a positive predictive value of 38% when the decision threshold was set to 0.4. Given that all of the images in our test set resulted in a recommendation of a biopsy, the use of our algorithm would have identified 15 images (25%) that were benign, potentially reducing the number of breast biopsies. CONCLUSIONS: Quantitative analysis of full-field digital mammograms can extract subtle shape, texture, and distribution features that may help to distinguish between benign and actionable amorphous calcifications.

The autism biomarkers consortium for clinical trials: evaluation of a battery of candidate eye-tracking biomarkers for use in autism clinical trials.

BACKGROUND: Eye tracking (ET) is a powerful methodology for studying attentional processes through quantification of eye movements. The precision, usability, and cost-effectiveness of ET render it a promising platform for developing biomarkers for use in clinical trials for autism spectrum disorder (ASD). METHODS: The autism biomarkers consortium for clinical trials conducted a multisite, observational study of 6-11-year-old children with ASD (n = 280) and typical development (TD, n = 119). The ET battery included: Activity Monitoring, Social Interactive, Static Social Scenes, Biological Motion Preference, and Pupillary Light Reflex tasks. A priori, gaze to faces in Activity Monitoring, Social Interactive, and Static Social Scenes tasks were aggregated into an Oculomotor Index of Gaze to Human Faces (OMI) as the primary outcome measure. This work reports on fundamental biomarker properties (data acquisition rates, construct validity, six-week stability, group discrimination, and clinical relationships) derived from these assays that serve as a base for subsequent development of clinical trial biomarker applications. RESULTS: All tasks exhibited excellent acquisition rates, met expectations for construct validity, had moderate or high six-week stabilities, and highlighted subsets of the ASD group with distinct biomarker performance. Within ASD, higher OMI was associated with increased memory for faces, decreased autism symptom severity, and higher verbal IQ and pragmatic communication skills. LIMITATIONS: No specific interventions were administered in this study, limiting information about how ET biomarkers track or predict outcomes in response to treatment. This study did not consider co-occurrence of psychiatric conditions nor specificity in comparison with non-ASD special populations, therefore limiting our understanding of the applicability of outcomes to specific clinical contexts-of-use. Research-grade protocols and equipment were used; further studies are needed to explore deployment in less standardized contexts. CONCLUSIONS: All ET tasks met expectations regarding biomarker properties, with strongest performance for tasks associated with attention to human faces and weakest performance associated with biological motion preference. Based on these data, the OMI has been accepted to the FDA's Biomarker Qualification program, providing a path for advancing efforts to develop biomarkers for use in clinical trials.

Memory Deficit in Patients With Temporal Lobe Epilepsy: Evidence From Eye Tracking Technology.

Objective: To explore quantitative measurements of the visual attention and neuroelectrophysiological relevance of memory deficits in temporal lobe epilepsy (TLE) by eye tracking and electroencephalography (EEG). Methods: Thirty-four TLE patients and twenty-eight healthy controls were invited to complete neurobehavioral assessments, cognitive oculomotor tasks, and 24-h video EEG (VEEG) recordings using an automated computer-based memory assessment platform with an eye tracker. Visit counts, visit time, and time of first fixation on areas of interest (AOIs) were recorded and analyzed in combination with interictal epileptic discharge (IED) characteristics from the bilateral temporal lobes. Results: The TLE patients had significantly worse Wechsler Digit Span scores [F(1, 58) = 7.49, p = 0.008]. In the Short-Term Memory Game with eye tracking, TLE patients took a longer time to find the memorized items [F(1, 57) = 17.30, p < 0.001]. They had longer first fixation [F(1, 57) = 4.06, p = 0.049] and more visit counts [F(1, 57) = 7.58, p = 0.008] on the target during the recall. Furthermore, the performance of the patients in the Digit Span task was negatively correlated with the total number of IEDs [r(28) = -0.463, p = 0.013] and the number of spikes per sleep cycle [r(28) = -0.420, p = 0.026]. Conclusion: Eye tracking appears to be a quantitative, objective measure of memory evaluation, demonstrating memory retrieval deficits but preserved visual attention in TLE patients. Nocturnal temporal lobe IEDs are closely associated with memory performance, which might be the electrophysiological mechanism for memory impairment in TLE.

Radiogenomic modeling predicts survival-associated prognostic groups in glioblastoma.

BACKGROUND: Combined whole-exome sequencing (WES) and somatic copy number alteration (SCNA) information can separate isocitrate dehydrogenase (IDH)1/2-wildtype glioblastoma into two prognostic molecular subtypes, which cannot be distinguished by epigenetic or clinical features. The potential for radiographic features to discriminate between these molecular subtypes has yet to be established. METHODS: Radiologic features (n = 35 340) were extracted from 46 multisequence, pre-operative magnetic resonance imaging (MRI) scans of IDH1/2-wildtype glioblastoma patients from The Cancer Imaging Archive (TCIA), all of whom have corresponding WES/SCNA data. We developed a novel feature selection method that leverages the structure of extracted MRI features to mitigate the dimensionality challenge posed by the disparity between a large number of features and the limited patients in our cohort. Six traditional machine learning classifiers were trained to distinguish molecular subtypes using our feature selection method, which was compared to least absolute shrinkage and selection operator (LASSO) feature selection, recursive feature elimination, and variance thresholding. RESULTS: We were able to classify glioblastomas into two prognostic subgroups with a cross-validated area under the curve score of 0.80 (±0.03) using ridge logistic regression on the 15-dimensional principle component analysis (PCA) embedding of the features selected by our novel feature selection method. An interrogation of the selected features suggested that features describing contours in the T2 signal abnormality region on the T2-weighted fluid-attenuated inversion recovery (FLAIR) MRI sequence may best distinguish these two groups from one another. CONCLUSIONS: We successfully trained a machine learning model that allows for relevant targeted feature extraction from standard MRI to accurately predict molecularly-defined risk-stratifying IDH1/2-wildtype glioblastoma patient groups.

Classifying Breast Histopathology Images with a Ductal Instance-Oriented Pipeline.

In this study, we propose the Ductal Instance-Oriented Pipeline (DIOP) that contains a duct-level instance segmentation model, a tissue-level semantic segmentation model, and three-levels of features for diagnostic classification. Based on recent advancements in instance segmentation and the Mask RCNN model, our duct-level segmenter tries to identify each ductal individual inside a microscopic image; then, it extracts tissue-level information from the identified ductal instances. Leveraging three levels of information obtained from these ductal instances and also the histopathology image, the proposed DIOP outperforms previous approaches (both feature-based and CNN-based) in all diagnostic tasks; for the four-way classification task, the DIOP achieves comparable performance to general pathologists in this unique dataset. The proposed DIOP only takes a few seconds to run in the inference time, which could be used interactively on most modern computers. More clinical explorations are needed to study the robustness and generalizability of this system in the future.

MLCD: A Unified Software Package for Cancer Diagnosis.

PURPOSE: Machine Learning Package for Cancer Diagnosis (MLCD) is the result of a National Institutes of Health/National Cancer Institute (NIH/NCI)-sponsored project for developing a unified software package from state-of-the-art breast cancer biopsy diagnosis and machine learning algorithms that can improve the quality of both clinical practice and ongoing research. METHODS: Whole-slide images of 240 well-characterized breast biopsy cases, initially assembled under R01 CA140560, were used for developing the algorithms and training the machine learning models. This software package is based on the methodology developed and published under our recent NIH/NCI-sponsored research grant (R01 CA172343) for finding regions of interest (ROIs) in whole-slide breast biopsy images, for segmenting ROIs into histopathologic tissue types and for using this segmentation in classifiers that can suggest final diagnoses. RESULT: The package provides an ROI detector for whole-slide images and modules for semantic segmentation into tissue classes and diagnostic classification into 4 classes (benign, atypia, ductal carcinoma in situ, invasive cancer) of the ROIs. It is available through the GitHub repository under the Massachusetts Institute of Technology license and will later be distributed with the Pathology Image Informatics Platform system. A Web page provides instructions for use. CONCLUSION: Our tools have the potential to provide help to other cancer researchers and, ultimately, to practicing physicians and will motivate future research in this field. This article describes the methodology behind the software development and gives sample outputs to guide those interested in using this package.