Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Ran Gilad-Bachrach has a total of 27 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    A generative approach for semantic auditing of electronic health records

    by Irena Girshovitz, Atai Ambus, Moni Shahar, Ran Gilad-Bachrach

    The reliability of clinical artificial intelligence (AI) depends on high-quality data, yet Electronic Health Records are often inconsistent with existing scientific knowledge. Current quality assessments are limited: they either focus on syntax or rely on labor-intensive manual rules to capture semantic nuances. To overcome these scalability barriers, we propose Medical Data Pecking, a methodology that adopts software unit testing principles for medical data validation. It introduces Semantic Data Coverage, employing Large Language Models to generate context-aware tests that “peck" for inconsistencies between observed data and epidemiological evidence. To demonstrate this methodology, we implemented a reference tool using a Retrieval-Augmented Generation architecture that synthesizes medical literature into executable code. When applied to three datasets, this implementation generated dozens of tests per cohort, identifying discrepancies between observed distributions and epidemiological priors. These discrepancies encompass both genuine data inconsistencies and expected cohort-selection effects. This work provides an initial framework for scalable semantic auditing, shifting assurance from manual rules to the generative and context-sensitive verification required for trustworthy AI.

  • Open Access

    Medical Data Pecking: A Context-Aware Approach for Automated Quality Evaluation of Structured Medical Data

    by Irena Girshovitz, Atai Ambus, Moni Shahar, Ran Gilad-Bachrach

    Background: The use of Electronic Health Records (EHRs) for epidemiological studies and artificial intelligence (AI) training is increasing rapidly. The reliability of the results depends on the accuracy and completeness of EHR data. However, EHR data often contain significant quality issues, including misrepresentations of subpopulations, biases, and systematic errors, as they are primarily collected for clinical and billing purposes. Existing quality assessment methods remain insufficient, lacking systematic procedures to assess data fitness for research. Methods: We present the Medical Data Pecking approach, which adapts unit testing and coverage concepts from software engineering to identify data quality concerns. We demonstrate our approach using the Medical Data Pecking Tool (MDPT), which consists of two main components: (1) an automated test generator that uses large language models and grounding techniques to create a test suite from data and study descriptions, and (2) a data testing framework that executes these tests, reporting potential errors and coverage. Results: We evaluated MDPT on three datasets: All of Us (AoU), MIMIC-III, and SyntheticMass, generating 55-73 tests per cohort across four conditions. These tests correctly identified 20-43 non-aligned or non-conforming data issues. We present a detailed analysis of the LLM-generated test suites in terms of reference grounding and value accuracy. Conclusion: Our approach incorporates external medical knowledge to enable context-sensitive data quality testing as part of the data analysis workflow to improve the validity of its outcomes. Our approach tackles these challenges from a quality assurance perspective, laying the foundation for further development such as additional data modalities and improved grounding methods.

  • Open Access

    Impact of Long-COVID in children: a large cohort study

    by Ziv Hersh, Yiska Loewenberg Weisband, Ariel Bogan, Adir Leibovich, Uri Obolski, Daniel Nevo, Ran Gilad-Bachrach

    Background: The impact of long-term Coronavirus disease 2019 (COVID-19) on the pediatric population is still not well understood. This study was designed to estimate the magnitude of COVID-19 long-term morbidity 3–6 months after the date of diagnosis. Methods: A retrospective study of all Clalit Health Services members in Israel aged 1–16 years who tested positive for SARS-CoV-2 between April 1, 2020 and March 31, 2021. Controls, who had no previous diagnosis of COVID-19, were one-to-one matched to 65,548 COVID-19-positive children and teens, and were assigned the infection dates of their matches as their index date. Matching included age, sex, socio-economic score, and societal sector. Individuals were excluded from the study if they had severe medical conditions before the diagnosis such as cancer, diabetes, chronic respiratory diseases, and/or abnormal physiological development. Generalized Estimating Equations were used to estimate the associations between COVID-19 and the use of medical services. The analysis focused on the 3–6 months after the infection date. Adjustments were made for demographics and for the use of medical services 6–12 and 3–6 months before the infection date. The latter was necessary because of observed disparities in medical service utilization between the groups before the COVID-19 diagnosis, despite the matching process. Results: Statistically significant differences were only found for referrals for mental health services [adjusted relative-risk (RR) 1·51, 95%CI 1·15 − 1·96; adjusted risk-difference (RD) 0·001, 95%CI 0·0006 − 0·002], and medication prescriptions of any kind (RR 1·03, 95%CI 1·01–1·06; RD 0·01 95%CI 0·004 − 0·02). Conclusions: The significant increase in medication prescriptions and mental health service referrals support the hypothesis that COVID-19 is associated with long-lasting morbidities in children and adolescents aged 1–16 years. However, the risk difference in both instances was small, suggesting a minor impact on medical services.


  • Open Access

    Inherent Inconsistencies of Feature Importance

    by Nimrod Harel, Uri Obolski, Ran Gilad-Bachrach

    The rapid advancement and widespread adoption of machine learning-driven technologies have underscored the practical and ethical need for creating interpretable artificial intelligence systems. Feature importance, a method that assigns scores to the contribution of individual features on prediction outcomes, seeks to bridge this gap as a tool for enhancing human comprehension of these systems. Feature importance serves as an explanation of predictions in diverse contexts, whether by providing a global interpretation of a phenomenon across the entire dataset or by offering a localized explanation for the outcome of a specific data point. Furthermore, feature importance is being used both for explaining models and for identifying plausible causal relations in the data, independently from the model. However, it is worth noting that these various contexts have traditionally been explored in isolation, with limited theoretical foundations. This paper presents an axiomatic framework designed to establish coherent relationships among the different contexts of feature importance scores. Notably, our work unveils a surprising conclusion: when we combine the proposed properties with those previously outlined in the literature, we demonstrate the existence of an inconsistency. This inconsistency highlights that certain essential properties of feature importance scores cannot coexist harmoniously within a single framework.


  • Open Access

    Lost in Translation: The Limits of Explainability in AI

    by Hofit Wasserman-Rozen, Ran Gilad-Bachrach, Niva Elkin-Koren

    As artificial intelligence becomes more prevalent, regulators are increasingly turning to legal measures, like “a right to explanation” to protect against potential risks raised by AI systems. However, are eXplainable AI (XAI) tools - the artificial intelligence tools that provide such explanations – up for the task? This paper critically examines XAI’s potential to facilitate the right to explanation by applying the prism of explanation’s role in law to different stakeholders. Inspecting the underlying functions of reason-giving reveals different objectives for each of the stakeholders involved. From the perspective of a decision-subject, reason-giving facilitates due process and acknowledges human agency. From a decision-maker’s perspective, reason-giving contributes to improving the quality of the decisions themselves. From an ecosystem perspective, reason-giving may strengthen the authority of the decision-making system toward different stakeholders by promoting accountability and legitimacy, and by providing better guidance. Applying this analytical framework to XAI’s generated explanations reveals that XAI fails to fulfill the underlying objectives of the right to explanation from the perspective of both the decision-subject and the decision-maker. In contrast, XAI is found to be extremely well-suited to fulfil the underlying functions of reason-giving from an ecosystems’ perspective, namely, strengthening the authority of the decision-making system. However, lacking all other virtues, this isolated ability may be misused or abused, eventually harming XAI’s intended human audience. The disparity between human decision-making and automated decisions makes XAI an insufficient and even a risky tool, rather than serving as a guardian of human rights. After conducting a rigorous analysis of these ramifications, this paper concludes by urging regulators and the XAI community to reconsider the pursuit of explainability and the right to explanation of AI systems.


  • Open Access

    Gait Detection from a Wrist-Worn Sensor Using Machine Learning Methods: A Daily Living Study in Older Adults and People with Parkinson’s Disease

    by Yonatan E. Brand, Dafna Schwartz, Eran Gazit, Aron S. Buchman, Ran Gilad-Bachrach, Jeffrey M. Hausdorff

    Remote assessment of the gait of older adults (OAs) during daily living using wrist-worn sensors has the potential to augment clinical care and mobility research. However, hand movements can degrade gait detection from wrist-sensor recordings. To address this challenge, we developed an anomaly detection algorithm and compared its performance to four previously published gait detection algorithms. Multiday accelerometer recordings from a wrist-worn and lower-back sensor (i.e., the “gold-standard” reference) were obtained in 30 OAs, 60% with Parkinson’s disease (PD). The area under the receiver operator curve (AUC) and the area under the precision–recall curve (AUPRC) were used to evaluate the performance of the algorithms. The anomaly detection algorithm obtained AUCs of 0.80 and 0.74 for OAs and PD, respectively, but AUPRCs of 0.23 and 0.31 for OAs and PD, respectively. The best performing detection algorithm, a deep convolutional neural network (DCNN), exhibited high AUCs (i.e., 0.94 for OAs and 0.89 for PD) but lower AUPRCs (i.e., 0.66 for OAs and 0.60 for PD), indicating trade-offs between precision and recall. When choosing a classification threshold of 0.9 (i.e., opting for high precision) for the DCNN algorithm, strong correlations (r > 0.8) were observed between daily living walking time estimates based on the lower-back (reference) sensor and the wrist sensor. Further, gait quality measures were significantly different in OAs and PD compared to healthy adults. These results demonstrate that daily living gait can be quantified using a wrist-worn sensor.


  • Open Access

    Detecting Impending Stroke From Cognitive Traits Evident in Internet Searches: Analysis of Archival Data

    by Sigal Shaklai, Ran Gilad-Bachrach, Elad Yom-Tov, Naftali Stern

    Background: Cerebrovascular disease is a leading cause of mortality and disability. Common risk assessment tools for stroke are based on the Framingham equation, which relies on traditional cardiovascular risk factors to predict an acute event in the near decade. However, no tools are currently available to predict a near/impending stroke, which might alert patients at risk to seek immediate preventive action (eg, anticoagulants for atrial fibrillation, control of hypertension). Objective: Here, we propose that an algorithm based on internet search queries can identify people at increased risk for a near stroke event. Methods: We analyzed queries submitted to the Bing search engine by 285 people who self-identified as having undergone a stroke event and 1195 controls with regard to attributes previously shown to reflect cognitive function. Controls included random people 60 years and above, or those of similar age who queried for one of nine control conditions. Results: The model performed well against all comparator groups with an area under the receiver operating characteristic curve of 0.985 or higher and a true positive rate (at a 1% false-positive rate) above 80% for separating patients from each of the controls. The predictive power rose as the stroke date approached and if data were acquired beginning 120 days prior to the event. Good prediction accuracy was obtained for a prospective cohort of users collected 1 year later. The most predictive attributes of the model were associated with cognitive function, including the use of common queries, repetition of queries, appearance of spelling mistakes, and number of queries per session. Conclusions: The proposed algorithm offers a screening test for a near stroke event. After clinical validation, this algorithm may enable the administration of rapid preventive intervention. Moreover, it could be applied inexpensively, continuously, and on a large scale with the aim of reducing stroke events.


  • Open Access

    Detecting Impending Stroke From Cognitive Traits Evident in Internet Searches: Analysis of Archival Data (Preprint)

    by Sigal Shaklai, Ran Gilad-Bachrach, Elad Yom-Tov, Naftali Stern

    BACKGROUND Cerebrovascular disease is a leading cause of mortality and disability. Common risk assessment tools for stroke are based on the Framingham equation, which relies on traditional cardiovascular risk factors to predict an acute event in the near decade. However, no tools are currently available to predict a near/impending stroke, which might alert patients at risk to seek immediate preventive action (eg, anticoagulants for atrial fibrillation, control of hypertension). OBJECTIVE Here, we propose that an algorithm based on internet search queries can identify people at increased risk for a near stroke event. METHODS We analyzed queries submitted to the Bing search engine by 285 people who self-identified as having undergone a stroke event and 1195 controls with regard to attributes previously shown to reflect cognitive function. Controls included random people 60 years and above, or those of similar age who queried for one of nine control conditions. RESULTS The model performed well against all comparator groups with an area under the receiver operating characteristic curve of 0.985 or higher and a true positive rate (at a 1% false-positive rate) above 80% for separating patients from each of the controls. The predictive power rose as the stroke date approached and if data were acquired beginning 120 days prior to the event. Good prediction accuracy was obtained for a prospective cohort of users collected 1 year later. The most predictive attributes of the model were associated with cognitive function, including the use of common queries, repetition of queries, appearance of spelling mistakes, and number of queries per session. CONCLUSIONS The proposed algorithm offers a screening test for a near stroke event. After clinical validation, this algorithm may enable the administration of rapid preventive intervention. Moreover, it could be applied inexpensively, continuously, and on a large scale with the aim of reducing stroke events.

  • Open Access

    Marginal Contribution Feature Importance -- an Axiomatic Approach for The Natural Case

    by Amnon Catav, Boyang Fu, Jason Ernst, Sriram Sankararaman, Ran Gilad-Bachrach

    When training a predictive model over medical data, the goal is sometimes to gain insights about a certain disease. In such cases, it is common to use feature importance as a tool to highlight significant factors contributing to that disease. As there are many existing methods for computing feature importance scores, understanding their relative merits is not trivial. Further, the diversity of scenarios in which they are used lead to different expectations from the feature importance scores. While it is common to make the distinction between local scores that focus on individual predictions and global scores that look at the contribution of a feature to the model, another important division distinguishes model scenarios, in which the goal is to understand predictions of a given model from natural scenarios, in which the goal is to understand a phenomenon such as a disease. We develop a set of axioms that represent the properties expected from a feature importance function in the natural scenario and prove that there exists only one function that satisfies all of them, the Marginal Contribution Feature Importance (MCI). We analyze this function for its theoretical and empirical properties and compare it to other feature importance scores. While our focus is the natural scenario, we suggest that our axiomatic approach could be carried out in other scenarios too.