PPPrzemysław Piterak

Przemysław Piterak

*This is an algorithmically generated author record based on publicly available information.

Przemysław Piterak has a total of 8 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    DEEP LEARNING–ASSISTED FRACTURE DETECTION ON PLAIN RADIOGRAPHS: DIAGNOSTIC ACCURACY, CLINICAL INTEGRATION, AND FUTURE DIRECTIONS

    by Monika Szlachta-Gubernat, Nicol Baran, Marta Bajkowska-Piterak, Adrianna Buż, Przemysław Piterak, Damian Danilczuk, Wiktor Adamiec, Jerzy Buszko, Michał Dyś

    Introduction and Purpose. Fractures are a major global health burden, and a non-trivial fraction of fractures presenting to emergency and urgent care settings are initially missed on plain radiographs, causing diagnostic delay and avoidable morbidity. This review appraises the current state of deep learning for fracture detection on conventional radiography, addressing standalone accuracy, the effect of algorithmic assistance on readers, and the conditions of clinical integration. Brief Description of the State of Knowledge. Across recent meta-analyses, pooled sensitivity and specificity for fracture detection on radiographs cluster in the low-to-mid nineties, with summary AUROC approaching 0.97 in the largest synthesis. The literature is unevenly distributed across the skeleton: the wrist and the hip dominate the evidence base, the pediatric elbow is moderately developed, and other regions remain represented by only a handful of studies. Algorithmic assistance lifts reader sensitivity by several percentage points without a specificity penalty, with the largest gains in clinicians who have less musculoskeletal imaging experience. Summary. Deep learning is mature for narrow fracture-detection tasks in the appendicular skeleton, but not yet for broad deployment claims. External validation across institutions and vendors, prospective evaluation tied to clinical endpoints, and standards-based integration into PACS and reporting infrastructure are the conditions that would move these tools from accuracy benchmarks toward measurable patient benefit. The clearest near-term value lies in decision support that narrows the diagnostic gap for less experienced readers.

  • Open Access

    AI-DRIVEN OCULOMICS: RETINAL IMAGING BIOMARKERS FOR SYSTEMIC DISEASE - FROM CARDIOVASCULAR RISK TO NEURODEGENERATION AND FOUNDATION MODELS

    by Adrianna Buż, Nicol Baran, Marta Bajkowska - Piterak, Jerzy Buszko, Damian Danilczuk, Wiktor Adamiec, Monika Szlachta-Gubernat, Przemysław Piterak, Michał Dyś

    Introduction and aim: The retina offers a unique, non-invasive window into systemic vascular and neurological health. Deep-learning algorithms applied to retinal photographs and optical coherence tomography (OCT) can predict cardiovascular events, detect chronic kidney disease, estimate biological age, and identify neurodegenerative diseases - a field termed oculomics. This review synthesizes current evidence on AI-driven oculomics across cardiovascular, cardio-renal-metabolic, neurodegenerative, and aging domains. Materials and methods: A narrative review was conducted in PubMed, Scopus, and Google Scholar for studies published 2018-2026. Original research, validation studies, and major reviews on AI-based prediction of systemic disease from retinal imaging were included. Results: Deep-learning models trained on fundus photographs predict major adverse cardiovascular events and ten-year atherosclerotic cardiovascular disease risk with areas under the receiver operating characteristic curve (AUROCs) of 0.70-0.89. Chronic kidney disease detection achieves AUROCs of 0.85-0.94. Alzheimer’s disease classifiers reach AUROCs of 0.93 in development cohorts and 0.73-0.91 in external validation, while Parkinson’s disease detection achieves AUROCs of 0.77-0.92. Retinal age-gap independently predicts mortality. Self-supervised foundation models (RETFound, VisionFM, EyeFound, RETFound-Green, MIRAGE, V-JEPA) show state-of-the-art performance, with volumetric OCT (V-JEPA) reaching AUROCs of 0.94. Conclusions: AI-driven oculomics is a rapidly maturing field with strong translational potential. Despite encouraging accuracies, prospective validation, algorithmic bias, regulatory approval, and clinical integration remain key challenges. Future work should prioritize multi-ethnic cohort studies, cost-effectiveness analyses, and regulatory pathways.

  • Open Access

    AMYOTROPHIC LATERAL SCLEROSIS: RISK FACTORS, DIAGNOSTIC DELAYS, MODERN DIAGNOSTIC METHODS, AND FUTURE POSSIBILITIES- LITERATURE REVIEW

    by Nicol Baran, Marta Bajkowska-Piterak, Monika Szlachta-Gubernat, Adrianna Buż, Przemysław Piterak, Damian Danilczuk, Wiktor Adamiec, Jerzy Buszko, Michał Dyś

    Introduction and purpose. Amyotrophic Lateral Sclerosis (ALS) is a progressive neurodegenerative disease affecting motor neurons. It affects about 4-5 persons per 100000 population and is characterized by varied clinical presentation and short survival period. ALS remains incurable; however, patients can receive multidisciplinary care and disease-modifying drugs. This article focuses on ALS protective and risk factors, diagnostic delays, modern diagnostic methods, and a future prospects for enhancing the diagnostic process. Material and methods. The review is based on research of articles published between 2020 and 2026 in databases such as Google Schoolar and PubMed. English-language publications were selected by the authors. Results and Conclusions. Head trauma, military service, chemicals, heavy metals, electrical injuries, magnetic field exposure, history of stroke, and hypertension are associated with higher risk of ALS. Diabetes mellitus, antidiabetic and antihypertensive drugs, higher premorbid BMI, live in urban, and kidney diseases may be protective factors. Diagnostic criteria and EMG are crucial tools in diagnostics. Gold Coast criteria are characterized by increased diagnostic sensitivity compared to previous criteria. Diagnostics remains a challenge, and definitive diagnosis is established with a significant delay due to patient- and healthcare system-dependent factors. ALS remains incurable. Nevertheless, eliminating diagnostic difficulties may provide a number of benefits for both patients and the healthcare system. Scientists attempt to facilitate the diagnostic process. Numerous publications highlight the potential of modern technologies. The results seem promising, however, these techniques have a number of limitations and involve a risk, so they should be treated with caution, and require further research and improvement.

  • Open Access

    SMARTPHONE-BASED AND INFRARED THERMOGRAPHY FOR HOME PREVENTION OF DIABETIC FOOT ULCERS: ACCESSIBILITY, ADOPTION, AND HEALTH-EQUITY PERSPECTIVE

    by Jerzy Buszko, Michał Dyś, Adrianna Buż, Damian Danilczuk, Marta Bajkowska-Piterak, Monika Szlachta-Gubernat, Nicol Baran, Przemysław Piterak, Wiktor Adamiec

    Home infrared temperature monitoring for diabetic foot ulcer (DFU) prevention is supported by multiple randomized controlled trials demonstrating reductions of roughly 60 to 90% in ulcer recurrence among high-risk patients (IWGDF risk categories 2 to 3) when compared with standard care without thermometry. These findings hold across diverse geographic and resource settings, including low-resource contexts in Indonesia and Peru. Technology platforms range from handheld infrared thermometers costing 30 to 100 US dollars to smartphone-attached thermal cameras, wireless thermometric mats, sensor-embedded socks rated 9 to 10 out of 10 for comfort, and wearable Internet-of-Things devices costing as little as 14.80 US dollars per unit. Adherence rates across technologies generally range from 70% to 88% over study periods of 6 to 24 months, and predictive systems can detect impending ulcers up to 37 days in advance with 97% sensitivity, though at the cost of a 57% false-positive rate at standard thresholds. Despite promising accessibility features, significant barriers remain, including physical limitations such as arthritis and vision loss, incorrect device use in nearly half of participants despite training, and reliance on assistance from others for interpreting results among older, low-income users, all of which disproportionately affect vulnerable populations. Critically, no study formally evaluated differential effectiveness by socioeconomic status, race or ethnicity, education level, or insurance coverage, leaving a substantial gap in the health-equity evidence base. The addition of mHealth reminders to thermometry did not improve adherence or outcomes beyond thermometry alone. While the technological and cost barriers to home thermometry are increasingly surmountable, the absence of equity-stratified analyses means that whether these tools reduce or inadvertently widen disparities in DFU outcomes cannot yet be determined from the available evidence.

  • Open Access

    HANDHELD AND POCKET-SIZED ULTRASOUND FOR ABDOMINAL AORTIC ANEURYSM SCREENING BEYOND THE RADIOLOGY SUITE: ACCESS, TASK-SHIFTING, AND REGULATORY CONSIDERATIONS

    by Michał Dyś, Jerzy Buszko, Adrianna Buż, Damian Danilczuk, Marta Bajkowska-Piterak, Monika Szlachta-Gubernat, Nicol Baran, Przemysław Piterak, Wiktor Adamiec

    Handheld and pocket-sized ultrasound devices enable accurate abdominal aortic aneurysm (AAA) screening across a wide range of non-radiology settings and operator types. This review synthesizes evidence from 26 sources spanning emergency departments, primary care offices, prehospital ambulances, and rural clinics across 12 countries. Non-radiologist operators, including family physicians, emergency physicians, paramedics, medical residents, complete novices, and nurses, achieved sensitivities of 96 to 100% and specificities of 94 to 100% for AAA detection when using focused, protocol-driven scanning. Training requirements ranged from 15 minutes for deep learning-guided nurses to 15 to 25 hours for unassisted clinicians, with measurement accuracy generally within 0.2 to 0.5 cm of reference standards. AI-guided systems reduced training to minutes while maintaining physician-comparable accuracy (100% sensitivity, 94 to 98% specificity, mean absolute error 2.8 mm), though this evidence derives from a single research group and requires broader validation. Obesity consistently reduced diagnostic accuracy, with sensitivity improving from 86% to 96% when large body habitus patients were excluded. Screening costs of $30 to $53 per examination, with devices costing approximately 10% of standard ultrasound machines, support economic viability; cost-utility modeling found opportunistic primary care scanning to be cost-saving for medium and large aneurysms. The principal barriers to broader implementation are not diagnostic accuracy but structural: the absence of standardized credentialing and training requirements across jurisdictions, unresolved coding and reimbursement pathways for non-radiologist point-of-care ultrasound (POCUS) screening, low screening uptake even with decision support, and additional regulatory complexity for AI-guided devices. Successful programs combined institutional endorsement, focused scanning protocols, short dedicated training, workflow adaptations such as nurse-directed ordering, and clear referral pathways for positive cases. Countries with centralized screening programs appear better positioned to integrate POCUS-based AAA screening than fragmented systems requiring individual practitioner navigation of reimbursement.

  • Open Access

    CONTACTLESS AND CONTACT-BASED SMARTPHONE PHOTOPLETHYSMOGRAPHY AS A MULTI-PARAMETER CARDIOMETABOLIC HOME SCREEN: BEYOND ATRIAL FIBRILLATION TOWARD BLOOD PRESSURE AND DYSGLYCEMIA ESTIMATION

    by Jerzy Buszko, Michał Dyś, Adrianna Buż, Damian Danilczuk, Marta Bajkowska-Piterak, Monika Szlachta-Gubernat, Nicol Baran, Przemysław Piterak, Wiktor Adamiec

    Smartphone photoplethysmography (PPG) demonstrates a stark maturity gradient across cardiometabolic parameters. For atrial fibrillation (AF) detection, meta-analytic evidence from 14 studies (n = 5,090) shows pooled sensitivity of 96% (95% CI 93 to 97%) and specificity of 97% (95% CI 95 to 98%), with individual validation studies reporting sensitivity up to 99.7% and specificity up to 99.7% using machine learning algorithms in ambulatory settings. AF detection performance is consistent across multiple smartphone devices, robust in unsupervised home use, and clinically impactful: population screening increased oral anticoagulation uptake from 56% to 74%, and digital post-ablation monitoring detected arrhythmia recurrence at more than double the rate of conventional ECG follow-up (38.5% versus 17.7%, OR 3.4). Signal quality rates for PPG (89 to 97%) were comparable to or exceeded those for single-lead ECG. Blood pressure (BP) estimation, by contrast, remains unreliable for population-level screening. Of 18 studies evaluated, only two clearly passed established AAMI/ESH/ISO validation protocols, both relying on per-individual calibration against a reference measurement. The two largest prospective trials (n = 965 and n = 62) explicitly failed validation standards, and contactless remote PPG showed critically low sensitivity for hypertension detection (systolic BP sensitivity 0.04) in a diverse-skinned field population. Diastolic BP was consistently estimated more accurately than systolic BP across approaches. Dysglycemia detection rests on a single large study demonstrating moderate screening discrimination (AUC 0.74 to 0.77) using a deep neural network applied to PPG waveforms, with the score independently associated with HbA1c. The evidence supports smartphone PPG as a clinically ready tool for AF detection and monitoring, particularly in populations aged over 65, but BP estimation requires individual calibration and cannot yet substitute for cuff-based measurement, and dysglycemia screening remains investigational.

  • Open Access

    FIRST-GENERATION BRAIN–COMPUTER INTERFACES: A NARRATIVE REVIEW OF CLINICAL TRANSLATION, COMPARATIVE SAFETY, AND NEURO-RIGHTS

    by Przemysław Piterak, Damian Danilczuk, Jerzy Buszko, Michał Dyś, Nicol Baran, Marta Bajkowska-Piterak, Monika Szlachta-Gubernat, Wiktor Adamiec, Adrianna Buż

    Objectives. This narrative review examines first-generation brain-computer interfaces (BCIs) across three domains: clinical translation in amyotrophic lateral sclerosis (ALS), locked-in syn-drome (LIS), tetraplegia, spinal cord injury (SCI), pharmacoresistant epilepsy and movement dis-orders; comparative safety covering neurological, infectious, technical and cybersecurity risks; and the emerging neuro-rights regulatory debate. Methods. Literature was retrieved from PubMed/MEDLINE, Scopus, Web of Science, IEEE Xplore and legal-academic databases, covering peer-reviewed clinical trials, systematic reviews, safety datasets, technical analyses and bioethics scholarship through 2026. Key Findings. The BrainGate dataset (14 adults, 12,203 implant-days) reported 68 device-related adverse events and six serious adverse events, with no intracranial infections, explantations or at-tributable deaths. The endovascular SWITCH trial found zero serious adverse events at 12 months. Safety risk follows access route and implant chronicity rather than a simple invasive/non-invasive binary. Closed-loop neurostimulation reduces seizure frequency by 60-65% in pharmacoresistant epilepsy; adaptive deep brain stimulation is the most mature movement-disorder application. The neuro-rights literature converges on mental privacy, integrity, autonomy and cognitive liberty as core protected interests. Conclusions. Responsible BCI deployment requires matching the least invasive platform to the indication, consent specific to device capabilities, cybersecurity treated as patient safety, and pro-spective neural-data governance built into device design.

  • Open Access

    THE IMPACT OF PADEL ON CARDIOMETABOLIC AND PSYCHOSOCIAL HEALTH: EVALUATING CLINICAL BENEFITS AGAINST THE RISK OF MUSCULOSKELETAL INJURIES

    by Jerzy Buszko, Adrianna Buż, Damian Danilczuk, Marta Bajkowska-Piterak, Monika Szlachta-Gubernat, Michał Dyś, Nicol Baran, Przemysław Piterak, Wiktor Adamiec

    Background. Padel currently ranks among the fastest-growing racket sports worldwide, mainly thanks to its social accessibility and relatively low technical entry barrier. However, there is a growing need to look at both sides of its health profile: the documented clinical benefits versus the considerable risk of musculoskeletal injury. Aim. This narrative review aimed to synthesise the existing literature comparing the cardiometabolic and psychosocial benefits of padel with the prevalence and patterns of musculoskeletal injuries. Material and methods. A narrative review was conducted using PubMed, Scopus, and Web of Science. The analysis focused on studies examining physiological demands, mental health outcomes, and injury patterns among amateur and professional padel players. Results. The evidence shows that regular padel practice tends to improve cardiovascular fitness and produces favourable acute changes in metabolic biomarkers, although the evidence for sustained changes in lipid profile is still mixed. Padel also seems to bring psychosocial benefits, including enhanced social engagement and reduced perceived stress. On the other hand, the sport’s biomechanical demands, with rapid decelerations and repetitive overhead strokes, carry a substantial injury burden, predominantly affecting the elbow (epicondylitis), knee, ankle, and shoulder. Conclusions. Padel is a useful tool for promoting health; but practitioners need to be aware of the injury burden it carries. Targeted prevention and proper physical conditioning matter, particularly for recreational participants.

  • Open Access

    VOICE AND COUGH AS DIGITAL BIOMARKERS (2025–2026): LINGUISTIC EQUITY, PRIVACY, AND THE PROMISE OF ACOUSTIC DIAGNOSTICS IN GLOBAL HEALTH

    by Wiktor Adamiec, Damian Danilczuk, Jerzy Buszko, Michał Dyś, Nicol Baran, Marta Bajkowska-Piterak, Przemysław Piterak, Monika Szlachta-Gubernat, Adrianna Buż

    Recent advances in artificial intelligence and mobile sensing have renewed interest in the human voice and cough as non-invasive, low-cost biomarkers of disease. While respiratory, neurological, and psychiatric applications have shown moderate-to-high diagnostic accuracies in classification studies, the translation of these tools from proof-of-concept into routine clinical practice remains incomplete. This narrative review synthesizes the 2025–2026 literature across five domains: (a) the biophysiological foundations that make voice and cough acoustically informative, (b) clinical evidence spanning respiratory, neurodegenerative, mental-health, cardiometabolic, pediatric, and laryngeal use cases, (c) the growing body of work on linguistic equity and structural bias in speech-based systems, (d) privacy and adversarial security concerns specific to voice as biometric data, and (e) deployment prospects in low- and middle-income countries, with tuberculosis screening as the most advanced use case. The analysis reveals that no single disease domain has reached the level of large-scale, independently validated deployment, though respiratory and laryngeal applications are closest. Structural language bias—particularly against speakers of tonal languages—remains a first-order obstacle, and current speaker de-identification systems demonstrably leak identity information. These findings point to three conditions for responsible scale-up: methodological standardization, inclusive multilingual datasets, and governance frameworks that treat voice data as sensitive biometric medical information.

  • Open Access

    ARTIFICIAL INTELLIGENCE IN EMERGENCY DEPARTMENT TRIAGE: A NARRATIVE REVIEW OF ALGORITHMIC ACCURACY, RACIAL AND SOCIOECONOMIC DISPARITIES, AND THE INTEGRATION OF SOCIAL DETERMINANTS OF HEALTH

    by Damian Danilczuk, Jerzy Buszko, Michał Dyś, Nicol Baran, Marta Bajkowska-Piterak, Przemysław Piterak, Monika Szlachta-Gubernat, Wiktor Adamiec, Adrianna Buż

    Background: Emergency department (ED) crowding represents a growing global public health challenge. Limitations of conventional five-level triage scales — including interrater variability and documented racial and socioeconomic disparities — have driven increasing interest in artificial intelligence (AI) for triage optimization. Existing literature has largely examined algorithmic accuracy, fairness, and social determinants of health (SDOH) in isolation. Objective: This narrative review synthesizes evidence across three interconnected domains: algorithmic accuracy of AI triage models compared with conventional scales; racial and socioeconomic disparities and the mechanisms producing algorithmic bias; and the integration of SDOH into AI-driven triage. Methods: Peer-reviewed publications were retrieved from PubMed/MEDLINE, Embase, Scopus, Web of Science, and IEEE Xplore (January 2015 through March 2026), combining structured keyword-based searches across emergency care, AI methodology, equity, and SDOH. Results: AI models — particularly gradient-boosted ensembles and NLP-augmented architectures — consistently outperform ESI, MTS, and CTAS in predicting critical endpoints, with AUROC gains of 0.05–0.10. However, biased proxy variables, measurement artifacts (e.g., pulse oximetry), unrepresentative training data, and biased outcome labels can amplify existing disparities. SDOH integration offers potential for improving equity but faces challenges including Z-code underutilization, definitional heterogeneity, and stigmatization risk. Emerging frameworks — the EU AI Act, FDA guidance, WHO principles, and TRIPOD+AI — provide a scaffold for responsible deployment. Conclusions: Realizing AI's benefits while avoiding harm requires rigorous external validation, transparent subgroup reporting, explicit fairness evaluation, post-deployment monitoring, and participatory design.