Cellulitis is a bacterial infection of the skin and the soft tissue beneath it, most often caused by streptococci or staphylococci that slip through cracks in the skin’s protective barrier. The result is a red, hot, swollen and painful area of skin, usually on the lower legs, often accompanied by fever and malaise. Although antibiotics can usually clear the infection, cellulitis imposes a substantial burden on health systems, because severe cases require intravenous therapy and hospital admission, and a considerable share of patients return with new episodes months or years later. Despite how frequently clinicians encounter the disease, risk factors reported in the literature have been inconsistent, which is precisely the gap the Vienna team set out to close.

The researchers designed their investigation as a retrospective case-control study, a design that compares people who have a condition with people who do not, and then looks backward to identify exposures or characteristics that differ between the two groups. In this case, the cases were 1,395 patients treated for cellulitis, while the controls numbered 667. Using regression models, the team analyzed which variables were associated with three distinct outcomes: the initial development of cellulitis, the occurrence of recurrent episodes, and prolonged hospitalization. Statistical models of this kind produce odds ratios, figures that express how much more likely an outcome is in the presence of a given factor, together with confidence intervals that indicate the precision of the estimate and p-values that signal statistical significance.

The results paint a remarkably consistent picture. Male sex emerged as a strong risk factor for developing cellulitis in the first place, with an odds ratio of 2.68, meaning men in the dataset had nearly triple the odds of women. Higher baseline age also mattered, with each additional year of age modestly but significantly raising the odds, reflected in an odds ratio of 1.03 per year. These demographic findings align with the clinical intuition that older patients and male patients are overrepresented on wards treating soft tissue infections, but the study places those impressions on firm quantitative footing.

The most striking findings concerned breaches of the skin barrier, which dwarfed every other risk factor in magnitude. Patients with tinea pedis, the fungal infection commonly known as athlete’s foot, had an odds ratio of 72.48 for developing cellulitis, an enormous effect that underscores how the fissures and scaling produced by a seemingly minor fungal problem can open the door to serious bacterial invasion. Ulcers carried an odds ratio of 45.79, traumatic wounds an odds ratio of 229.26, stasis dermatitis an odds ratio of 10.81, and eczema an odds ratio of 13.14. In other words, almost any condition that compromises the integrity of the skin, whether through inflammation, ulceration or injury, dramatically multiplies the risk that bacteria will gain entry and establish an infection in the underlying tissue.

Fluid-related conditions formed a second major cluster of risk. Venous insufficiency, in which faulty valves in the leg veins allow blood to pool and fluid to leak into the tissues, was associated with an odds ratio of 2.36. Lymphoedema, the swelling that results from impaired drainage of lymph fluid, carried an odds ratio of 11.21, and oedema not otherwise specified an odds ratio of 13.12. The biological logic is compelling: fluid-rich, swollen tissue is thought to impair local immune defenses and circulation, creating an environment in which bacteria that have breached the skin can proliferate more easily. Clinicians have long observed that patients with chronically swollen legs are prone to cellulitis, and this study quantifies just how pronounced that vulnerability is.

Metabolic and chronic systemic conditions also left their mark. Diabetes mellitus was associated with an odds ratio of 1.58 for disease onset, chronic kidney disease with an odds ratio of 2.29, and obesity with an odds ratio of 2.74. Each of these conditions is known to affect immunity, tissue perfusion or wound healing in ways that could plausibly facilitate infection. Diabetes impairs both immune cell function and microvascular circulation, chronic kidney disease is linked to immune dysfunction and fluid overload, and obesity is associated with chronic low-grade inflammation, altered lymphatic drainage and skin fold problems that can compromise the barrier function of the skin.

Perhaps the most clinically consequential finding is that obesity did not merely raise the risk of a first episode. It was also a consistent risk factor for multiple disease episodes, with an odds ratio of 1.73, and for prolonged hospitalization, with an odds ratio of 1.72. This means that patients with obesity were not only more likely to develop cellulitis in the first place but also more likely to experience the disease repeatedly and to require longer hospital stays when they did. In an era when obesity rates continue to climb globally, this finding suggests that the burden of cellulitis on health systems may grow in parallel, and it highlights weight management as a potential long-term strategy for reducing both the incidence and the severity of the disease.

The authors conclude that male sex, higher age, breaches of the skin barrier, oedema, diabetes mellitus, chronic kidney disease and obesity may act as strong risk factors for cellulitis, with obesity additionally predisposing patients to recurrent episodes and extended hospitalization. The practical implications are significant. For patients with athlete’s foot, ulcers, eczema or leg swelling, meticulous skin care and prompt treatment of these conditions could represent a highly effective form of prevention, given the enormous odds ratios attached to barrier defects. For patients with lymphoedema or venous insufficiency, compression therapy and careful management of swelling may reduce infection risk. And for the growing population of patients with obesity, the study adds another compelling reason to address weight as a modifiable factor in infectious disease risk.

As with any retrospective case-control study, the findings describe associations rather than proven causal relationships, and the retrospective design means the researchers relied on previously recorded data rather than prospective observation. Nevertheless, the large number of patients analyzed and the breadth of risk factors examined make this one of the most comprehensive assessments of cellulitis epidemiology available. By systematically separating the factors that drive the first infection from those that drive recurrence and prolonged hospital stay, the Vienna team has given clinicians a clearer map of where to focus preventive efforts, from treating fungal foot infections and managing leg swelling to recognizing obesity as a factor that shapes the entire trajectory of this common and costly infection.

Subject of Research: Risk factors for the onset, recurrence and prolonged hospitalization of cellulitis patients

Article Title: Risk factors for onset, recurrence and prolonged hospitalization in patients with cellulitis: a retrospective case-control study

Article References: Borst, C., Steiner, I., Stegmüller, S., Peichl, D., Egger, E., Symmank, D., Chromy, D., & Weninger, W. (2026). Risk factors for onset, recurrence and prolonged hospitalization in patients with cellulitis: a retrospective case-control study. BMC Infectious Diseases. https://doi.org/10.1186/s12879-026-14500-2

Image Credits: AI Generated

DOI: 10.1186/s12879-026-14500-2

Keywords: cellulitis, skin infection, risk factors, obesity, tinea pedis, lymphoedema, diabetes mellitus, chronic kidney disease, case-control study, hospitalization, dermatology, infectious diseases

The research, conducted at Loghman Hakim Hospital, a tertiary referral centre affiliated with Shahid Beheshti University of Medical Sciences in Tehran, was published in Health Science Reports. It followed the Standards for Reporting of Diagnostic Accuracy Studies (STARD 2015) and enrolled 127 adults with diabetes and clinically suspected foot infection, each of whom underwent both ultrasound and magnetic resonance imaging (MRI) as part of routine care. The stakes are considerable: diabetic foot ulcers affect an estimated 19 to 34 percent of diabetic patients over their lifetime, and people with these ulcers face roughly an eight-fold higher risk of lower extremity amputation compared with non-diabetics. Osteomyelitis is present in 10 to 15 percent of moderate infections and in about half of severe ones, making early and reliable detection a decisive factor in whether a limb is saved.

The technical logic of the study is straightforward but demanding. MRI, performed on a 1.5 Tesla scanner with a full protocol of T1-weighted, T2-weighted, proton-density fat-saturated, diffusion-weighted and contrast-enhanced sequences, served as the reference standard for osteomyelitis and deep abscess, in line with Infectious Diseases Society of America and International Working Group on the Diabetic Foot guidance. On MRI, osteomyelitis was defined by bone marrow oedema appearing dark on T1-weighted images and bright on fluid-sensitive sequences, restricted diffusion, and cortical destruction adjacent to the ulcer. Ultrasound criteria included periosteal elevation or thickening, subperiosteal abscess, and cortical erosion with fluid in direct contact with bone. Ultrasound was always performed first, and each modality was interpreted by a different experienced radiologist blinded to the other’s findings, preserving the independence of the two readings.

The headline result is stark. MRI identified osteomyelitis in 106 of the 127 participants, or 83.5 percent, yet ultrasound detected only 17.9 percent of those MRI-positive cases, with a confidence interval of 11.2 to 26.6 percent. In other words, more than four out of five bone infections visible on MRI were invisible to ultrasound. Specificity was high, at 95.2 percent, meaning that when ultrasound did suggest bone infection it was usually right, with a positive predictive value of 95 percent. But the negative predictive value collapsed to just 18.7 percent overall, and in severe ulcers it fell to an almost meaningless 1.1 percent. A clean ultrasound scan in a severely affected foot, the data show, provides almost no reassurance whatsoever that the underlying bone is healthy.

The severity subgroup analysis sharpens the clinical message. Among the 19 patients with moderate ulcers, classified as Wagner grades 1 and 2, MRI found no osteomyelitis at all, and ultrasound correctly agreed in every case. But among the 108 patients with severe ulcers, grades 3 through 5, MRI confirmed bone infection in 98.1 percent, and ultrasound missed the overwhelming majority of them. Overall accuracy for osteomyelitis detection in severe ulcers was a mere 18.5 percent. The authors argue that these figures quantify the danger of false reassurance: a clinician who relies on a negative ultrasound to exclude bone infection in a severely ulcerated foot risks delayed diagnosis, inadequate surgical debridement, and ultimately amputation.

The picture changes dramatically when the target shifts from bone to soft tissue. For abscess detection, ultrasound achieved perfect sensitivity of 100 percent in both moderate and severe ulcers, meaning a negative scan reliably rules out a drainable collection. Specificity, however, was poor, at 30.4 percent overall and as low as 22.6 percent in severe ulcers, producing a flood of false positives and a positive predictive value of only 46.6 percent overall. The practical implication is that ultrasound works well as a rule-out tool for abscess but should not be used alone to confirm one before committing a patient to surgical drainage. For cellulitis, ultrasound identified the condition in 126 of 127 cases, agreeing with both the clinical diagnosis and MRI in 99.2 percent of participants, although the authors caution that because every participant had clinically diagnosed cellulitis, these concordance figures describe detection rather than true discriminative accuracy.

Why does ultrasound fail so badly at bone? The physics and anatomy offer an explanation. Ultrasound sound waves reflect strongly at soft tissue interfaces and cannot penetrate cortical bone, so the modality can only infer infection from secondary signs at the bone surface, such as periosteal reaction or cortical irregularity. Early marrow infection, which MRI captures through signal changes deep within the bone, produces no visible surface change for the ultrasound probe to detect. The study’s authors also note operator dependence and probe quality as limiting factors, and they contrast their findings with earlier literature: a 1999 study by Enderle and colleagues reported ultrasound sensitivity of 78.6 percent for chronic osteomyelitis, and paediatric studies have shown better performance in acute haematogenous osteomyelitis, where accessible affected areas and different disease biology may favour surface detection.

The study is candid about its limitations. Bone biopsy with histopathology and microbiology remains the definitive gold standard for osteomyelitis, but subjecting every participant to an invasive bone procedure was not ethically justifiable, so MRI served as the reference standard, and MRI itself is imperfect. Charcot neuroarthropathy, with its reactive marrow oedema, can mimic infection on MRI and may have inflated apparent disease in some cases. Each modality was read by a single radiologist, so interobserver reproducibility could not be assessed. Requiring contrast-enhanced MRI excluded patients with advanced kidney disease, a group in whom chronic renal impairment is common and in whom ultrasound may in practice be the only feasible option. Patients with indeterminate MRI bone findings, described as osteitis without confluent marrow replacement, were excluded rather than adjudicated, a recognised source of optimistic bias in diagnostic accuracy studies. The predominance of severe ulcers in the cohort also introduces spectrum bias toward the very patients in whom ultrasound performed worst.

Microbiological data from the cohort add epidemiological texture. Staphylococcus aureus was the most prevalent pathogen at 25.2 percent, followed by Escherichia coli at 18.1 percent, Staphylococcus epidermidis at 14.2 percent, Enterococcus species at 13.4 percent, Pseudomonas aeruginosa at 9.4 percent, Klebsiella pneumoniae at 8.7 percent and Acinetobacter baumannii at 7.9 percent. Most ulcers were severe: 37 percent were Wagner grade 3, 31.5 percent grade 4 and 16.5 percent grade 5, with the forefoot the most commonly involved region at 48 percent and the plantar surface the most frequent site of surface involvement at 40.9 percent. No imaging-related complications were recorded for either modality.

The authors propose a tiered diagnostic protocol as the study’s practical takeaway. Ultrasound, cheap, portable and radiation-free, earns its place as a first-line triage tool for soft-tissue complications, rapidly identifying abscesses that need drainage and confirming cellulitis at the bedside. But its diagnostic ceiling at the bone surface is now quantified, and it is low. Where osteomyelitis is clinically suspected and ultrasound is negative, MRI remains a mandatory next step, consistent with current IWGDF guidance that recommends plain radiography first, followed by MRI when suspicion persists. In a condition where every day of delay narrows the window between debridement and amputation, the message for clinicians is unambiguous: trust ultrasound to map the soft tissues, but never let a negative scan silence suspicion of infection in the bone.

Subject of Research: Diagnostic accuracy of ultrasonography versus MRI for detecting osteomyelitis and soft-tissue infections in diabetic foot ulcers

Article Title: Ultrasonography Cannot Exclude Osteomyelitis in Diabetic Foot Ulcers: A Cross‐Sectional Diagnostic Accuracy Study Against Magnetic Resonance Imaging

Article References: Haghighi‐Morad, M., Vahedi, H., Fard, A. Y., Pordanjani, B. M., & Mahmoudi, S. T. S. (2026). Ultrasonography Cannot Exclude Osteomyelitis in Diabetic Foot Ulcers: A Cross‐Sectional Diagnostic Accuracy Study Against Magnetic Resonance Imaging. Endocrinology, Diabetes & Metabolism, 9(5), Article e70320. https://doi.org/10.1002/edm2.70320

Image Credits: AI Generated

DOI: 10.1002/edm2.70320

Keywords: diabetic foot ulcer, osteomyelitis, ultrasonography, MRI, diagnostic accuracy, amputation, abscess, cellulitis, diabetes complications, medical imaging, Wagner classification, IWGDF guidelines

The stakes of getting this diagnosis right are remarkably high. When cellulitis is mistakenly diagnosed in patients who actually have pseudocellulitis conditions such as venous stasis dermatitis, lipodermatosclerosis, or inflammatory reactions to other causes, those patients are often admitted unnecessarily and treated with days of intravenous antibiotics they do not need. Landmark research published in JAMA Dermatology estimated the annual cost of misdiagnosed lower extremity cellulitis in the United States at tens of millions of dollars, alongside avoidable hospital-acquired complications and delays in treating the actual underlying disease. Conversely, missing genuine cellulitis can allow a rapidly spreading bacterial infection to progress toward sepsis, making accurate and timely differentiation a genuine clinical imperative.

Traditional approaches to this diagnostic dilemma have relied on clinical prediction rules, such as the ALT-70 model and the Dundee score, which incorporate features like asymmetry of the leg, temperature, leukocyte counts, and tachycardia to improve diagnostic accuracy. While these models have shown promise in validation studies, their adoption has been limited by the need for laboratory values and bedside calculations, and their performance in real-world settings remains variable. The Ohio State team, led by first author Priya Gupta and corresponding author Abraham M. Korman, took a fundamentally different approach by asking whether the visual appearance of the affected skin itself contains enough diagnostic signal to be captured by a machine learning classifier.

What makes this study particularly notable is the tool the researchers chose. Rather than training a bespoke deep learning network from scratch, which typically requires substantial computing infrastructure and machine learning expertise, the team used Google’s Teachable Machine, a freely available web-based platform designed to let non-specialists build image classifiers with remarkably little technical overhead. Teachable Machine applies transfer learning, taking a neural network pre-trained on millions of general images and fine-tuning its final layers on a small set of custom examples. This approach drastically reduces the data requirements, allowing a clinically driven research team to develop a working classifier in a fraction of the time and cost associated with conventional model development.

The researchers assembled image datasets from two complementary sources: publicly available clinical photographs published in peer-reviewed manuscripts and inpatient images obtained at The Ohio State University under Institutional Review Board protocol 2023H0077, approved in March 2023, with written informed consent from all patients for the use of identifiable images. Published images were de-identified and therefore did not require additional review. This dual-source strategy balanced the reproducibility of textbook-quality published photographs against the messier but more clinically realistic appearance of cellulitis on hospital wards, where bandages, leg elevation, lighting variation, and concurrent edema complicate the visual picture.

The model’s performance was evaluated using standard statistical measures that readers of the machine learning literature will recognize. The area under the receiver operating characteristic curve, or AUROC, summarizes a classifier’s ability to discriminate between classes across all decision thresholds, with values approaching 1.0 indicating near-perfect separation and 0.5 representing chance-level guessing. The team also reported confidence intervals to quantify uncertainty around their estimates and Cohen’s kappa, a statistic that measures agreement beyond what would be expected by chance and is particularly valued in diagnostic studies because it corrects for the inflation that arises when a condition is common.

The work builds directly on a growing program of research by overlapping members of the same group. In a 2026 study published in the International Journal of Dermatology, Fosu and colleagues, including several of the cellulitis study’s co-authors, demonstrated that Google’s Teachable Machine could identify validated cases of pyoderma gangrenosum, an ulcerative skin disorder that is notoriously difficult to distinguish from other wound types. That earlier result provided proof of concept that the platform could capture the subtle morphological signatures of inflammatory dermatoses, and the cellulitis study extends the paradigm to one of the highest-volume and most consequential diagnostic challenges in acute care medicine.

The clinical vision behind this line of work is augmentation rather than replacement. Emergency physicians, hospitalists, and even primary care providers see few cases of confirmed cellulitis relative to dermatology specialists, and interobserver agreement on the diagnosis is historically modest. A point-of-care image classifier could function as a second opinion, flagging cases whose appearance deviates from typical cellulitis and prompting consideration of pseudocellulitis diagnoses before empiric antibiotics and hospital admission are initiated. In telemedicine contexts, where the examination is already mediated by photographs, such a tool could integrate naturally into existing workflows.

Important limitations temper the enthusiasm. The datasets analyzed in the study are not publicly available due to patient privacy and institutional restrictions, though they are available from the corresponding author on reasonable request. Models trained on photographs from a single academic center and published literature may struggle to generalize across skin tones, imaging devices, and clinical environments, a persistent challenge in dermatology-focused artificial intelligence. The authors also disclose that the machine learning tool was used to create and test the classifier itself, with the team reviewing and editing all outputs and taking full responsibility for the published content, an increasingly common transparency practice in AI-assisted research. Prospective validation in diverse, real-world emergency and inpatient settings will be essential before such a tool can influence bedside decisions.

Even with those caveats, the study represents a compelling demonstration of how democratized machine learning platforms can move from novelty to clinical relevance. The project was supported by the Samuel J. Roessler Memorial Medical Student Research Scholarship, underscoring that meaningful contributions to diagnostic artificial intelligence no longer require industrial-scale resources. As image-classification tools of this kind accumulate validated use cases across dermatology, from ulcerative disorders to soft tissue infections, the prospect of inexpensive, rapidly deployable diagnostic adjuncts for some of medicine’s most error-prone judgment calls is moving steadily closer to practice.

Subject of Research: Machine learning-based image classification for the diagnosis of lower extremity cellulitis

Article Title: Augmenting the diagnosis of cellulitis using image-based machine learning

Article References: Gupta, P., Fosu, N. A., Bittinger, Z., Bobulsky, G. S., Conroy, M., Eisinger, G. J., Kaffenberger, B. H., Kirven, R. M., & Korman, A. M. (2026). Augmenting the diagnosis of cellulitis using image-based machine learning. Archives of Dermatological Research, 318(1), Article 469. https://doi.org/10.1007/s00403-026-04964-2

Image Credits: AI Generated

DOI: 10.1007/s00403-026-04964-2

Keywords: cellulitis, machine learning, teachable machine, dermatology, misdiagnosis, pseudocellulitis, image classification, diagnostic support, emergency medicine, skin and soft tissue infection, transfer learning, predictive medicine