Medical Image Segmentation Review: The Success of U-Net.

Automatic medical image segmentation is a crucial topic in the medical domain and successively a critical counterpart in the computer-aided diagnosis paradigm. U-Net is the most widespread image segmentation architecture due to its flexibility, optimized modular design, and success in all medical image modalities. Over the years, the U-Net model has received tremendous attention from academic and industrial researchers who have extended it to address the scale and complexity created by medical tasks. These extensions are commonly related to enhancing the U-Net's backbone, bottleneck, or skip connections, or including representation learning, or combining it with a Transformer architecture, or even addressing probabilistic prediction of the segmentation map. Having a compendium of different previously proposed U-Net variants makes it easier for machine learning researchers to identify relevant research questions and understand the challenges of the biological tasks that challenge the model. In this work, we discuss the practical aspects of the U-Net model and organize each variant model into a taxonomy. Moreover, to measure the performance of these strategies in a clinical application, we propose fair evaluations of some unique and famous designs on well-known datasets. Furthermore, we provide a comprehensive implementation library with trained models. In addition, for ease of future studies, we created an online list of U-Net papers with their possible official implementation. All information is gathered in a GitHub repository https://github.com/NITR098/Awesome-U-Net.

Merlin: A Vision Language Foundation Model for 3D Computed Tomography.

Over 85 million computed tomography (CT) scans are performed annually in the US, of which approximately one quarter focus on the abdomen. Given the current shortage of both general and specialized radiologists, there is a large impetus to use artificial intelligence to alleviate the burden of interpreting these complex imaging studies while simultaneously using the images to extract novel physiological insights. Prior state-of-the-art approaches for automated medical image interpretation leverage vision language models (VLMs) that utilize both the image and the corresponding textual radiology reports. However, current medical VLMs are generally limited to 2D images and short reports. To overcome these shortcomings for abdominal CT interpretation, we introduce Merlin - a 3D VLM that leverages both structured electronic health records (EHR) and unstructured radiology reports for pretraining without requiring additional manual annotations. We train Merlin using a high-quality clinical dataset of paired CT scans (6+ million images from 15,331 CTs), EHR diagnosis codes (1.8+ million codes), and radiology reports (6+ million tokens) for training. We comprehensively evaluate Merlin on 6 task types and 752 individual tasks. The non-adapted (off-the-shelf) tasks include zero-shot findings classification (31 findings), phenotype classification (692 phenotypes), and zero-shot cross-modal retrieval (image to findings and image to impressions), while model adapted tasks include 5-year chronic disease prediction (6 diseases), radiology report generation, and 3D semantic segmentation (20 organs). We perform internal validation on a test set of 5,137 CTs, and external validation on 7,000 clinical CTs and on two public CT datasets (VerSe, TotalSegmentator). Beyond these clinically-relevant evaluations, we assess the efficacy of various network architectures and training strategies to depict that Merlin has favorable performance to existing task-specific baselines. We derive data scaling laws to empirically assess training data needs for requisite downstream task performance. Furthermore, unlike conventional VLMs that require hundreds of GPUs for training, we perform all training on a single GPU. This computationally efficient design can help democratize foundation model training, especially for health systems with compute constraints. We plan to release our trained models, code, and dataset, pending manual removal of all protected health information.

Measurement of adolescent psychological wellbeing: A test of factor structure and measurement invariance.

Gains in holistic approaches to adult mental health have been associated with increasing interest in understanding psychological wellbeing (PWB) among adolescents. Empirical examination of measurement models for PWB in adolescence is lacking. Thus, the current study examined PWB in a longitudinal, diverse sample of 433 adolescents (non-Latinx Black: 37.6%; non-Latinx White: 25.9%; Latinx: 36.5%; Male adolescents: 50.1%). A one-factor, correlated six-factor and hierarchical models were examined across racial/ethnic (White, Black, and Hispanic) and gender (female, male) identities, after which the best fitting model was selected to undergo invariance testing. A one-factor structure was superior, and exhibited strict invariance across racial/ethnic and gender identities at each wave of the study, as well as longitudinal invariance within the entire sample.

A High Threshold of Biotherapeutic Aggregate Numbers is Needed to Induce an Immunogenic Response In Vitro, In Vivo, and in the Clinic.

BACKGROUND AND PURPOSE: There is concern that subvisible aggregates in biotherapeutic drug products pose a risk to patient safety. We investigated the threshold of biotherapeutic aggregates needed to induce immunogenic responses. METHODS AND RESULTS: Highly aggregated samples were tested in cell-based assays and induced cellular responses in a manner that depended on the number of particles. The threshold of immune activation varied by disease state (cancer, rheumatoid arthritis, allergy), concomitant therapies, and particle number. Compared to healthy donors, disease state patients showed an equal or lower response at the late phase (7 days), suggesting they may not have a higher risk of responding to aggregates. Xeno-het mice were used to assess the threshold of immune activation in vivo. Although highly aggregated samples (~ 1,600,000 particles/mL) induced a weak and transient immunogenic response in mice, a 100-fold dilution of this sample (~ 16,000 particles/mL) did not induce immunogenicity. To confirm this result, subvisible particles (up to ~ 18,000 particles/mL, containing aggregates and silicone oil droplets) produced under representative administration practices (created upon infusion of a drug product through an IV catheter) did not induce a response in cell-based assays or appear to increase the rate of adverse events or immunogenicity during phase 3 clinical trials. CONCLUSION: The ability of biotherapeutic aggregates to elicit an immune response in vitro, in vivo, and in the clinic depends on high numbers of particles. This suggests that there is a high threshold for aggregates to induce an immunogenic response which is well beyond that seen in standard biotherapeutic drug products.

PTEN-regulated PI3K-p110 and AKT isoform plasticity controls metastatic prostate cancer progression.

PTEN loss, one of the most frequent mutations in prostate cancer (PC), is presumed to drive disease progression through AKT activation. However, two transgenic PC models with Akt activation plus Rb loss exhibited different metastatic development: Pten/RbPE:-/- mice produced systemic metastatic adenocarcinomas with high AKT2 activation, whereas RbPE:-/- mice deficient for the Src-scaffolding protein, Akap12, induced high-grade prostatic intraepithelial neoplasias and indolent lymph node dissemination, correlating with upregulated phosphotyrosyl PI3K-p85α. Using PC cells isogenic for PTEN, we show that PTEN-deficiency correlated with dependence on both p110ß and AKT2 for in vitro and in vivo parameters of metastatic growth or motility, and with downregulation of SMAD4, a known PC metastasis suppressor. In contrast, PTEN expression, which dampened these oncogenic behaviors, correlated with greater dependence on p110α plus AKT1. Our data suggest that metastatic PC aggressiveness is controlled by specific PI3K/AKT isoform combinations influenced by divergent Src activation or PTEN-loss pathways.

Thirty-Day Outcomes After Acute Total Hip Arthroplasty Combined With Internal Fixation of Acetabular Fractures: A Multi-Institutional Database Analysis.

INTRODUCTION: The purposes of this study were to characterize the 30-day surgical risk of patients undergoing open reduction and internal fixation (ORIF) and total hip arthroplasty stratified by an acetabular fracture pattern and to compare postoperative complications with ORIF alone using a large database. METHODS: The American College of Surgeons National Surgical Quality Improvement Program database was queried to determine 30-day outcomes after the combined hip procedure (CHP) compared with ORIF alone between 2005 and 2020. Current Procedural Terminology codes categorized fracture patterns. Univariate analysis was performed using the chi-square, Fisher exact, or Wilcoxon rank sum test. Logistic regression models were fitted to evaluate for any differences in postoperative complications. Total hospital length of stay was compared. RESULTS: A total of 1,187 patients were identified. One hundred eighty-four patients underwent a CHP, consisting of 99 acetabular wall fractures, 45 elementary acetabular fractures, and 40 associated acetabular fractures. There was no notable difference in any surgical site infection, thromboembolic events, transfusion rates, 30-day revision surgery, and readmission, regardless of the fracture pattern when controlling for comorbidities. Total hospital length of stay was shorter for patients who underwent a CHP for acetabular wall fractures or elementary acetabular fractures (P < 0.001). CONCLUSION: This combined surgical approach appears to have a similar 30-day risk profile when compared with ORIF alone regardless of the fracture pattern.

Adaptive DNA amplification of synthetic gene circuit opens a way to overcome cancer chemoresistance.

Drug resistance continues to impede the success of cancer treatments, creating a need for experimental model systems that are broad, yet simple, to allow the identification of mechanisms and novel countermeasures applicable to many cancer types. To address these needs, we investigated a set of engineered mammalian cell lines with synthetic gene circuits integrated into their genome that evolved resistance to Puromycin. We identified DNA amplification as the mechanism underlying drug resistance in 4 out of 6 replicate populations. Triplex-forming oligonucleotide (TFO) treatment combined with Puromycin could efficiently suppress the growth of cell populations with DNA amplification. Similar observations in human cancer cell lines suggest that TFOs could be broadly applicable to mitigate drug resistance, one of the major difficulties in treating cancer.

Proximal fracture line extension in humeral shaft fractures.

Introduction: Humeral shaft fractures are common injuries treated by orthopaedic surgeons. The purpose of this study is to evaluate displaced diaphyseal humerus fractures and describe the incidence and characteristics associated with non or minimally displaced fracture line extension into the proximal metadiaphyseal region of the humerus. Methods and materials: All adult patients with diaphyseal humeral shaft fractures located within the distal two-thirds of the humeral shaft, treated at a single level I trauma institution between 2007 and 2020, were retrospectively identified. 202 patients with 203 fractures of the humeral shaft were included. Fracture patterns were classified according to AO/OTA classification and fracture line extension into the proximal metadiaphyseal region was evaluated on radiographs. Patient demographics, management details, and radiographic outcomes were obtained from review of the electronic medical record. Results: Of 203 diaphyseal humerus fractures, 11.8 % (n = 24) had non or minimally displaced proximal extension of their main fracture line. This included 43.7 % (n = 7) of all proximal third junction diaphyseal fractures, 10.7 % (n = 16) of all middle third diaphyseal fractures, and 2.6 % (n = 1) of all distal third diaphyseal fractures. Patients with proximal fracture extension were, on average, older (61.7 versus 44.4 years, p < 0.001), and a higher percentage were female (75 % versus 45.5 %, p < 0.01) compared to patients without fracture proximal extension. Fractures with proximal extension were all closed fractures (n = 24), were more often sustained from low-energy fall (87.5 % versus 35.2 %, p < 0.001), and were more often spiral type fractures (62.5 % versus 17.2 %). Fractures with proximal extension were more often treated non-operatively (58.3 % versus 42.1 %, p < 0.01), but were found to have a higher rate of nonunion after non-operative treatment (17.6 % versus 8.1 %) compared to fractures without proximal extension. All operatively treated fractures that had proximal metaphyseal extension were secured with a fixation construct to achieve fixation proximal to the extent of the fracture line, most often into the humeral head and neck. Operative management with proximal fixation into the humeral head was also pursued for a patient with nonunion, including persistent lucency of the proximal extension line, after failed non-operative treatment. Mean follow-up was 35.5 weeks (range: 0-607 weeks). Conclusions: Proximal fracture line extension in the setting of diaphyseal humerus fractures is not uncommon. Detection and consideration of this sometimes subtle finding is important when planning to treat these injuries operatively.

MDS-Related Anemia Is Associated with Impaired Quality of Life but Improvement Is Not Always Achieved by Increased Hemoglobin Level.

Quality of life is impaired in MDS, but the role of hemoglobin level is unclear. To study the Hb-QoL correlation at diagnosis and 1 year later, patients filled out the EQ-5D questionnaire, assessing their mobility, self care, daily activities, pain/discomfort, and anxiety/depression, using scores of 0 (normal), 1 (mild/moderate), or 2 (poor). They also evaluated their health using a visual analogue scale, scoring from 0 (poor) to 100 (excellent). The anemia subgroups were: none/normal (Hb ≥ 12.5 g/dL), mild (10 ≤ Hb < 12.5), moderate (9 ≤ Hb < 10), severe (8 ≤ Hb < 9), or very severe (Hb < 8). LR-MDS patients (n = 127) and inpatient controls (n = 141) participated. The anemic patients had a poor QoL and the MDS patients had a lower QoL with a lower Hb. The controls had no QoL difference among the various anemia subgroups. In addition, the MDS QoL sharply decreased with an Hb of < 9. The MDS patients showed a wide QoL variability, i.e., different QoL scores in the same Hb subgroup, suggesting that other factors affect QoL (e.g., age and comorbidities). After 1 year (n = 61), the QoL was still poor for most MDS patients (including 27 patients with an increased Hb). In summary: (1) a poor QoL in MDS-anemia is non-linear, suggesting other influencing factors on QoL. (2) The sharp QoL drop with Hb < 9 g/dL challenges the transfusion Hb threshold. (3) The QoL in anemic MDS patients might differ from that in non-MDS patients. (4) Raising Hb, while recommended, does not guarantee an improved QoL.

Opportunistic detection of type 2 diabetes using deep learning from frontal chest radiographs.

Deep learning (DL) models can harness electronic health records (EHRs) to predict diseases and extract radiologic findings for diagnosis. With ambulatory chest radiographs (CXRs) frequently ordered, we investigated detecting type 2 diabetes (T2D) by combining radiographic and EHR data using a DL model. Our model, developed from 271,065 CXRs and 160,244 patients, was tested on a prospective dataset of 9,943 CXRs. Here we show the model effectively detected T2D with a ROC AUC of 0.84 and a 16% prevalence. The algorithm flagged 1,381 cases (14%) as suspicious for T2D. External validation at a distinct institution yielded a ROC AUC of 0.77, with 5% of patients subsequently diagnosed with T2D. Explainable AI techniques revealed correlations between specific adiposity measures and high predictivity, suggesting CXRs' potential for enhanced T2D screening.

Quantitative measurement of the requirement of diverse protein degradation pathways in MHC class I peptide presentation.

Peptides from degradation of intracellular proteins are continuously displayed by major histocompatibility complex (MHC) class I. To better understand origins of these peptides, we performed a comprehensive census of the class I peptide repertoire in the presence and absence of ubiquitin-proteasome system (UPS) activity upon developing optimized methodology to enrich for and quantify these peptides. Whereas most class I peptides are dependent on the UPS for their generation, a surprising 30%, enriched in peptides of mitochondrial origin, appears independent of the UPS. A further ~10% of peptides were found to be dependent on the proteasome but independent of ubiquitination for their generation. Notably, clinically achievable partial inhibition of the proteasome resulted in display of atypical peptides. Our results suggest that generation of MHC class I•peptide complexes is more complex than previously recognized, with UPS-dependent and UPS-independent components; paradoxically, alternative protein degradation pathways also generate class I peptides when canonical pathways are impaired.

Testing the temporal precedence of family functioning and child psychopathology in the LONGSCAN sample.

Family functioning may serve as protective or risk factors in the development of youth psychopathology. However, few studies have examined the potentially reciprocal relation between child psychopathology and family functioning. To fill this gap in the literature, this study tested for time-ordered associations between measures of family functioning (e.g., cohesion, conflict, and emotional expressiveness) and child psychopathology (e.g., total behavior problems, externalizing, and internalizing problems) using data from the Longitudinal Studies of Child Abuse and Neglect (LONGSCAN; N = 1143, 52.3% female, Nwaves = 5). We used a random-intercept cross-lagged panel model to identify whether child psychopathology preceded and predicted family functioning, the reverse, or both processes occurred simultaneously. At the between-person level, families who tended to have more cohesion, who lacked conflict, and who expressed their emotions had lower levels of child psychopathology. At the within-person level in childhood, we found minimal evidence for time-ordered associations. In adolescence, however, a clear pattern whereby early psychopathology consistently predicted subsequent family functioning emerged, and the reverse direction was rarely found. Results indicate a complex dynamic relation between the family unit and child that have important implications for developmental models that contextualize risk and resilience within the family unit.

Treatment of Vancouver B2 Femur Fractures With Open Reduction Internal Fixation Versus Revision Arthroplasty.

Background Vancouver B2 periprosthetic femur fractures have traditionally been treated with revision arthroplasty. However, there is increasing evidence that open reduction and internal fixation (ORIF) may be a valid alternative treatment strategy. The purpose of this study was to compare the outcomes of ORIF versus revision arthroplasty for the treatment of Vancouver B2 fractures and evaluate the influence of the treating surgeon's fellowship training on treatment selection. Methodology This was a retrospective cohort study of 31 patients treated for Vancouver B2 periprosthetic fractures (16 ORIF and 15 revision arthroplasty) at a single academic Level 1 trauma center. Outcome measures included one-year mortality, revision, reoperation, infection, and blood loss. Results There were no statistically significant differences in revision, reoperation, or infection at an average follow-up of 65 weeks. Median estimated blood loss was higher in the arthroplasty group (700 cc versus 400 cc; P = 0.04). There were five deaths in the ORIF group versus one in the revision group (P = 0.18). Cases treated by surgeons with fellowship training in arthroplasty were more likely to be treated with revision arthroplasty (10/11, 90.9%) than those treated by surgeons with fellowship training in trauma (5/15, 33.3%; P < 0.01). Conclusions There was no difference in outcomes between the two treatment strategies, but revision was associated with higher blood loss. The appropriate treatment method should be based on surgeon familiarity and patients' characteristics.

PTEN regulated PI3K-p110 and AKT isoform plasticity controls metastatic prostate cancer progression.

PTEN loss, one of the most frequent mutations in prostate cancer (PC), is presumed to drive disease progression through AKT activation. However, two transgenic PC models with Akt activation plus Rb loss exhibited different metastasis development: Pten/RbPE:-/- mice produced systemic metastatic adenocarcinomas with high AKT2 activation, whereas RbPE:-/- mice deficient for the Src-scaffolding protein, Akap12, induced high-grade prostatic intraepithelial neoplasias and indolent lymph node disseminations, correlating with upregulated phosphotyrosyl PI3K-p85α. Using PC cells isogenic for PTEN, we show that PTEN-deficiency correlated with dependence on both p110ß and AKT2 for in vitro and in vivo parameters of metastatic growth or motility, and with downregulation of SMAD4, a known PC metastasis suppressor. In contrast, PTEN expression, which dampened these oncogenic behaviors, correlated with greater dependence on p110α plus AKT1. Our data suggest that metastatic PC aggressiveness is controlled by specific PI3K/AKT isoform combinations influenced by divergent Src activation or PTEN-loss pathways.

Interpersonal Well-Being and Suicidal Outcomes in a Nationally Representative Study of Adolescents: A Translational Study.

Adolescent suicide continues to rise despite burgeoning research on interpersonal risk for suicide. This may reflect challenges in applying developmental psychopathology research into clinical settings. In response, the present study used a translational analytic plan to examine indices of social well-being most accurate and statistically fair for indexing adolescent suicide. Data from the National Comorbidity Survey Replication Adolescent Supplement were used. Adolescents aged 13-17 (N = 9,900) completed surveys on traumatic events, current relationships, and suicidal thoughts and attempts. Both frequentist (e.g., receiver operating characteristics) and Bayesian (e.g., Diagnostic Likelihood Ratios; DLRs) techniques provided insight into classification, calibration, and statistical fairness. Final algorithms were compared to a machine learning-informed algorithm. Overall, parental care and family cohesion best classified suicidal ideation, while these indices and school engagement best classified attempts. Multi-indicator algorithms suggested adolescents at high risk across these indices were approximately 3-times more likely to engage in ideation (DLR = 3.26) and 5-times more likely to engage in attempts (DLR = 4.53). Although equitable for attempts, models for ideation underperformed in non-White adolescents. Supplemental, machine learning-informed algorithms performed similarly, suggesting non-linear and interactive effects did not improve model performance. Future directions for interpersonal theories for suicide are discussed and clinical implications for suicide screening are demonstrated.

Nonmonotone invasion landscape by noise-aware control of metastasis activator levels.

A major pharmacological assumption is that lowering disease-promoting protein levels is generally beneficial. For example, inhibiting metastasis activator BACH1 is proposed to decrease cancer metastases. Testing such assumptions requires approaches to measure disease phenotypes while precisely adjusting disease-promoting protein levels. Here we developed a two-step strategy to integrate protein-level tuning, noise-aware synthetic gene circuits into a well-defined human genomic safe harbor locus. Unexpectedly, engineered MDA-MB-231 metastatic human breast cancer cells become more, then less and then more invasive as we tune BACH1 levels up, irrespective of the native BACH1. BACH1 expression shifts in invading cells, and expression of BACH1's transcriptional targets confirm BACH1's nonmonotone phenotypic and regulatory effects. Thus, chemical inhibition of BACH1 could have unwanted effects on invasion. Additionally, BACH1's expression variability aids invasion at high BACH1 expression. Overall, precisely engineered, noise-aware protein-level control is necessary and important to unravel disease effects of genes to improve clinical drug efficacy.

Sagittal deformity of Garden type I and II geriatric femoral neck fractures is frequently misclassified by lateral radiographs.

The objective of this study was to determine the validity and inter-rater reliability of radiographic assessment of sagittal deformity of femoral neck fractures. Design: This is a retrospective cohort study. Setting: Level 1 trauma center. Patients/Participants: Thirty-one patients 65 years or older who sustained low-energy, Garden type I/II femoral neck fractures imaged with biplanar radiographs and either computed tomography or magnetic resonance imaging were included. Main Outcome Measurements: Preoperative sagittal tilt was measured on lateral radiographs and compared with the tilt identified on advanced imaging. Fractures were defined as "high-risk" if posterior tilt was ≥20 degrees or anterior tilt was >10 degrees. Results: Of 31 Garden type I/II femoral neck fractures, advanced imaging identified 10 high-risk fractures including 8 (25.8%) with posterior tilt ≥20 degrees and 2 (6.5%) with anterior tilt >10 degrees. Overall, there was no significant difference between sagittal tilt measured using lateral radiographs and advanced imaging (P = 0.84), and the 3 raters had good agreement between their measurements of sagittal tilt on lateral radiographs (interclass correlation coefficient 0.79, 95% confidence interval [0.65, 0.88], P < 0.01). However, for high-risk fractures, radiographic measurements from lateral radiographs alone resulted in greater variability and underestimation of tilt by 5.2 degrees (95% confidence interval [-18.68, 8.28]) when compared with computed tomography/magnetic resonance imaging. Owing to this underestimation of sagittal tilt, the raters misclassified high-risk fractures as "low-risk" in most cases (averaging 6.3 of 10, 63%, range 6 - 7) when using lateral radiographs while low-risk fractures were rarely misclassified as high-risk (averaging 1.7 of 21, 7.9%, range 1 - 3, P = 0.01). Conclusions: Lateral radiographs frequently lead surgeons to misclassify high-risk sagittal tilt of low-energy femoral neck fractures as low-risk. Further research is necessary to improve the assessment of sagittal plane deformity for these injuries. Level of Evidence: Level IV diagnostic study.

Recommendations for Imaging of the Temporomandibular Joint. Position Statement from the American Academy of Oral and Maxillofacial Radiology and the American Academy of Orofacial Pain.

This position statement was developed by an ad hoc committee of the American Academy of Oral and Maxillofacial Radiology and the American Academy of Orofacial Pain. The committee reviewed the pertinent literature and drafted recommendations for imaging. This joint statement provides evidence-based recommendations and clinical guidance for applying appropriate diagnostic imaging to evaluate the temporomandibular joint (TMJ). This manuscript guides the design of TMJ imaging examinations, addresses in-office CBCT imaging, and provides timely evidence-based recommendations to evaluate the TMJ bony components, also addressing the use of MRI and other modalities to evaluate TMJ involvement in different pathologic conditions.