-
BMC Medical Informatics and Decision... Aug 2023Differentiating between Crohn's disease (CD) and intestinal tuberculosis (ITB) with endoscopy is challenging. We aim to perform more accurate endoscopic diagnosis...
BACKGROUND
Differentiating between Crohn's disease (CD) and intestinal tuberculosis (ITB) with endoscopy is challenging. We aim to perform more accurate endoscopic diagnosis between CD and ITB by building a trustworthy AI differential diagnosis application.
METHODS
A total of 1271 electronic health record (EHR) patients who had undergone colonoscopies at Peking Union Medical College Hospital (PUMCH) and were clinically diagnosed with CD (n = 875) or ITB (n = 396) were used in this study. We build a workflow to make diagnoses with EHRs and mine differential diagnosis features; this involves finetuning the pretrained language models, distilling them into a light and efficient TextCNN model, interpreting the neural network and selecting differential attribution features, and then adopting manual feature checking and carrying out debias training.
RESULTS
The accuracy of debiased TextCNN on differential diagnosis between CD and ITB is 0.83 (CR F1: 0.87, ITB F1: 0.77), which is the best among the baselines. On the noisy validation set, its accuracy was 0.70 (CR F1: 0.87, ITB: 0.69), which was significantly higher than that of models without debias. We also find that the debiased model more easily mines the diagnostically significant features. The debiased TextCNN unearthed 39 diagnostic features in the form of phrases, 17 of which were key diagnostic features recognized by the guidelines.
CONCLUSION
We build a trustworthy AI differential diagnosis application for differentiating between CD and ITB focusing on accuracy, interpretability and robustness. The classifiers perform well, and the features which had statistical significance were in agreement with clinical guidelines.
Topics: Humans; Crohn Disease; Diagnosis, Differential; Tuberculosis, Gastrointestinal; Colonoscopy
PubMed: 37582768
DOI: 10.1186/s12911-023-02257-6 -
Annals of Medicine Dec 2023The persistent spread of SARS-CoV-2 makes diagnosis challenging because COVID-19 symptoms are hard to differentiate from those of other respiratory illnesses. The...
OBJECTIVE
The persistent spread of SARS-CoV-2 makes diagnosis challenging because COVID-19 symptoms are hard to differentiate from those of other respiratory illnesses. The reverse transcription-polymerase chain reaction test is the current golden standard for diagnosing various respiratory diseases, including COVID-19. However, this standard diagnostic method is prone to erroneous and false negative results (10% -15%). Therefore, finding an alternative technique to validate the RT-PCR test is paramount. Artificial intelligence (AI) and machine learning (ML) applications are extensively used in medical research. Hence, this study focused on developing a decision support system using AI to diagnose mild-moderate COVID-19 from other similar diseases using demographic and clinical markers. Severe COVID-19 cases were not considered in this study since fatality rates have dropped considerably after introducing COVID-19 vaccines.
METHODS
A custom stacked ensemble model consisting of various heterogeneous algorithms has been utilized for prediction. Four deep learning algorithms have also been tested and compared, such as one-dimensional convolutional neural networks, long short-term memory networks, deep neural networks and Residual Multi-Layer Perceptron. Five explainers, namely, Shapley Additive Values, Eli5, QLattice, Anchor and Local Interpretable Model-agnostic Explanations, have been utilized to interpret the predictions made by the classifiers.
RESULTS
After using Pearson's correlation and particle swarm optimization feature selection, the final stack obtained a maximum accuracy of 89%. The most important markers which were useful in COVID-19 diagnosis are Eosinophil, Albumin, T. Bilirubin, ALP, ALT, AST, HbA1c and TWBC.
CONCLUSION
The promising results suggest using this decision support system to diagnose COVID-19 from other similar respiratory illnesses.
Topics: Humans; COVID-19; Artificial Intelligence; SARS-CoV-2; COVID-19 Vaccines; COVID-19 Testing
PubMed: 37436038
DOI: 10.1080/07853890.2023.2233541 -
BMC Cancer Dec 2023The Syrian decade-long war has severely affected the healthcare system, including almost vanishing cancer screening practices, war-destroyed medical facilities, and lack...
BACKGROUND
The Syrian decade-long war has severely affected the healthcare system, including almost vanishing cancer screening practices, war-destroyed medical facilities, and lack of continuous medical education. This study aims to present data on the affected breast cancer screening practices, methods of diagnosis, and stages distribution in Syria.
METHODS
Medical charts of breast cancer patients treated at Albairouni University Hospital between January 2019 and May 2022 were retrospectively reviewed. Eligible patients were women diagnosed with primary breast cancer. Exclusion criteria included females receiving neoadjuvant chemotherapy and incomplete charts. Data regarding the patient's age, city of residence, marital status, number of children, smoking habits, method of diagnosis, tumor size (T), lymph nodes (N), and distal metastasis (M) were collected. We used Microsoft Excel and Statistical Package for the Social Sciences (SPSS) to analyze data. Descriptive methodology (frequency [n], percentage) was used.
RESULTS
The number of charts reviewed was 4,500. The number of remaining charts after applying the exclusion criteria was 2,367. The mean age was 51.8 (SD = 11.3). More than half of the patients (58.3%) came from outside Damascus -where the hospital is located- and its suburbs. Less than 5% of the population detected cancer by screening mammography. Only 32.4% of patients were diagnosed by a biopsy, while surgical procedures (lumpectomy and mastectomy) were used to diagnose 64.8% of the population. At the time of diagnosis, only 8% of patients presented with local-stage disease (stages 0 & I), 73% had a regional disease (stages II & III), and 19% had metastatic breast cancer (stage IV).
CONCLUSION
Our retrospective chart review analysis is the first comprehensive review in Syria for female breast cancer patients. We found a significant low percentage of patients diagnosed based on a screening mammogram, much higher surgical biopsies rather than a simple imaging-guided biopsy, and much lower than the national average of early-stage disease. Our alarming findings can serve as the base for future strategies to raise the population's health awareness, create more serious national screening campaigns, and adopt a multidisciplinary approach to the disease in Syria.
Topics: Child; Female; Humans; Middle Aged; Breast Neoplasms; Early Detection of Cancer; Mammography; Mastectomy; Neoplasm Staging; Retrospective Studies; Syria
PubMed: 38097985
DOI: 10.1186/s12885-023-11740-2 -
Rhinology Dec 2023Nasal bone fractures are common in children but can be challenging to diagnose accurately in the first days due to swelling and tenderness. While X-rays and computed...
BACKGROUND
Nasal bone fractures are common in children but can be challenging to diagnose accurately in the first days due to swelling and tenderness. While X-rays and computed tomography have limitations, ultrasound may be a radiation-free and cost-effective alternative for diagnosing and treating nasal fractures.
METHODS
A prospective cohort study at a tertiary referral hospital between 2021-2023. Children who had sustained nasal trauma were included. A radiologist and a non-radiologist blindly reviewed ultrasound scans, and the results were compared to the physical examination performed by a senior otolaryngologist. If closed reduction was necessary, ultrasound was employed during the procedure. The primary outcome was the assessment of nasal fractures in children using ultrasound; Secondary outcomes included success rates for closed reduction and test reliability.
RESULTS
Of the 50 children (mean age: 11 years, interquartile range: 6-15 years, 36 [72%] males), 22 (44%) were clinically diagnosed with a nasal fracture. Interobserver reliability for nasal fracture by ultrasound was 92%, with a Cohen's kappa coefficient of k=0.91. The sensitivity and specificity of ultrasound in detecting nasal fractures were 90% and 89%, respectively, with positive and negative predictive values of 86% and 93%, respectively. Closed reduction was performed on 18 children, with (n=11) or without (n=7) ultrasound, with the former showing better alignment results (82% vs. 71%).
CONCLUSIONS
Ultrasound has a high negative predictive value in identifying nasal fractures in children with swollen noses during presentation. This enables to avoid further unnecessary referrals and interventions. Ultrasound-guided closed reduction of nasal fractures demonstrates improved outcomes; however, further large-scale randomized studies are required to validate our findings.
Topics: Male; Humans; Child; Female; Prospective Studies; Reproducibility of Results; Skull Fractures; Sensitivity and Specificity; Predictive Value of Tests; Ultrasonography; Nasal Bone
PubMed: 37594057
DOI: 10.4193/Rhin23.176 -
Journal of Dental Education Dec 2023Dentin hypersensitivity (DH) affects patients' oral health-related quality of life, but is not always optimally treated in dental offices. The objectives were to assess...
OBJECTIVES
Dentin hypersensitivity (DH) affects patients' oral health-related quality of life, but is not always optimally treated in dental offices. The objectives were to assess dentists' DH-related education, knowledge, and professional behavior and explore relationships between education, knowledge, and behavior.
METHODS
Survey data were collected from 220 ADA members in the United States. Descriptive and correlational analyses were performed.
RESULTS
About half of the respondents agreed/strongly agreed that their dental school had educated them well about diagnosing DH in classroom-based (53.6%) and clinical settings (48.9%). Lower percentages agreed being well educated about treating DH (40.9%/37.3%). The majority self-educated themselves about DH after dental school by attending continuing education courses in person or online (60.6%/36.8%), reading articles (64.1%), or consulting with colleagues (59.6%). The majority knew that patients with DH describe their pain as stimulated (91.4%) and that recessed gingiva (89.6%), abrasion lesions (72.3%), tooth whitening (63.1%), erosion lesions (58.6%), and abfraction lesions (51.4%) are risk factors for DH. The majority diagnosed DH with patient self-reporting, confirmed by exams (81.8%), applying air blasts (53.7%), or cold-water (52.3%). They treated patients with DH often/very often with over-the-counter desensitizing agents (90%), and prescribing fluoride formulations toothpaste (82.8%) and/or potassium nitrate toothpastes (60.9%). In their offices, the majority (73.2%) educated their patients often/very often about DH and used fluoride dental varnish for treating DH (71.8%). The more recently respondents had graduated from dental school, the more positively they described their dental school education (r = 0.14; p < 0.05), the more ways to diagnose DH they used (r = 0.16; p < 0.05) and the more often they used fluoride dental varnish in their offices (r = 0.23; p < 0.001). The more dentists had educated themselves, the more methods for diagnosing DH they used (r = 0.23; p < 0.001) and the more often they used potassium oxalate products (r = 0.19; p < 0.01), Arginine/calcium products (r = 0.19; p < 0.01) and dentin bonding (r = 0.22; p < 0.001).
CONCLUSIONS
More recently graduating from dental school correlates with more positive evaluations of DH-related dental school education. The finding that most dentists engage in self-education about DH after dental school should motivate dental educators to increase education about this topic not only in dental school, but also in continuing education courses.
Topics: Humans; Fluorides; Dentin Sensitivity; Quality of Life; Educational Status; Toothpastes; Dentists; Treatment Outcome
PubMed: 37650366
DOI: 10.1002/jdd.13363 -
BJGP Open Dec 2023Rather than first diagnosing and then deciding on treatment, GPs may intuitively decide on treatment and justify this through choice of diagnosis.
BACKGROUND
Rather than first diagnosing and then deciding on treatment, GPs may intuitively decide on treatment and justify this through choice of diagnosis.
AIM
To investigate the relationship between choice of a medicalising diagnosis and antibiotic treatment for throat-related consultations.
DESIGN & SETTING
A retrospective cohort study in a large database of UK electronic primary care records between 1 January 2010 and 1 January 2020.
METHOD
All first throat-related consultations were included, categorised as either pharyngitis/tonsillitis or sore throat. The outcome was any antibiotic prescription on the consultation date. GP-level random effects on prescribing and on diagnosis were estimated in a series of mixed-effects regression models, including age, sex, weekday, month, and clinician characteristics as fixed effects. GPs were grouped into quintiles by antibiotic prescribing propensity, and described the proportion of patients they diagnosed with pharyngitis/tonsillitis or sore throat in each quintile.
RESULTS
The analysis dataset included 393 590 throat-related consultations with 6881 staff. Diagnosis of pharyngitis/tonsillitis was strongly associated with antibiotic prescribing (adjusted odds ratio = 13.41, 95% confidence interval = 12.8 to 14.04). GP random effect accounted for 18% of variation in prescribing and for 26% of variation in diagnosis. GPs in the lowest quintile of antibiotic prescribing propensity diagnosed pharyngitis/tonsillitis on 31% of occasions, compared with 55% in the highest quintile.
CONCLUSION
There is substantial variation among GPs in diagnosis and treatment of throat-related problems. Preference for a medicalising diagnosis is associated with a preference for antibiotics, suggesting a common propensity to both diagnose and treat.
PubMed: 37429635
DOI: 10.3399/BJGPO.2023.0056 -
International Journal of Molecular... Sep 2023Histology diagnosis is essential for the monitoring and management of kidney transplant patients. Nowadays, the accuracy and reproducibility of histology have been...
Histology diagnosis is essential for the monitoring and management of kidney transplant patients. Nowadays, the accuracy and reproducibility of histology have been criticized when compared with molecular microscopy diagnostic system (MMDx). Our cohort included 95 renal allograft biopsies with both histology and molecular diagnoses. Discrepancies between histology and molecular diagnosis were assessed for each biopsy. Among the 95 kidney allograft biopsies, a total of 6 cases (6%) showed clear ( = 4) or borderline ( = 2) discrepancies between histology and molecular diagnoses. Four out of the six (67%) were cases with pathologically and clinically confirmed active infections that were diagnosed as mild to moderate T-cell-mediated rejection (TCMR) with MMDx. Two cases showed pathological changes that were not sufficient to make a definitive diagnosis of active rejection via histology, while MMDx results showed antibody-mediated rejection (ABMR). In addition, there were six cases with recurrent or de novo glomerular diseases diagnosed only via histology. All other biopsy results were in an agreement. Our results indicate that histology diagnosis of kidney allograft biopsy is superior to molecular diagnosis in the setting of infections and glomerular diseases; however, MMDx can provide helpful information to confirm the diagnosis of active ABMR.
Topics: Humans; Kidney Transplantation; Reproducibility of Results; Graft Rejection; Kidney Diseases; Biopsy; Antibodies; Kidney; Allografts
PubMed: 37762119
DOI: 10.3390/ijms241813817 -
Tomography (Ann Arbor, Mich.) Oct 2023Neuroimaging has a key role in identifying small-vessel vasculitis from common diseases it mimics, such as multiple sclerosis. Oftentimes, a multitude of these... (Review)
Review
Neuroimaging Scoring Tools to Differentiate Inflammatory Central Nervous System Small-Vessel Vasculitis: A Need for Artificial Intelligence/Machine Learning?-A Scoping Review.
Neuroimaging has a key role in identifying small-vessel vasculitis from common diseases it mimics, such as multiple sclerosis. Oftentimes, a multitude of these conditions present similarly, and thus diagnosis is difficult. To date, there is no standardized method to differentiate between these diseases. This review identifies and presents existing scoring tools that could serve as a starting point for integrating artificial intelligence/machine learning (AI/ML) into the clinical decision-making process for these rare diseases. A scoping literature review of EMBASE and MEDLINE included 114 articles to evaluate what criteria exist to diagnose small-vessel vasculitis and common mimics. This paper presents the existing criteria of small-vessel vasculitis conditions and mimics them to guide the future integration of AI/ML algorithms to aid in diagnosing these conditions, which present similarly and non-specifically.
Topics: Humans; Artificial Intelligence; Machine Learning; Vasculitis; Neuroimaging; Central Nervous System
PubMed: 37888736
DOI: 10.3390/tomography9050144 -
European Heart Journal. Cardiovascular... Sep 2023Traditionally, congestive heart failure (HF) was phenotyped by echocardiography or other imaging techniques according to left ventricular (LV) ejection fraction (LVEF)....
Traditionally, congestive heart failure (HF) was phenotyped by echocardiography or other imaging techniques according to left ventricular (LV) ejection fraction (LVEF). The more recent echocardiographic modality speckle tracking strain is complementary to LVEF, as it is more sensitive to diagnose mild systolic dysfunction. Furthermore, when LV systolic dysfunction is associated with a small, hypertrophic ventricle, EF is often normal or supernormal, whereas LV global longitudinal strain can reveal reduced contractility. In addition, segmental strain patterns may be used to identify specific cardiomyopathies, which in some cases can be treated with patient-specific medicine. In HF with preserved EF (HFpEF), a diagnostic hallmark is elevated LV filling pressure, which can be diagnosed with good accuracy by applying a set of echocardiographic parameters. Patients with HFpEF often have normal filling pressure at rest, and a non-invasive or invasive diastolic stress test may be used to identify abnormal elevation of filling pressure during exercise. The novel parameter LV work index, which incorporates afterload, is a promising tool for quantification of LV contractile function and efficiency. Another novel modality is shear wave imaging for diagnosing stiff ventricles, but clinical utility remains to be determined. In conclusion, echocardiographic imaging of cardiac function should include LV strain as a supplementary method to LVEF. Echocardiographic parameters can identify elevated LV filling pressure with good accuracy and may be applied in the diagnostic workup of patients suspected of HFpEF.
Topics: Humans; Heart Failure; Stroke Volume; Echocardiography; Ventricular Function, Left; Ventricular Dysfunction, Left; Hemodynamics
PubMed: 37542477
DOI: 10.1093/ehjci/jead196 -
Cancer Medicine Aug 2023Endoscopic ultrasonography-guided fine-needle aspiration/biopsy (EUS-FNA/B) is considered to be a first-line procedure for the pathological diagnosis of pancreatic...
BACKGROUND AND AIMS
Endoscopic ultrasonography-guided fine-needle aspiration/biopsy (EUS-FNA/B) is considered to be a first-line procedure for the pathological diagnosis of pancreatic cancer owing to its high accuracy and low complication rate. The number of new cases of pancreatic ductal adenocarcinoma (PDAC) is increasing, and its accurate pathological diagnosis poses a challenge for cytopathologists. Our aim was to develop a hyperspectral imaging (HSI)-based convolution neural network (CNN) algorithm to aid in the diagnosis of pancreatic EUS-FNA cytology specimens.
METHODS
HSI images were captured of pancreatic EUS-FNA cytological specimens from benign pancreatic tissues (n = 33) and PDAC (n = 39) prepared using a liquid-based cytology method. A CNN was established to test the diagnostic performance, and Attribution Guided Factorization Visualization (AGF-Visualization) was used to visualize the regions of important classification features identified by the model.
RESULTS
A total of 1913 HSI images were obtained. Our ResNet18-SimSiam model achieved an accuracy of 0.9204, sensitivity of 0.9310 and specificity of 0.9123 (area under the curve of 0.9625) when trained on HSI images for the differentiation of PDAC cytological specimens from benign pancreatic cells. AGF-Visualization confirmed that the diagnoses were based on the features of tumor cell nuclei.
CONCLUSIONS
An HSI-based model was developed to diagnose cytological PDAC specimens obtained using EUS-guided sampling. Under the supervision of experienced cytopathologists, we performed multi-staged consecutive in-depth learning of the model. Its superior diagnostic performance could be of value for cytologists when diagnosing PDAC.
Topics: Humans; Endoscopic Ultrasound-Guided Fine Needle Aspiration; Cytology; Deep Learning; Pancreatic Neoplasms; Carcinoma, Pancreatic Ductal
PubMed: 37455599
DOI: 10.1002/cam4.6335