Machine learning-based medical imaging diagnosis in patients with temporomandibular disorders: a diagnostic test accuracy systematic review and meta-analysis

OBJECTIVES

Temporomandibular disorders (TMDs) are the second most common musculoskeletal condition which are challenging tasks for most clinicians. Recent research used machine learning (ML) algorithms to diagnose TMDs intelligently. This study aimed to systematically evaluate the quality of these studies and assess the diagnostic accuracy of existing models.

MATERIALS AND METHODS

Twelve databases (Europe PMC, Embase, etc.) and two registers were searched for published and unpublished studies using ML algorithms on medical images. Two reviewers extracted the characteristics of studies and assessed the methodological quality using the QUADAS-2 tool independently.

RESULTS

A total of 28 studies (29 reports) were included: one was at unclear risk of bias and the others were at high risk. Thus the certainty of evidence was quite low. These studies used many types of algorithms including 8 machine learning models (logistic regression, support vector machine, random forest, etc.) and 15 deep learning models (Resnet152, Yolo v5, Inception V3, etc.). The diagnostic accuracy of a few models was relatively satisfactory. The pooled sensitivity and specificity were 0.745 (0.660-0.814) and 0.770 (0.700-0.828) in random forest, 0.765 (0.686-0.829) and 0.766 (0.688-0.830) in XGBoost, and 0.781 (0.704-0.843) and 0.781 (0.704-0.843) in LightGBM.

CONCLUSIONS

Most studies had high risks of bias in Patient Selection and Index Test. Some algorithms are relatively satisfactory and might be promising in intelligent diagnosis. Overall, more high-quality studies and more types of algorithms should be conducted in the future.

CLINICAL RELEVANCE

We evaluated the diagnostic accuracy of the existing models and provided clinicians with much advice about the selection of algorithms. This study stated the promising orientation of future research, and we believe it will promote the intelligent diagnosis of TMDs.

Quantifying error introduced by iterative closest point image registration

OBJECTIVES

The aim of this paper was to quantify the analysis error introduced by iterative closest point (ICP) image registration. We also investigated whether a subsequent subtraction process can reduce process error.

METHODS

We tested metrology and two 3D inspection software using calibration standards at 0.39 μm, and 2.64 μm and mathematically perfect defects (softgauges) at 2 and 20 μm, on free form surfaces of increasing complexity and area, both with and without registration. Errors were calculated in percentage relative to the size of the defect being measured. Data were analysed in GraphPad Prism 9, normal and two-way ANOVA with post-hoc Tukey's was applied. Significance was inferred at p < 0.05.

RESULTS

Using ICP registration introduced errors from 0 % to 15.63 % of the defect size depending on the surface complexity and size of the defect. Significant differences were observed in analysis measurements between metrology and 3D inspection software and within different 3D inspection software, however, one did not show clear superiority over another. Even in the absence of registration, defects at 0.39 μm, and 2.64 μm produced substantial measurement error (13.39-77.50 % of defect size) when using 3D inspection software. Adding an additional data subtraction process reduced registration error to negligible levels (<1 % independent of surface complexity or area).

CONCLUSIONS

Commercial 3D inspection software introduces error during direct measurements below 3 μm. When using an ICP registration, errors over 15 % of the defect size can be introduced regardless of the accuracy of adjacent registration surfaces. Analysis output between software are not consistently repeatable or comparable and do not utilise ISO standards. Subtracting the datasets and analysing the residual difference reduced error to negligible levels.

CLINICAL SIGNIFICANCE

This paper quantifies the significant errors and inconsistencies introduced during the registration process even when 3D datasets are true and precise. This may impact on research diagnostics and clinical performance. An additional data processing step of scan subtraction can reduce this error but increases computational complexity.

Natural Language Processing: Chances and Challenges in Dentistry

INTRODUCTION

Natural language processing (NLP) is an intersection between Computer Science and Linguistic which aims to enable machines to process and understand human language. We here summarized applications and limitations of NLP in dentistry.

DATA AND SOURCES

Narrative review.

FINDINGS

NLP has evolved increasingly fast. For the dental domain, relevant NLP applications are text classification (e.g., symptom classification) and natural language generation and understanding (e.g., clinical chatbots assisting professionals in office work and patient communication). Analyzing large quantities of text will allow understanding diseases and their trajectories and support a more precise and personalized care. Speech recognition systems may serve as virtual assistants and facilitate automated documentation. However, to date, NLP has rarely been applied in dentistry. Existing research focuses mainly on rule-based solutions for narrow tasks. Technologies such as Recurrent Neural Networks and Transformers have been shown to surpass the language processing capabilities of such rule-based solutions in many fields, but are data-hungry (i.e., rely on large amounts of training data), which limits their application in the dental domain at present. Technologies such as federated or transfer learning or data sharing concepts may allow to overcome this limitation, while challenges in terms of explainability, reproducibility, generalizability and evaluation of NLP in dentistry remain to be resolved for enabling approval of such technologies in medical devices and services.

CONCLUSIONS

NLP will become a cornerstone of a number of applications in dentistry. The community is called to action to improve the current limitations and foster reliable, high-quality dental NLP.

CLINICAL SIGNIFICANCE

NLP for text classification (e.g., dental symptom classification) and language generation and understanding (e.g., clinical chatbots, speech recognition) will support administrative tasks in dentistry, provide deeper insights for clinicians and support research and education.

Real-time optical detection of endodontic infection using bacterial autofluorescence

OBJECTIVES

For successful root canal treatment (RCT), it is essential to objectively assess the presence and activity of bacteria in the root canal system. However, current methods rely on subjective observations of root canal exudates. This study aimed to confirm whether real-time optical detection using bacterial autofluorescence can evaluate endodontic infection status by assessing the red fluorescence (RF) detected from root canal exudates.

METHODS

During RCT, endodontic paper points were used to collect root canal exudates scored using conventional organoleptic tests to assess the severity of root canal infections. RF on the paper points was assessed using quantitative light-induced fluorescence (QLF) technology. RF intensity and area from the paper points were quantified, and their correlations with infection severity were assessed using their organoleptic scores. The oral microbiome composition of RF samples was compared with non-red fluorescent (non-RF) samples.

RESULTS

The RF detection rate was nil and >98% in the non-infectious and severe groups. The RF intensity and area significantly increased with infection severity (p<0.001) and showed strong correlations with organoleptic scores (r=0.72, 0.82, respectively). The diagnostic accuracy for detecting root canal infection using RF intensity was good to excellent (AUC = 0.81-0.95) and increased with infection severity. The microbial diversity of the RF samples was significantly lower than that of the non-RF samples. Gram-negative anaerobic bacteria such as Prevotella and Porphyromonas were more predominant in RF samples.

CONCLUSIONS

Optical detection using bacterial autofluorescence can objectively evaluate endodontic infection status in real-time by assessing the RF of endodontic root canal exudates.

CLINICAL SIGNIFICANCE

This real-time optical technology can be utilised to detect endodontic bacterial infection without conventional incubation, allowing clinicians to determine the endpoint of chemomechanical debridement and increase the positive outcomes of RCTs.

Alternative Diagnostic Models of the Psychotic Disorders: Evidence-Based Choices

Standard diagnostic systems, the predominantly categorical DSM-5 and ICD-11, have limitations in validity, utility, and predictive and descriptive power. For psychotic disorders, these issues were partly addressed in current versions, but additional modifications are thought to be needed. Changes should be evidence based. We reviewed categorical, modified-categorical, and continuum-based models versus factor-based models of psychosis. Factors are clusters of symptoms or single prominent aspects of illness. Consistent evidence from studies of the genetics, pathobiology, and clinical presentation of psychotic disorders all support an underlying structure of factors, not categories, as best characterizing psychoses. Factors are not only the best fit but also comprehensive, as they can encompass any key feature of illness, including symptoms and course, as well as determinants of risk or response. Factors are inherently dimensional, even multidimensional, as are the psychoses themselves, and they provide the detail needed for either grouping or distinguishing patients for treatment decisions. The tools for making factor-based diagnoses are available, reliable, and concordant with actual practices used for clinical assessments. If needed, factors can be employed to create categories similar to those in current use. In addition, they can be used to define unique groupings of patients relevant to specific treatments or studies of the psychoses. Lastly, factor-based classifications are concordant with other comprehensive approaches to psychiatric nosology, including personalized (precision treatment) models and hierarchical models, both of which are currently being explored. Factors might be considered as the right primary structural choice for future versions of standard diagnostic systems, both DSM and ICD.

At the Core of Depression: A Diagnostic Interview of the Core Features of Depression

INTRODUCTION

Valid and reliable methods for diagnosing depression are essential. The present study aimed to test the performance of a new diagnostic interview for depression focusing on the core symptoms of depression.

METHOD

We developed a diagnostic interview for depression: the CORE Diagnostic Interview, CORE-DI, which assesses each of the core features of depression on the four dimensions: quality, reactivity, globality, and fluctuations over time. The diagnostic performance of this interview was tested in a clinical study including 83 individuals presenting with various depressive symptoms, who were interviewed independently (1) by means of the CORE-DI and the Mini-International Neuropsychiatric Interview (M.I.N.I.), and (2) by highly skilled specialists in depression representing gold standard diagnoses.

RESULTS

We compared the outcome of the CORE-DI, the M.I.N.I., and the diagnosis made by clinicians, respectively, versus the gold standard diagnosis, using diagnostic efficiency statistics. The CORE-DI diagnosed depression with a high specificity (0.91, 95% CI: 0.85-0.97, for International Classification of Diseases [ICD]-10 criteria and 0.88, 95% CI: 0.81-0.95, for Diagnostic and Statistical Manual of Mental Disorders [DSM-5] criteria) compared to both M.I.N.I (specificity 0.44, 95% CI: 0.33-0.55) and clinical diagnoses (specificity 0.76, 95% CI: 0.67-0.85). The sensitivity of the CORE-DI was 0.61 (95% CI: 0.55-0.72) for ICD-10 criteria and 0.67 (95% CI: 0.57-0.77) for DSM-5 criteria.

DISCUSSION/CONCLUSION

The CORE-DI increased the specificity of the depression diagnosis substantially compared to clinical diagnoses and the diagnoses obtained by M.I.N.I. The results point to the usefulness of an elaborated and systematic assessment of the core symptoms in the examination of patients with depressive symptoms and thereby indicate a way for further development of specific diagnostic tools for depression in both clinical and research settings. However, it should be noted that the sensitivity of the CORE-DI was modest, and the psychometric properties of the CORE-DI might be different in other settings with higher or lower prevalence or severity of depressive symptoms.

Conceptual Issues in Acute and Transient Psychotic Disorders

Short-lived psychotic disorders as currently listed under "acute and transient psychotic disorder," ICD-11 Classification of Mental, Behavioural, and Neurodevelopmental Disorders, and "brief psychotic disorder," Diagnostic and Statistical Manual of Mental Disorders (DSM-5), constitute a point of divergence in the classification of psychotic disorders between the 2 diagnostic systems, which reveals the lack of knowledge about these conditions. Whether this is due to conceptual shortcomings inherent to the categories themselves and which spill over onto research or reflects a mismatch between the diagnostic criteria used and research techniques needs clarification. This study aimed to examine conceptual issues involved in the development of the above categories and shows that little continuity exists between earlier nosological concepts such as bouffée délirante, cycloid psychosis, and reactive psychosis and modern descriptive categories used to classify short-lived psychotic disorders. It seems likely that shortcomings in terms of symptom completeness, specificity, and heterogeneity, in addition to changes in definition and diagnostic criteria in successive DSM and ICD versions, have hampered empirical research, making it difficult to enhance the understanding of these conditions and achieve a closer concordance between the 2 classificatory systems.

Tournament leave-pair-out cross-validation for receiver operating characteristic analysis

Receiver operating characteristic analysis is widely used for evaluating diagnostic systems. Recent studies have shown that estimating an area under receiver operating characteristic curve with standard cross-validation methods suffers from a large bias. The leave-pair-out cross-validation has been shown to correct this bias. However, while leave-pair-out produces an almost unbiased estimate of area under receiver operating characteristic curve, it does not provide a ranking of the data needed for plotting and analyzing the receiver operating characteristic curve. In this study, we propose a new method called tournament leave-pair-out cross-validation. This method extends leave-pair-out by creating a tournament from pair comparisons to produce a ranking for the data. Tournament leave-pair-out preserves the advantage of leave-pair-out for estimating area under receiver operating characteristic curve, while it also allows performing receiver operating characteristic analyses. We have shown using both synthetic and real-world data that tournament leave-pair-out is as reliable as leave-pair-out for area under receiver operating characteristic curve estimation and confirmed the bias in leave-one-out cross-validation on low-dimensional data. As a case study on receiver operating characteristic analysis, we also evaluate how reliably sensitivity and specificity can be estimated from tournament leave-pair-out receiver operating characteristic curves.

Multiphoton Microscopy for Caries Detection with ICDAS Classification

Dentin carious lesion is a dynamic process that involves demineralization and collagen denaturation. Collagen type I is the major protein in dentin and it has been investigated based on its optical properties. Multiphoton microscopy (MPM) is a nonlinear imaging technique that reveals the caries process using the collagen two-photon excitation fluorescence (2PEF) and its second-harmonic generation (SHG). Combining the histological and the International Caries Detection and Assessment System (ICDAS) classifications with nonlinear optical spectroscopy (NLOS), 2PEF and SHG intensities of enamel and dentin were highly altered during the caries process. It has been proven that the ratio SHG/2PEF is a relevant indicator of the organic matrix denaturation [Terrer et al.: J Dent Res 2016; 96: 574-579]. In the present study, a series of measurable signals is made to detect early stages of carious lesion according to the ICDAS classification and to explore the relationship between these measures and the ICDAS scale. Comparison of the efficiency of nonlinear optical signals for caries detection with the ICDAS classification is essential to evaluate their potential for clinical application. In our study, the use of the NLOS measured by MPM allowed us to monitor a quantitative parameter (SHG/2PEF ratio) according to the dentin carious lesion state (ICDAS and histological examination). Three coherent new groups were defined (ICDAS 0/1; ICDAS 2/3; ICDAS 4/5/6), where the carious process can be clearly described with a statistically significant decrease of the SHG/2PEF ratio.

Hepatocellular carcinoma imaging systems: why they exist, how they have evolved, and how they differ

Over the past 16 years, several scientific organizations have proposed systems that incorporate imaging for surveillance, diagnosis, staging, treatment, and monitoring of treatment response of hepatocellular carcinoma (HCC). These systems are needed to standardize the acquisition, interpretation, and reporting of liver imaging examinations; help differentiate benign from malignant observations; improve consistency between radiologists; and provide guidance for management of HCC. This review article discusses the historical evolution of HCC imaging systems. We indicate the features differentiating these systems, including target population, screening and surveillance algorithm, diagnostic imaging modalities, diagnostic scope, expertise and technical requirements, terminology, major and ancillary imaging features, staging and transplant eligibility, and assessment of treatment response. We highlight the potential benefits of unifying the systems, which we anticipate will enable sharing, pooling, and meta-analysis of data; facilitate multi-center trials; and accelerate dissemination of knowledge.

Red fluorescence of dental plaque in children -A cross-sectional study

OBJECTIVES

The relation between the presence of red fluorescent plaque and the caries status in children was studied. In addition, the microbial composition of dental plaque from sites with red fluorescent plaque (RFP) and from sites with no red fluorescent plaque (NFP) was assessed.

METHODS

Fluorescence photographs were taken from fifty children (6-14 years old) with overnight plaque. Full-mouth caries scores (ICDAS II) were obtained. The composition of a saliva sample and two plaque samples (RFP and NFP) was assessed using 16S rDNA sequencing.

RESULTS

At the site level, no clinically relevant correlations were found between the presence of RFP and the caries status. At the subject level, a weak correlation was found between RFP and the caries status when non-cavitated lesions were included (r_s=0.37, p=0.007). The microbial composition of RFP differed significantly from NFP. RFP had more anaerobes and more Gram-negative bacterial taxa. The most discriminative operational taxonomic units (OTUs) for RFP were Corynebacterium, Leptotrichia, Porphyromonas and Selenomonas, while the most discriminative OTUs for NFP were Neisseria, Actinomyces, Streptococcus and Rothia.

CONCLUSIONS

There were no clinical relevant correlations in this cross-sectional study between the presence of RFP and (early) caries lesions. There were differences in the composition of these phenotypically different plaque samples: RFP contained more Gram-negative, anaerobic taxa and was more diverse than NFP.

CLINICAL SIGNIFICANCE

The study outcomes provide more insight in the possibilities to use plaque fluorescence in oral health risk assessments.

Psychopharmacological Practice: The DSM Versus The Brain

In 1952, the Diagnostic and Statistical Manual of Mental Disorders (DSM) system of creating, validating, studying and employing a diagnostic system in clinical psychiatric practice was introduced. There have been several updates and revisions to this manual and, regardless of its a theoretical framework, it actually does have a framework and presupposition. Essentially the DSM dictates that all psychiatric disorders are syndromes, or a collection of symptoms that commonly occur together and impair psychosocial functioning. These syndromes allow for homogenous groups of patients to be studied and psychotherapies and pharmacotherapies to be developed. This editorial will examine the DSM system with regards to its applicability to central nervous system dysfunction where psychiatric disorders are concerned. Specifically, the brain does not follow categorical, or syndromal, constructs. In fact, the psychiatric patient likely inherits several risk genes that promote abnormal proteins along several neuropathways in the brain. These abnormalities create dysfunctional neurocircuits which create individual psychiatric symptoms, but not a categorical syndrome or diagnosis. The concept that the DSM may be excellent for clinical diagnostic purposes, but less correct in its assumptions for a psychopharmacologist's treatment approaches will be discussed.