An insight into diagnosis of depression using machine learning techniques: a systematic review

Diagnostic accuracy of deep learning using speech samples in depression: a systematic review and meta-analysis

OBJECTIVE

This study aims to conduct a systematic review and meta-analysis of the diagnostic accuracy of deep learning (DL) using speech samples in depression.

MATERIALS AND METHODS

This review included studies reporting diagnostic results of DL algorithms in depression using speech data, published from inception to January 31, 2024, on PubMed, Medline, Embase, PsycINFO, Scopus, IEEE, and Web of Science databases. Pooled accuracy, sensitivity, and specificity were obtained by random-effect models. The diagnostic Precision Study Quality Assessment Tool (QUADAS-2) was used to assess the risk of bias.

RESULTS

A total of 25 studies met the inclusion criteria and 8 of them were used in the meta-analysis. The pooled estimates of accuracy, specificity, and sensitivity for depression detection models were 0.87 (95% CI, 0.81-0.93), 0.85 (95% CI, 0.78-0.91), and 0.82 (95% CI, 0.71-0.94), respectively. When stratified by model structure, the highest pooled diagnostic accuracy was 0.89 (95% CI, 0.81-0.97) in the handcrafted group.

DISCUSSION

To our knowledge, our study is the first meta-analysis on the diagnostic performance of DL for depression detection from speech samples. All studies included in the meta-analysis used convolutional neural network (CNN) models, posing problems in deciphering the performance of other DL algorithms. The handcrafted model performed better than the end-to-end model in speech depression detection.

CONCLUSIONS

The application of DL in speech provided a useful tool for depression detection. CNN models with handcrafted acoustic features could help to improve the diagnostic performance.

PROTOCOL REGISTRATION

The study protocol was registered on PROSPERO (CRD42023423603).

Neural substrates of predicting anhedonia symptoms in major depressive disorder via connectome-based modeling

MAIN PROBLEM

Anhedonia is a critical diagnostic symptom of major depressive disorder (MDD), being associated with poor prognosis. Understanding the neural mechanisms underlying anhedonia is of great significance for individuals with MDD, and it encourages the search for objective indicators that can reliably identify anhedonia.

METHODS

A predictive model used connectome-based predictive modeling (CPM) for anhedonia symptoms was developed by utilizing pre-treatment functional connectivity (FC) data from 59 patients with MDD. Node-based FC analysis was employed to compare differences in FC patterns between melancholic and non-melancholic MDD patients. The support vector machines (SVM) method was then applied for classifying these two subtypes of MDD patients.

RESULTS

CPM could successfully predict anhedonia symptoms in MDD patients (positive network: r = 0.4719, p < 0.0020, mean squared error = 23.5125, 5000 iterations). Compared to non-melancholic MDD patients, melancholic MDD patients showed decreased FC between the left cingulate gyrus and the right parahippocampus gyrus (p_bonferroni = 0.0303). This distinct FC pattern effectively discriminated between melancholic and non-melancholic MDD patients, achieving a sensitivity of 93.54%, specificity of 67.86%, and an overall accuracy of 81.36% using the SVM method.

CONCLUSIONS

This study successfully established a network model for predicting anhedonia symptoms in MDD based on FC, as well as a classification model to differentiate between melancholic and non-melancholic MDD patients. These findings provide guidance for clinical treatment.

Enhancing accuracy and privacy in speech-based depression detection through speaker disentanglement

Speech signals are valuable biomarkers for assessing an individual's mental health, including identifying Major Depressive Disorder (MDD) automatically. A frequently used approach in this regard is to employ features related to speaker identity, such as speaker-embeddings. However, over-reliance on speaker identity features in mental health screening systems can compromise patient privacy. Moreover, some aspects of speaker identity may not be relevant for depression detection and could serve as a bias factor that hampers system performance. To overcome these limitations, we propose disentangling speaker-identity information from depression-related information. Specifically, we present four distinct disentanglement methods to achieve this - adversarial speaker identification (SID)-loss maximization (ADV), SID-loss equalization with variance (LEV), SID-loss equalization using Cross-Entropy (LECE) and SID-loss equalization using KL divergence (LEKLD). Our experiments, which incorporated diverse input features and model architectures, have yielded improved F1 scores for MDD detection and voice-privacy attributes, as quantified by Gain in Voice Distinctiveness G V   D and De-Identification Scores (DeID). On the DAIC-WOZ dataset (English), LECE using ComparE16 features results in the best F1-Scores of 80% which represents the audio-only SOTA depression detection F1-Score along with a G V   D of -1.1 dB and a DeID of 85%. On the EATD dataset (Mandarin), ADV using raw-audio signal achieves an F1-Score of 72.38% surpassing multi-modal SOTA along with a G V   D of -0.89 dB dB and a DeID of 51.21%. By reducing the dependence on speaker-identity-related features, our method offers a promising direction for speech-based depression detection that preserves patient privacy.

Leveraging Novel Technologies and Artificial Intelligence to Advance Practice-Oriented Research

Mental health services are experiencing notable transformations as innovative technologies and artificial intelligence (AI) are increasingly utilized in a growing number of studies and services.These cutting-edge technologies carry the promise of substantial improvements in the field of mental health. Nevertheless, questions emerge about the alignment of novel technologies and AI systems with human needs, especially in the context of vulnerable populations receiving mental healthcare. The practice-oriented research (POR) model is pivotal in seamlessly integrating these emerging technologies into clinical research and practice. It underscores the importance of tight collaboration between clinicians and researchers, all driven by the central goal of ensuring and elevating client well-being. This paper focuses on how novel technologies can enhance the POR model and highlights its pivotal role in integrating these technologies into clinical research and practice. We discuss two key phases: pre-treatment, and during treatment. For each phase, we describe the challenges, present the major technological innovations, describe recent studies exemplifying technology use, and suggest future directions. Ethical concerns and the importance of aligning humans and technology are also considered, in addition to implications for practice and training.

Improving speech depression detection using transfer learning with wav2vec 2.0 in low-resource environments

Depression, a pervasive global mental disorder, profoundly impacts daily lives. Despite numerous deep learning studies focused on depression detection through speech analysis, the shortage of annotated bulk samples hampers the development of effective models. In response to this challenge, our research introduces a transfer learning approach for detecting depression in speech, aiming to overcome constraints imposed by limited resources. In the context of feature representation, we obtain depression-related features by fine-tuning wav2vec 2.0. By integrating 1D-CNN and attention pooling structures, we generate advanced features at the segment level, thereby enhancing the model's capability to capture temporal relationships within audio frames. In the realm of prediction results, we integrate LSTM and self-attention mechanisms. This incorporation assigns greater weights to segments associated with depression, thereby augmenting the model's discernment of depression-related information. The experimental results indicate that our model has achieved impressive F1 scores, reaching 79% on the DAIC-WOZ dataset and 90.53% on the CMDC dataset. It outperforms recent baseline models in the field of speech-based depression detection. This provides a promising solution for effective depression detection in low-resource environments.

Post-surgery survival and associated factors for cardiac patients in Ethiopia: applications of machine learning, semi-parametric and parametric modelling

INTRODUCTION

Living in poverty, especially in low-income countries, are more affected by cardiovascular disease. Unlike the developed countries, it remains a significant cause of preventable heart disease in the Sub-Saharan region, including Ethiopia. According to the Ethiopian Ministry of Health statement, around 40,000 cardiac patients have been waiting for surgery in Ethiopia since September 2020. There is insufficient information about long-term cardiac patients' post-survival after cardiac surgery in Ethiopia. Therefore, the main objective of the current study was to determine the long-term post-cardiac surgery patients' survival status in Ethiopia.

METHODS

All patients attended from 2012 to 2023 throughout the country were included in the current study. The total number of participants was 1520 heart disease patients. The data collection procedure was conducted from February 2022- January 2023. Machine learning algorithms were applied. Gompertz regression was used also for the multivariable analysis report.

RESULTS

From possible machine learning models, random survival forest were preferred. It emphasizes, the most important variable for clinical prediction was SPO2, Age, time to surgery waiting time, and creatinine value and it accounts, 42.55%, 25.17%,11.82%, and 12.19% respectively. From the Gompertz regression, lower saturated oxygen, higher age, lower ejection fraction, short period of cardiac center stays after surgery, prolonged waiting time to surgery, and creating value were statistically significant predictors of death outcome for post-cardiac surgery patients' survival in Ethiopia.

CONCLUSION

Some of the risk factors for the death of post-cardiac surgery patients are identified in the current investigation. Particular attention should be given to patients with prolonged waiting times and aged patients. Since there were only two fully active cardiac centers in Ethiopia it is far from an adequate number of centers for more than 120 million population, therefore, the study highly recommended to increase the number of cardiac centers that serve as cardiac surgery in Ethiopia.

Adapting the Number of Questions Based on Detected Psychological Distress for Cognitive Behavioral Therapy With an Embodied Conversational Agent: Comparative Study

BACKGROUND

The high prevalence of mental illness is a critical social problem. The limited availability of mental health services is a major factor that exacerbates this problem. One solution is to deliver cognitive behavioral therapy (CBT) using an embodied conversational agent (ECA). ECAs make it possible to provide health care without location or time constraints. One of the techniques used in CBT is Socratic questioning, which guides users to correct negative thoughts. The effectiveness of this approach depends on a therapist's skill to adapt to the user's mood or distress level. However, current ECAs do not possess this skill. Therefore, it is essential to implement this adaptation ability to the ECAs.

OBJECTIVE

This study aims to develop and evaluate a method that automatically adapts the number of Socratic questions based on the level of detected psychological distress during a CBT session with an ECA. We hypothesize that this adaptive approach to selecting the number of questions will lower psychological distress, reduce negative emotional states, and produce more substantial cognitive changes compared with a random number of questions.

METHODS

In this study, which envisions health care support in daily life, we recruited participants aged from 18 to 65 years for an experiment that involved 2 different conditions: an ECA that adapts a number of questions based on psychological distress detection or an ECA that only asked a random number of questions. The participants were assigned to 1 of the 2 conditions, experienced a single CBT session with an ECA, and completed questionnaires before and after the session.

RESULTS

The participants completed the experiment. There were slight differences in sex, age, and preexperimental psychological distress levels between the 2 conditions. The adapted number of questions condition showed significantly lower psychological distress than the random number of questions condition after the session. We also found a significant difference in the cognitive change when the number of questions was adapted based on the detected distress level, compared with when the number of questions was fewer than what was appropriate for the level of distress detected.

CONCLUSIONS

The results show that an ECA adapting the number of Socratic questions based on detected distress levels increases the effectiveness of CBT. Participants who received an adaptive number of questions experienced greater reductions in distress than those who received a random number of questions. In addition, the participants showed a greater amount of cognitive change when the number of questions matched the detected distress level. This suggests that adapting the question quantity based on distress level detection can improve the results of CBT delivered by an ECA. These results illustrate the advantages of ECAs, paving the way for mental health care that is more tailored and effective.

Machine learning (ML) techniques to predict breast cancer in imbalanced datasets: a systematic review

Knowledge discovery in databases (KDD) is crucial in analyzing data to extract valuable insights. In medical outcome prediction, KDD is increasingly applied, particularly in diseases with high incidence, mortality, and costs, like cancer. ML techniques can develop more accurate predictive models for cancer patients' clinical outcomes, aiding informed healthcare decision-making. However, cancer prediction modeling faces challenges because of the unbalanced nature of the datasets, where there is a small minority category of patients with a cancer diagnosis compared to a majority category of cancer-free patients. Imbalanced datasets pose statistical hurdles like bias and overfitting when developing accurate prediction models. This systematic review focuses on breast cancer prediction articles published from 2008 to 2023. The objective is to examine ML methods used in three critical steps of KDD: preprocessing, data mining, and interpretation which address the imbalanced data problem in breast cancer prediction. This work synthesizes prior research in ML methods for breast cancer prediction. The findings help identify effective preprocessing strategies, including balancing and feature selection methods, robust predictive models, and evaluation metrics of those models. The study aims to inform healthcare providers and researchers about effective techniques for accurate breast cancer prediction.

Identification of rice leaf diseases and deficiency disorders using a novel DeepBatch technique

Rice is one of the most widely consumed foods all over the world. Various diseases and deficiency disorders impact the rice crop's growth, thereby hampering the rice yield. Therefore, proper crop monitoring is very important for the early diagnosis of diseases or deficiency disorders. Diagnosis of diseases and disorders requires specialized manpower, which is not scalable and accessible to all farmers. To address this issue, machine learning and deep learning (DL)-driven automated systems are designed, which may help the farmers in diagnosing disease/deficiency disorders in crops so that proper care can be taken on time. Various studies have used transfer learning (TL) models in the recent past. In recent studies, further improvement in rice disease and deficiency disorder diagnosis system performance is achieved by performing the ensemble of various TL models. However, in all these DL-based studies, the segmentation of the region of interest is not done beforehand and the infected-region extraction is left for the DL model to handle automatically. Therefore, this article proposes a novel framework for the diagnosis of rice-infected leaves based on DL-based segmentation with bitwise logical AND operation and DL-based classification. The rice diseases covered in this study are bacterial leaf blight, brown spot, and leaf smut. The rice nutrient deficiencies like nitrogen (N), phosphorous (P), and potassium (K) were also included. The results of the experiment conducted on these datasets showed that the performance of DeepBatch was significantly improved as compared to the conventional technique.

A systematic review on machine learning and deep learning techniques in the effective diagnosis of Alzheimer's disease

Alzheimer's disease (AD) is a brain-related disease in which the condition of the patient gets worse with time. AD is not a curable disease by any medication. It is impossible to halt the death of brain cells, but with the help of medication, the effects of AD can be delayed. As not all MCI patients will suffer from AD, it is required to accurately diagnose whether a mild cognitive impaired (MCI) patient will convert to AD (namely MCI converter MCI-C) or not (namely MCI non-converter MCI-NC), during early diagnosis. There are two modalities, positron emission tomography (PET) and magnetic resonance image (MRI), used by a physician for the diagnosis of Alzheimer's disease. Machine learning and deep learning perform exceptionally well in the field of computer vision where there is a requirement to extract information from high-dimensional data. Researchers use deep learning models in the field of medicine for diagnosis, prognosis, and even to predict the future health of the patient under medication. This study is a systematic review of publications using machine learning and deep learning methods for early classification of normal cognitive (NC) and Alzheimer's disease (AD).This study is an effort to provide the details of the two most commonly used modalities PET and MRI for the identification of AD, and to evaluate the performance of both modalities while working with different classifiers.

Emotion regulation in bipolar disorder type-I: multivariate analysis of fMRI data

BACKGROUND

Bipolar disorder type-I (BD-I) patients are known to show emotion regulation abnormalities. In a previous fMRI study using an explicit emotion regulation paradigm, we compared responses from 19 BD-I patients and 17 matched healthy controls (HC). A standard general linear model-based univariate analysis revealed that BD patients showed increased activations in inferior frontal gyrus when instructed to decrease their emotional response as elicited by neutral images. We implemented multivariate pattern recognition analyses on the same data to examine if we could classify conditions within-group as well as HC versus BD.

METHODS

We reanalyzed explicit emotion regulation data using a multivariate pattern recognition approach, as implemented in PRONTO software. The original experimental paradigm consisted of a full 2 × 2 factorial design, with valence (Negative/Neutral) and instruction (Look/Decrease) as within subject factors.

RESULTS

The multivariate models were able to accurately classify different task conditions when HC and BD were analyzed separately (63.24%-75.00%, p = 0.001-0.012). In addition, the models were able to correctly classify HC versus BD with significant accuracy in conditions where subjects were instructed to downregulate their felt emotion (59.60%-60.84%, p = 0.014-0.018). The results for HC versus BD classification demonstrated contributions from the salience network, several occipital and frontal regions, inferior parietal lobes, as well as other cortical regions, to achieve above-chance classifications.

CONCLUSIONS

Our multivariate analysis successfully reproduced some of the main results obtained in the previous univariate analysis, confirming that these findings are not dependent on the analysis approach. In particular, both types of analyses suggest that there is a significant difference of neural patterns between conditions within each subject group. The multivariate approach also revealed that reappraisal conditions provide the most informative activity for differentiating HC versus BD, irrespective of emotional valence (negative or neutral). The current results illustrate the importance of investigating the cognitive control of emotion in BD. We also propose a set of candidate regions for further study of emotional control in BD.

Semantic segmentation model for land cover classification from satellite images in Gambella National Park, Ethiopia

This work uses machine learning approaches to present semantic segmentation for land cover classification in Gambella National Park (GNP). Land cover classification has become more accurate due to developments in remote sensing data. Land cover classification from satellite images has been studied, but the methodologies and satellite data employed so far are not suitable for research regions with the possibility to find heterogeneous land cover classes within small areas. Previous studies found issues with the satellite images coarser spatial resolution, the use of standard statistical methods as classifiers, and the difficulty in optimal patch size selection when patch-based classification is used. To address these issues, we suggested a deep learning-based semantic segmentation model that could be utilized as a pixel-level land cover classification technique. The suggested technique employed high-resolution Sentinel-2 satellite images of our study area (GNP) as a dataset and constructed and assessed pixel-level classification models. As a deep learning-based classification model, we have used the Link-Net architecture and its encoder part was modified further to incorporate the state-of-the-art architecture called ResNet34. The developed models, support vector machine with CNN features (CNN–SVM), random forest with CNN features (CNN-RF), LinkNet model with ResNet-34 as encoder (LinkNet-ResNet34), attain average F1-Score values of 81%,82%, and 87.4% respectively.

Role of Ultrasound Imaging in the Prediction of TRIM67 in Brain Metastases From Breast Cancer

Objectives

Ultrasound (US) imaging is a relatively novel strategy to monitor the activity of the blood–brain barrier, which can facilitate the diagnosis and treatment of neurovascular-related metastatic tumors. The purpose of this study was to investigate the clinical significance of applying a combination of US imaging outcomes and the associated genes. This was performed to construct line drawings to facilitate the prediction of brain metastases arising from breast cancer.

Methods

The RNA transcript data from The Cancer Genome Atlas (TCGA) database was obtained for breast cancer, and the differentially expressed genes (DEGs) associated with tumor and brain tumor metastases were identified. Subsequently, key genes associated with survival prognosis were subsequently identified from the DEGs.

Results

Tripartite motif-containing protein 67 (TRIM67) was identified and the differential; in addition, the survival analyses of the TCGA database revealed that it was associated with brain tumor metastases and overall survival prognosis. Applying independent clinical cohort data, US-related features (microcalcification and lymph node metastasis) were associated with breast cancer tumor metastasis. Furthermore, ultrasonographic findings of microcalcifications showed correlations with TRIM67 expression. The study results revealed that six variables [stage, TRIM67, tumor size, regional lymph node staging (N), age, and HER2 status] were suitable predictors of tumor metastasis by applying support vector machine–recursive feature elimination. Among these, US-predicted tumor size correlated with tumor size classification, whereas US-predicted lymph node metastasis correlated with tumor N classification. The TRIM67 upregulation was accompanied by upregulation of the integrated breast cancer pathway; however, it leads to the downregulation of the miRNA targets in ECM and membrane receptors and the miRNAs involved in DNA damage response pathways.

Conclusions

The TRIM67 is a risk factor associated with brain metastases from breast cancer and it is considered a prognostic survival factor. The nomogram constructed from six variables—stage, TRIM67, tumor size, N, age, HER2 status—is an appropriate predictor to estimate the occurrence of breast cancer metastasis.