LSCS-Net: A lightweight skin cancer segmentation network with densely connected multi-rate atrous convolution.

The rising occurrence and notable public health consequences of skin cancer, especially of the most challenging form known as melanoma, have created an urgent demand for more advanced approaches to disease management. The integration of modern computer vision methods into clinical procedures offers the potential for enhancing the detection of skin cancer . The UNet model has gained prominence as a valuable tool for this objective, continuously evolving to tackle the difficulties associated with the inherent diversity of dermatological images. These challenges stem from diverse medical origins and are further complicated by variations in lighting, patient characteristics, and hair density. In this work, we present an innovative end-to-end trainable network crafted for the segmentation of skin cancer . This network comprises an encoder-decoder architecture, a novel feature extraction block, and a densely connected multi-rate Atrous convolution block. We evaluated the performance of the proposed lightweight skin cancer segmentation network (LSCS-Net) on three widely used benchmark datasets for skin lesion segmentation: ISIC 2016, ISIC 2017, and ISIC 2018. The generalization capabilities of LSCS-Net are testified by the excellent performance on breast cancer and thyroid nodule segmentation datasets. The empirical findings confirm that LSCS-net attains state-of-the-art results, as demonstrated by a significantly elevated Jaccard index.

ECG-based cardiac arrhythmias detection through ensemble learning and fusion of deep spatial-temporal and long-range dependency features.

Cardiac arrhythmia is one of the prime reasons for death globally. Early diagnosis of heart arrhythmia is crucial to provide timely medical treatment. Heart arrhythmias are diagnosed by analyzing the electrocardiogram (ECG) of patients. Manual analysis of ECG is time-consuming and challenging. Hence, effective automated detection of heart arrhythmias is important to produce reliable results. Different deep-learning techniques to detect heart arrhythmias such as Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), Transformer, and Hybrid CNN-LSTM were proposed. However, these techniques, when used individually, are not sufficient to effectively learn multiple features from the ECG signal. The fusion of CNN and LSTM overcomes the limitations of CNN in the existing studies as CNN-LSTM hybrids can extract spatiotemporal features. However, LSTMs suffer from long-range dependency issues due to which certain features may be ignored. Hence, to compensate for the drawbacks of the existing models, this paper proposes a more comprehensive feature fusion technique by merging CNN, LSTM, and Transformer models. The fusion of these models facilitates learning spatial, temporal, and long-range dependency features, hence, helping to capture different attributes of the ECG signal. These features are subsequently passed to a majority voting classifier equipped with three traditional base learners. The traditional learners are enriched with deep features instead of handcrafted features. Experiments are performed on the MIT-BIH arrhythmias database and the model performance is compared with that of the state-of-art models. Results reveal that the proposed model performs better than the existing models yielding an accuracy of 99.56%.

Estimating age and gender from electrocardiogram signals: A comprehensive review of the past decade.

Twelve lead electrocardiogram signals capture unique fingerprints about the body's biological processes and electrical activity of heart muscles. Machine learning and deep learning-based models can learn the embedded patterns in the electrocardiogram to estimate complex metrics such as age and gender that depend on multiple aspects of human physiology. ECG estimated age with respect to the chronological age reflects the overall well-being of the cardiovascular system, with significant positive deviations indicating an aged cardiovascular system and a higher likelihood of cardiovascular mortality. Several conventional, machine learning, and deep learning-based methods have been proposed to estimate age from electronic health records, health surveys, and ECG data. This manuscript comprehensively reviews the methodologies proposed for ECG-based age and gender estimation over the last decade. Specifically, the review highlights that elevated ECG age is associated with atherosclerotic cardiovascular disease, abnormal peripheral endothelial dysfunction, and high mortality, among many other cardiovascular disorders. Furthermore, the survey presents overarching observations and insights across methods for age and gender estimation. This paper also presents several essential methodological improvements and clinical applications of ECG-estimated age and gender to encourage further improvements of the state-of-the-art methodologies.

Unveiling the future of breast cancer assessment: a critical review on generative adversarial networks in elastography ultrasound.

Elastography Ultrasound provides elasticity information of the tissues, which is crucial for understanding the density and texture, allowing for the diagnosis of different medical conditions such as fibrosis and cancer. In the current medical imaging scenario, elastograms for B-mode Ultrasound are restricted to well-equipped hospitals, making the modality unavailable for pocket ultrasound. To highlight the recent progress in elastogram synthesis, this article performs a critical review of generative adversarial network (GAN) methodology for elastogram generation from B-mode Ultrasound images. Along with a brief overview of cutting-edge medical image synthesis, the article highlights the contribution of the GAN framework in light of its impact and thoroughly analyzes the results to validate whether the existing challenges have been effectively addressed. Specifically, This article highlights that GANs can successfully generate accurate elastograms for deep-seated breast tumors (without having artifacts) and improve diagnostic effectiveness for pocket US. Furthermore, the results of the GAN framework are thoroughly analyzed by considering the quantitative metrics, visual evaluations, and cancer diagnostic accuracy. Finally, essential unaddressed challenges that lie at the intersection of elastography and GANs are presented, and a few future directions are shared for the elastogram synthesis research.

Deep learning for ECG Arrhythmia detection and classification: an overview of progress for period 2017-2023.

Cardiovascular diseases are a leading cause of mortality globally. Electrocardiography (ECG) still represents the benchmark approach for identifying cardiac irregularities. Automatic detection of abnormalities from the ECG can aid in the early detection, diagnosis, and prevention of cardiovascular diseases. Deep Learning (DL) architectures have been successfully employed for arrhythmia detection and classification and offered superior performance to traditional shallow Machine Learning (ML) approaches. This survey categorizes and compares the DL architectures used in ECG arrhythmia detection from 2017-2023 that have exhibited superior performance. Different DL models such as Convolutional Neural Networks (CNNs), Multilayer Perceptrons (MLPs), Transformers, and Recurrent Neural Networks (RNNs) are reviewed, and a summary of their effectiveness is provided. This survey provides a comprehensive roadmap to expedite the acclimation process for emerging researchers willing to develop efficient algorithms for detecting ECG anomalies using DL models. Our tailored guidelines bridge the knowledge gap allowing newcomers to align smoothly with the prevailing research trends in ECG arrhythmia detection. We shed light on potential areas for future research and refinement in model development and optimization, intending to stimulate advancement in ECG arrhythmia detection and classification.

Unsupervised anomaly appraisal of cleft faces using a StyleGAN2-based model adaptation technique.

A novel machine learning framework that is able to consistently detect, localize, and measure the severity of human congenital cleft lip anomalies is introduced. The ultimate goal is to fill an important clinical void: to provide an objective and clinically feasible method of gauging baseline facial deformity and the change obtained through reconstructive surgical intervention. The proposed method first employs the StyleGAN2 generative adversarial network with model adaptation to produce a normalized transformation of 125 faces, and then uses a pixel-wise subtraction approach to assess the difference between all baseline images and their normalized counterparts (a proxy for severity of deformity). The pipeline of the proposed framework consists of the following steps: image preprocessing, face normalization, color transformation, heat-map generation, morphological erosion, and abnormality scoring. Heatmaps that finely discern anatomic anomalies visually corroborate the generated scores. The proposed framework is validated through computer simulations as well as by comparison of machine-generated versus human ratings of facial images. The anomaly scores yielded by the proposed computer model correlate closely with human ratings, with a calculated Pearson's r score of 0.89. The proposed pixel-wise measurement technique is shown to more closely mirror human ratings of cleft faces than two other existing, state-of-the-art image quality metrics (Learned Perceptual Image Patch Similarity and Structural Similarity Index). The proposed model may represent a new standard for objective, automated, and real-time clinical measurement of faces affected by congenital cleft deformity.

MetaAnalyst: a user-friendly tool for metagenomic biomarker detection and phenotype classification.

BACKGROUND: Many metagenomic studies have linked the imbalance in microbial abundance profiles to a wide range of diseases. These studies suggest utilizing the microbial abundance profiles as potential markers for metagenomic-associated conditions. Due to the inevitable importance of biomarkers in understanding the disease progression and the development of possible therapies, various computational tools have been proposed for metagenomic biomarker detection. However, most existing tools require prior scripting knowledge and lack user friendly interfaces, causing considerable time and effort to install, configure, and run these tools. Besides, there is no available all-in-one solution for running and comparing various metagenomic biomarker detection simultaneously. In addition, most of these tools just present the suggested biomarkers without any statistical evaluation for their quality. RESULTS: To overcome these limitations, this work presents MetaAnalyst, a software package with a simple graphical user interface (GUI) that (i) automates the installation and configuration of 28 state-of-the-art tools, (ii) supports flexible study design to enable studying the dataset under different scenarios smoothly, iii) runs and evaluates several algorithms simultaneously iv) supports different input formats and provides the user with several preprocessing capabilities, v) provides a variety of metrics to evaluate the quality of the suggested markers, and vi) presents the outcomes in the form of publication quality plots with various formatting capabilities as well as Excel sheets. CONCLUSIONS: The utility of this tool has been verified through studying a metagenomic dataset under four scenarios. The executable file for MetaAnalyst along with its user manual are made available at https://github.com/mshawaqfeh/MetaAnalyst .

Gauging Facial Abnormality Using Haar-Cascade Object Detector.

The overriding clinical and academic challenge that inspires this work is the lack of a universally accepted, objective, and feasible method of measuring facial deformity; and, by extension, the lack of a reliable means of assessing the benefits and shortcomings of craniofacial surgical interventions. We propose a machine learning-based method to create a scale of facial deformity by producing numerical scores that reflect the level of deformity. An object detector that is constructed using a cascade function of Haar features has been trained with a rich dataset of normal faces in addition to a collection of images that does not contain faces. After that, the confidence score of the face detector was used as a gauge of facial abnormality. The scores were compared with a benchmark that is based on human appraisals obtained using a survey of a range of facial deformities. Interestingly, the overall Pearson's correlation coefficient of the machine scores with respect to the average human score exceeded 0.96.

Toward a Universal Measure of Facial Difference Using Two Novel Machine Learning Models.

A sensitive, objective, and universally accepted method of measuring facial deformity does not currently exist. Two distinct machine learning methods are described here that produce numerical scores reflecting the level of deformity of a wide variety of facial conditions. METHODS: The first proposed technique utilizes an object detector based on a cascade function of Haar features. The model was trained using a dataset of 200,000 normal faces, as well as a collection of images devoid of faces. With the model trained to detect normal faces, the face detector confidence score was shown to function as a reliable gauge of facial abnormality. The second technique developed is based on a deep learning architecture of a convolutional autoencoder trained with the same rich dataset of normal faces. Because the convolutional autoencoder regenerates images disposed toward their training dataset (ie, normal faces), we utilized its reconstruction error as an indicator of facial abnormality. Scores generated by both methods were compared with human ratings obtained using a survey of 80 subjects evaluating 60 images depicting a range of facial deformities [rating from 1 (abnormal) to 7 (normal)]. RESULTS: The machine scores were highly correlated to the average human score, with overall Pearson's correlation coefficient exceeding 0.96 (P < 0.00001). Both methods were computationally efficient, reporting results within 3 seconds. CONCLUSIONS: These models show promise for adaptation into a clinically accessible handheld tool. It is anticipated that ongoing development of this technology will facilitate multicenter collaboration and comparison of outcomes between conditions, techniques, operators, and institutions.

Personalized quantification of facial normality: a machine learning approach.

What is a normal face? A fundamental task for the facial reconstructive surgeon is to answer that question as it pertains to any given individual. Accordingly, it would be important to be able to place the facial appearance of a patient with congenital or acquired deformity numerically along their own continuum of normality, and to measure any surgical changes against such a personalized benchmark. This has not previously been possible. We have solved this problem by designing a computerized model that produces realistic, normalized versions of any given facial image, and objectively measures the perceptual distance between the raw and normalized facial image pair. The model is able to faithfully predict human scoring of facial normality. We believe this work represents a paradigm shift in the assessment of the human face, holding great promise for development as an objective tool for surgical planning, patient education, and as a means for clinical outcome measurement.

Robust Recurrent CNV Detection in the Presence of Inter-Subject Variability.

The study of recurrent copy number variations (CNVs) plays an important role in understanding the onset and evolution of complex diseases such as cancer. Array-based comparative genomic hybridization (aCGH) is a widely used microarray based technology for identifying CNVs. However, due to high noise levels and inter-sample variability, detecting recurrent CNVs from aCGH data remains a challenging topic. This paper proposes a novel method for identification of the recurrent CNVs. In the proposed method, the noisy aCGH data is modeled as the superposition of three matrices: a full-rank matrix of weighted piece-wise generating signals accounting for the clean aCGH data, a Gaussian noise matrix to model the inherent experimentation errors and other sources of error, and a sparse matrix to capture the sparse inter-sample (sample-specific) variations. We demonstrated the ability of our method to separate accurately recurrent CNVs from sample-specific variations and noise in both simulated (artificial) data and real data. The proposed method produced more accurate results than current state-of-the-art methods used in recurrent CNV detection and exhibited robustness to noise and sample-specific variations.

Single-Cell Expression Variability Implies Cell Function.

As single-cell RNA sequencing (scRNA-seq) data becomes widely available, cell-to-cell variability in gene expression, or single-cell expression variability (scEV), has been increasingly appreciated. However, it remains unclear whether this variability is functionally important and, if so, what are its implications for multi-cellular organisms. Here, we analyzed multiple scRNA-seq data sets from lymphoblastoid cell lines (LCLs), lung airway epithelial cells (LAECs), and dermal fibroblasts (DFs) and, for each cell type, selected a group of homogenous cells with highly similar expression profiles. We estimated the scEV levels for genes after correcting the mean-variance dependency in that data and identified 465, 466, and 364 highly variable genes (HVGs) in LCLs, LAECs, and DFs, respectively. Functions of these HVGs were found to be enriched with those biological processes precisely relevant to the corresponding cell type's function, from which the scRNA-seq data used to identify HVGs were generated-e.g., cytokine signaling pathways were enriched in HVGs identified in LCLs, collagen formation in LAECs, and keratinization in DFs. We repeated the same analysis with scRNA-seq data from induced pluripotent stem cells (iPSCs) and identified only 79 HVGs with no statistically significant enriched functions; the overall scEV in iPSCs was of negligible magnitude. Our results support the "variation is function" hypothesis, arguing that scEV is required for cell type-specific, higher-level system function. Thus, quantifying and characterizing scEV are of importance for our understating of normal and pathological cellular processes.

Single-cell RNA sequencing of a European and an African lymphoblastoid cell line.

In biomedical research, lymphoblastoid cell lines (LCLs), often established by in vitro infection of resting B cells with Epstein-Barr virus, are commonly used as surrogates for peripheral blood lymphocytes. Genomic and transcriptomic information on LCLs has been used to study the impact of genetic variation on gene expression in humans. Here we present single-cell RNA sequencing (scRNA-seq) data on GM12878 and GM18502-two LCLs derived from the blood of female donors of European and African ancestry, respectively. Cells from three samples (the two LCLs and a 1:1 mixture of the two) were prepared separately using a 10x Genomics Chromium Controller and deeply sequenced. The final dataset contained 7,045 cells from GM12878, 5,189 from GM18502, and 5,820 from the mixture, offering valuable information on single-cell gene expression in highly homogenous cell populations. This dataset is a suitable reference for population differentiation in gene expression at the single-cell level. Data from the mixture provide additional valuable information facilitating the development of statistical methods for data normalization and batch effect correction.

Simulating variance heterogeneity in quantitative genome wide association studies.

BACKGROUND: Analyzing Variance heterogeneity in genome wide association studies (vGWAS) is an emerging approach for detecting genetic loci involved in gene-gene and gene-environment interactions. vGWAS analysis detects variability in phenotype values across genotypes, as opposed to typical GWAS analysis, which detects variations in the mean phenotype value. RESULTS: A handful of vGWAS analysis methods have been recently introduced in the literature. However, very little work has been done for evaluating these methods. To enable the development of better vGWAS analysis methods, this work presents the first quantitative vGWAS simulation procedure. To that end, we describe the mathematical framework and algorithm for generating quantitative vGWAS phenotype data from genotype profiles. Our simulation model accounts for both haploid and diploid genotypes under different modes of dominance. Our model is also able to simulate any number of genetic loci causing mean and variance heterogeneity. CONCLUSIONS: We demonstrate the utility of our simulation procedure through generating a variety of genetic loci types to evaluate common GWAS and vGWAS analysis methods. The results of this evaluation highlight the challenges current tools face in detecting GWAS and vGWAS loci.

SparseNCA: Sparse Network Component Analysis for Recovering Transcription Factor Activities with Incomplete Prior Information.

Network component analysis (NCA) is an important method for inferring transcriptional regulatory networks (TRNs) and recovering transcription factor activities (TFAs) using gene expression data, and the prior information about the connectivity matrix. The algorithms currently available crucially depend on the completeness of this prior information. However, inaccuracies in the measurement process may render incompleteness in the available knowledge about the connectivity matrix. Hence, computationally efficient algorithms are needed to overcome the possible incompleteness in the available data. We present a sparse network component analysis algorithm (sparseNCA), which incorporates the effect of incompleteness in the estimation of TRNs by imposing an additional sparsity constraint using the norm, which results in a greater estimation accuracy. In order to improve the computational efficiency, an iterative re-weighted method is proposed for the NCA problem which not only promotes sparsity but is hundreds of times faster than the norm based solution. The performance of sparseNCA is rigorously compared to that of FastNCA and NINCA using synthetic data as well as real data. It is shown that sparseNCA outperforms the existing state-of-the-art algorithms both in terms of estimation accuracy and consistency with the added advantage of low computational complexity. The performance of sparseNCA compared to its predecessors is particularly pronounced in case of incomplete prior information about the sparsity of the network. Subnetwork analysis is performed on the E.coli data which reiterates the superior consistency of the proposed algorithm.

Reliable Biomarker discovery from Metagenomic data via RegLRSD algorithm.

BACKGROUND: Biomarker detection presents itself as a major means of translating biological data into clinical applications. Due to the recent advances in high throughput sequencing technologies, an increased number of metagenomics studies have suggested the dysbiosis in microbial communities as potential biomarker for certain diseases. The reproducibility of the results drawn from metagenomic data is crucial for clinical applications and to prevent incorrect biological conclusions. The variability in the sample size and the subjects participating in the experiments induce diversity, which may drastically change the outcome of biomarker detection algorithms. Therefore, a robust biomarker detection algorithm that ensures the consistency of the results irrespective of the natural diversity present in the samples is needed. RESULTS: Toward this end, this paper proposes a novel Regularized Low Rank-Sparse Decomposition (RegLRSD) algorithm. RegLRSD models the bacterial abundance data as a superposition between a sparse matrix and a low-rank matrix, which account for the differentially and non-differentially abundant microbes, respectively. Hence, the biomarker detection problem is cast as a matrix decomposition problem. In order to yield more consistent and solid biological conclusions, RegLRSD incorporates the prior knowledge that the irrelevant microbes do not exhibit significant variation between samples belonging to different phenotypes. Moreover, an efficient algorithm to extract the sparse matrix is proposed. Comprehensive comparisons of RegLRSD with the state-of-the-art algorithms on three realistic datasets are presented. The obtained results demonstrate that RegLRSD consistently outperforms the other algorithms in terms of reproducibility performance and provides a marker list with high classification accuracy. CONCLUSIONS: The proposed RegLRSD algorithm for biomarker detection provides high reproducibility and classification accuracy performance regardless of the dataset complexity and the number of selected biomarkers. This renders RegLRSD as a reliable and powerful tool for identifying potential metagenomic biomarkers.

Inferring microbial interaction networks from metagenomic data using SgLV-EKF algorithm.

BACKGROUND: Inferring the microbial interaction networks (MINs) and modeling their dynamics are critical in understanding the mechanisms of the bacterial ecosystem and designing antibiotic and/or probiotic therapies. Recently, several approaches were proposed to infer MINs using the generalized Lotka-Volterra (gLV) model. Main drawbacks of these models include the fact that these models only consider the measurement noise without taking into consideration the uncertainties in the underlying dynamics. Furthermore, inferring the MIN is characterized by the limited number of observations and nonlinearity in the regulatory mechanisms. Therefore, novel estimation techniques are needed to address these challenges. RESULTS: This work proposes SgLV-EKF: a stochastic gLV model that adopts the extended Kalman filter (EKF) algorithm to model the MIN dynamics. In particular, SgLV-EKF employs a stochastic modeling of the MIN by adding a noise term to the dynamical model to compensate for modeling uncertainties. This stochastic modeling is more realistic than the conventional gLV model which assumes that the MIN dynamics are perfectly governed by the gLV equations. After specifying the stochastic model structure, we propose the EKF to estimate the MIN. SgLV-EKF was compared with two similarity-based algorithms, one algorithm from the integral-based family and two regression-based algorithms, in terms of the achieved performance on two synthetic data-sets and two real data-sets. The first data-set models the randomness in measurement data, whereas, the second data-set incorporates uncertainties in the underlying dynamics. The real data-sets are provided by a recent study pertaining to an antibiotic-mediated Clostridium difficile infection. The experimental results demonstrate that SgLV-EKF outperforms the alternative methods in terms of robustness to measurement noise, modeling errors, and tracking the dynamics of the MIN. CONCLUSIONS: Performance analysis demonstrates that the proposed SgLV-EKF algorithm represents a powerful and reliable tool to infer MINs and track their dynamics.

Consistent metagenomic biomarker detection via robust PCA.

BACKGROUND: Recent developments of high throughput sequencing technologies allow the characterization of the microbial communities inhabiting our world. Various metagenomic studies have suggested using microbial taxa as potential biomarkers for certain diseases. In practice, the number of available samples varies from experiment to experiment. Therefore, a robust biomarker detection algorithm is needed to provide a set of potential markers irrespective of the number of available samples. Consistent performance is essential to derive solid biological conclusions and to transfer these findings into clinical applications. Surprisingly, the consistency of a metagenomic biomarker detection algorithm with respect to the variation in the experiment size has not been addressed by the current state-of-art algorithms. RESULTS: We propose a consistency-classification framework that enables the assessment of consistency and classification performance of a biomarker discovery algorithm. This evaluation protocol is based on random resampling to mimic the variation in the experiment size. Moreover, we model the metagenomic data matrix as a superposition of two matrices. The first matrix is a low-rank matrix that models the abundance levels of the irrelevant bacteria. The second matrix is a sparse matrix that captures the abundance levels of the bacteria that are differentially abundant between different phenotypes. Then, we propose a novel Robust Principal Component Analysis (RPCA) based biomarker discovery algorithm to recover the sparse matrix. RPCA belongs to the class of multivariate feature selection methods which treat the features collectively rather than individually. This provides the proposed algorithm with an inherent ability to handle the complex microbial interactions. Comprehensive comparisons of RPCA with the state-of-the-art algorithms on two realistic datasets are conducted. Results show that RPCA consistently outperforms the other algorithms in terms of classification accuracy and reproducibility performance. CONCLUSIONS: The RPCA-based biomarker detection algorithm provides a high reproducibility performance irrespective of the complexity of the dataset or the number of selected biomarkers. Also, RPCA selects biomarkers with quite high discriminative accuracy. Thus, RPCA is a consistent and accurate tool for selecting taxanomical biomarkers for different microbial populations. REVIEWERS: This article was reviewed by Masanori Arita and Zoltan Gaspari.

Anatomical Region-Specific In Vivo Wireless Communication Channel Characterization.

In vivo wireless body area networks and their associated technologies are shaping the future of healthcare by providing continuous health monitoring and noninvasive surgical capabilities, in addition to remote diagnostic and treatment of diseases. To fully exploit the potential of such devices, it is necessary to characterize the communication channel, which will help to build reliable and high-performance communication systems. This paper presents an in vivo wireless communication channel characterization for male torso both numerically and experimentally (on a human cadaver) considering various organs at 915 MHz and 2.4 GHz. A statistical path loss (PL) model is introduced, and the anatomical region-specific parameters are provided. It is found that the mean PL in decibel scale exhibits a linear decaying characteristic rather than an exponential decaying profile inside the body, and the power decay rate is approximately twice at 2.4 GHz as compared to 915 MHz. Moreover, the variance of shadowing increases significantly as the in vivo antenna is placed deeper inside the body since the main scatterers are present in the vicinity of the antenna. Multipath propagation characteristics are also investigated to facilitate proper waveform designs in the future wireless healthcare systems, and a root-mean-square delay spread of 2.76 ns is observed at 5 cm depth. Results show that the in vivo channel exhibit different characteristics than the classical communication channels, and location dependence is very critical for accurate, reliable, and energy-efficient link budget calculations.

CrossLink: a novel method for cross-condition classification of cancer subtypes.

BACKGROUND: We considered the prediction of cancer classes (e.g. subtypes) using patient gene expression profiles that contain both systematic and condition-specific biases when compared with the training reference dataset. The conventional normalization-based approaches cannot guarantee that the gene signatures in the reference and prediction datasets always have the same distribution for all different conditions as the class-specific gene signatures change with the condition. Therefore, the trained classifier would work well under one condition but not under another. METHODS: To address the problem of current normalization approaches, we propose a novel algorithm called CrossLink (CL). CL recognizes that there is no universal, condition-independent normalization mapping of signatures. In contrast, it exploits the fact that the signature is unique to its associated class under any condition and thus employs an unsupervised clustering algorithm to discover this unique signature. RESULTS: We assessed the performance of CL for cross-condition predictions of PAM50 subtypes of breast cancer by using a simulated dataset modeled after TCGA BRCA tumor samples with a cross-validation scheme, and datasets with known and unknown PAM50 classification. CL achieved prediction accuracy >73 %, highest among other methods we evaluated. We also applied the algorithm to a set of breast cancer tumors derived from Arabic population to assign a PAM50 classification to each tumor based on their gene expression profiles. CONCLUSIONS: A novel algorithm CrossLink for cross-condition prediction of cancer classes was proposed. In all test datasets, CL showed robust and consistent improvement in prediction performance over other state-of-the-art normalization and classification algorithms.