pith. sign in

arxiv: 1907.10418 · v1 · pith:YU7RAHRQnew · submitted 2019-07-23 · 📡 eess.IV · cs.LG· stat.ML

Improving Malaria Parasite Detection from Red Blood Cell using Deep Convolutional Neural Networks

Aimon Rahman , Hasib Zunair , M Sohel Rahman , Jesia Quader Yuki , Sabyasachi Biswas , Md Ashraful Alam , Nabila Binte Alam , M.R.C. Mahdy This is my paper

Pith reviewed 2026-05-24 17:24 UTC · model grok-4.3

classification 📡 eess.IV cs.LGstat.ML
keywords malaria detectionconvolutional neural networksdeep learningred blood cell smearsimage classificationNIH Malaria Datasetend-to-end learningmedical imaging
0
0 comments X

The pith

A deep convolutional neural network identifies malaria parasites in red blood cell patches with 97.77 percent accuracy through end-to-end learning.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes that a deep convolutional neural network can detect malaria parasites directly from raw patches of red blood cell smears. It replaces manual feature engineering with automatic extraction and classification in one model. The approach reaches nearly 97.77 percent accuracy on the NIH Malaria Dataset after testing multiple architectures and pre-processing steps. Five-fold cross-validation plus a holdout test support that the model generalizes to unseen patches. This matters because malaria diagnosis has long depended on time-consuming expert review of microscope slides.

Core claim

The authors show that training a deep convolutional neural network end-to-end on segmented red blood cell patches from the NIH Malaria Dataset produces a classifier that reaches an accuracy of almost 97.77 percent. The network performs feature extraction and classification without hand-crafted descriptors. Experiments compare several architectures, apply standard pre-processing, use five-fold cross-validation to select the best model, and confirm performance on a separate holdout set.

What carries the argument

Deep convolutional neural network that performs both feature extraction and classification directly from raw segmented patches of red blood smears.

If this is right

  • The method removes the need to design separate hand-engineered features for each new imaging setup.
  • High accuracy on cross-validated and holdout data indicates the model can process new patches without retraining from scratch.
  • Standard pre-processing steps can be added to the pipeline to raise performance further.
  • The same end-to-end structure could be applied to other microscopic blood cell classification tasks.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Portable microscopes paired with this model could support field diagnosis where trained microscopists are scarce.
  • Retraining the network on images from different microscope brands or staining protocols would test robustness across real deployments.
  • Combining the classifier output with patient metadata might improve overall diagnostic decisions beyond patch-level accuracy.

Load-bearing premise

The NIH dataset patches carry accurate labels that represent typical clinical samples without systematic errors or imaging artifacts.

What would settle it

Applying the trained model to a fresh collection of blood smear images labeled independently by multiple experts and observing accuracy well below 90 percent would falsify the performance claim.

Figures

Figures reproduced from arXiv: 1907.10418 by Aimon Rahman, Hasib Zunair, Jesia Quader Yuki, Md Ashraful Alam, M.R.C. Mahdy, M Sohel Rahman, Nabila Binte Alam, Sabyasachi Biswas.

Figure 01
Figure 01. Figure 01: Samples drawn from NIH Malaria dataset which are uninfected red blood cells. It is seen that the images have varying color distributions which are resulted from different stains during data acquisition [PITH_FULL_IMAGE:figures/full_fig_p006_01.png] view at source ↗
Figure 02
Figure 02. Figure 02: Samples drawn from NIH Malaria dataset which are malaria infected parasite red blood cells. The images show various forms of parasite in the red blood cells [PITH_FULL_IMAGE:figures/full_fig_p006_02.png] view at source ↗
Figure 03
Figure 03. Figure 03: Intensity distribution of an isolated cell before and after stain normalization. Each graph represents corresponding color channels (Red, Blue & Green). The y-axis is the number of pixels and x-axis is the range of pixel values (0-255). Stain normalization results to a narrow domain without changing the semantic meaning of the image [PITH_FULL_IMAGE:figures/full_fig_p009_03.png] view at source ↗
Figure 4
Figure 4. Figure 4 [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 04
Figure 04. Figure 04: Before and after stain normalization applied to red blood cell (RBC) patches. All RBCs have been transformed to have less color variations. This reduces the stain variation from training images while preserving the semantic meaning of images. From the figure it is evident that the semantic meaning is preserved after stain normalization is applied to the red blood cell patches. It is also seen that the sta… view at source ↗
Figure 05
Figure 05. Figure 05: Before and after standardization of RBC patch. Images have been rescaled to have mean 0 and standard deviation of 1. Earlier, each RBC image had pixel values between 0 to 255 [PITH_FULL_IMAGE:figures/full_fig_p011_05.png] view at source ↗
Figure 06
Figure 06. Figure 06: Augmented Images of a single malaria blood sample to increase data points. Each image has been flipped, rorated, translated, blurred, cropped etc. to have better representation of test set. The Network Architecture As has been alluded to in an earlier section, we have employed three different experimental settings. We have built a custom network from scratch (referred to as Custom henceforth). We also hav… view at source ↗
Figure 7
Figure 7. Figure 7: 19 layers custom architecture of Neural network with 2-way softmax activation. The input dimensions are 200 x 200 and ReLu is used as activation for convolution layers [PITH_FULL_IMAGE:figures/full_fig_p014_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: VGG-16 architecture as a baseline with 2-way softmax activation. The input dimensions are 200 x 200 and ReLU is used as activation for convolution layers. The adaptive learning rate method, named Adadelta [46], is used to optimize the weights and biases of the network with a starting learning rate of10−2 . Adadelta is a more robust extension of other optimizers such as Adagrad, that adapts learning rates b… view at source ↗
Figure 9
Figure 9. Figure 9: Ensemble of three models and their combined predictions. The three different deep learning models used in this experiment (Custom CNN, TL-VGG16 & CNN-Ex SVM) are combined to produce a better prediction for test set [PITH_FULL_IMAGE:figures/full_fig_p019_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Test Time Augmentation used as post processing. Images have been augmented several times and their weighted average predictions are considered [PITH_FULL_IMAGE:figures/full_fig_p020_10.png] view at source ↗
Figure 12
Figure 12. Figure 12: Example of mislabeled parasite infected blood cells confirmed by pathologist. Images shown here have a ground truth of normal but are evaluated as parasite by a pathologist. In the cases of false positives, it is seen that the model miss-classifies the red blood cells which are in an early stage of malaria - consisting of only a few parasites. Furthermore, we show from error analysis that the mislabeled d… view at source ↗
Figure 15
Figure 15. Figure 15: Training and validation logs of the TL-VGG16 model after running for 100 epochs. a) Accuracy plotted in every epoch throughout the training regimen. The network achieves a training accuracy of 98% and validation accuracy of 97.7%. b) Loss plotted in every epoch showing a maximum validation loss of 0.12. Discussion A. We have experimented with several preprocessing and post-processing techniques to analyze… view at source ↗
read the original abstract

Malaria is a female anopheles mosquito-bite inflicted life-threatening disease which is considered endemic in many parts of the world. This article focuses on improving malaria detection from patches segmented from microscopic images of red blood cell smears by introducing a deep convolutional neural network. Compared to the traditional methods that use tedious hand engineering feature extraction, the proposed method uses deep learning in an end-to-end arrangement that performs both feature extraction and classification directly from the raw segmented patches of the red blood smears. The dataset used in this study was taken from National Institute of Health named NIH Malaria Dataset. The evaluation metric accuracy and loss along with 5-fold cross validation was used to compare and select the best performing architecture. To maximize the performance, existing standard pre-processing techniques from the literature has also been experimented. In addition, several other complex architectures have been implemented and tested to pick the best performing model. A holdout test has also been conducted to verify how well the proposed model generalizes on unseen data. Our best model achieves an accuracy of almost 97.77%.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 1 minor

Summary. The manuscript proposes a deep convolutional neural network for end-to-end malaria parasite detection directly from raw segmented red blood cell patches in the NIH Malaria Dataset. It experiments with multiple architectures and pre-processing techniques, selects the best model via 5-fold cross-validation on accuracy and loss, conducts a holdout test, and reports a peak accuracy of 97.77%.

Significance. If the performance holds under leakage-free partitioning, the end-to-end CNN approach on a public dataset offers a reproducible baseline for automated malaria detection that avoids manual feature engineering. The multi-architecture comparison and pre-processing ablation are positive elements, but missing methodological details substantially limit the result's current utility and verifiability.

major comments (3)
  1. [Abstract] Abstract: The central claim of 97.77% holdout accuracy is presented without any description of the selected CNN architecture, its hyperparameters, the precise train/validation/test split ratios, or whether splits were performed at the patch level versus the patient/slide level. This information is load-bearing for assessing reproducibility and generalization.
  2. [Abstract] Abstract: The 5-fold cross-validation and holdout evaluation provide no indication that patches from the same blood smear or patient were kept together in a single fold or set. Because the NIH dataset contains multiple patches per slide/patient, patch-level random splitting risks data leakage via shared imaging or staining artifacts, directly undermining the generalization claim to unseen clinical samples.
  3. [Abstract] Abstract: No quantitative baseline results (e.g., against hand-crafted feature methods or prior CNNs on the same dataset), ablation studies, or error bars/confidence intervals are reported, so the improvement over traditional methods cannot be evaluated.
minor comments (1)
  1. [Abstract] Abstract: The phrasing 'female anopheles mosquito-bite inflicted' is awkward and could be clarified for precision.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment below and indicate where revisions will be made to improve clarity, reproducibility, and methodological transparency.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim of 97.77% holdout accuracy is presented without any description of the selected CNN architecture, its hyperparameters, the precise train/validation/test split ratios, or whether splits were performed at the patch level versus the patient/slide level. This information is load-bearing for assessing reproducibility and generalization.

    Authors: The abstract is intentionally concise as a high-level overview. The full manuscript details the tested architectures (VGG, ResNet, and custom CNN variants), hyperparameter selection via grid search, the 5-fold CV protocol, and an 80/10/10 patch-level split for the holdout test. To address the concern, we will revise the abstract to specify the best-performing architecture, key hyperparameters, and the patch-level split ratios. revision: yes

  2. Referee: [Abstract] Abstract: The 5-fold cross-validation and holdout evaluation provide no indication that patches from the same blood smear or patient were kept together in a single fold or set. Because the NIH dataset contains multiple patches per slide/patient, patch-level random splitting risks data leakage via shared imaging or staining artifacts, directly undermining the generalization claim to unseen clinical samples.

    Authors: This is a valid methodological concern. Our experiments used random patch-level splitting, consistent with the standard protocol and prior publications on the NIH Malaria Dataset (which is distributed as individual patches). We will revise the Methods section to explicitly describe the splitting procedure and add a discussion of this limitation, including its implications for generalization to new patients. revision: partial

  3. Referee: [Abstract] Abstract: No quantitative baseline results (e.g., against hand-crafted feature methods or prior CNNs on the same dataset), ablation studies, or error bars/confidence intervals are reported, so the improvement over traditional methods cannot be evaluated.

    Authors: The manuscript already performs internal comparisons across multiple architectures and pre-processing methods, serving as ablation on model and data choices. However, we did not include external baselines from hand-crafted features or prior CNN papers, nor report per-fold standard deviations. We will add a comparison table with published results on the same dataset and include error bars from the 5-fold CV in the revised manuscript. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical ML benchmark on external dataset

full rationale

The paper reports training and evaluating CNN architectures on the public NIH Malaria Dataset using standard 5-fold CV plus holdout testing, with accuracy as the metric. No equations, parameter fits, or derivations are present that reduce the reported 97.77% accuracy to an input by construction. No self-citation chains or ansatzes are invoked as load-bearing for any claim. The evaluation is self-contained against the external benchmark and does not exhibit any of the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The performance claim rests on the representativeness of the NIH dataset, the validity of standard CNN training, and the post-hoc selection of the best architecture and preprocessing steps from experiments.

free parameters (2)
  • CNN architecture and hyperparameters
    Multiple complex architectures were implemented and the best selected based on validation accuracy and loss.
  • Pre-processing techniques
    Existing standard pre-processing methods were experimented with to maximize performance.
axioms (2)
  • domain assumption The NIH Malaria Dataset consists of accurately labeled segmented patches suitable for supervised training and evaluation.
    All reported metrics depend on this dataset without independent verification of labeling quality or domain coverage.
  • domain assumption Deep CNNs trained end-to-end via backpropagation can extract discriminative features from raw image patches for this classification task.
    Core premise of the deep learning approach stated in the abstract.

pith-pipeline@v0.9.0 · 5754 in / 1442 out tokens · 44468 ms · 2026-05-24T17:24:28.533643+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

59 extracted references · 59 canonical work pages · 5 internal anchors

  1. [1]

    Malaria: fact sheet

    World Health Organization. Malaria: fact sheet. No. WHO -EM/MAC/035/E. World Health Organization. Regional Office for the Eastern Mediterranean, 2014

    work page 2014

  2. [2]

    Emergency department management of mosquito- borne illness: malaria, dengue, and West Nile virus

    Caraballo, Hector, and Kevin King. "Emergency department management of mosquito- borne illness: malaria, dengue, and West Nile virus." Emergency medicine practice 16, no. 5 (2014): 1-23

    work page 2014

  3. [3]

    Malaria burden and control in Bangladesh and prospects for elimination: an epidemiological and economic assessment

    Haque, Ubydul, Hans J. Overgaard, Archie CA Clements, Dou glas E. Norris, Nazrul Islam, Jahirul Karim, Shyamal Roy et al. "Malaria burden and control in Bangladesh and prospects for elimination: an epidemiological and economic assessment." The Lancet Global Health 2, no. 2 (2014): e98-e105

    work page 2014

  4. [4]

    Malaria prevalence in endemic districts of Bangladesh

    Haque, Ubydul, Syed Masud Ahmed, Shahed Hossain, Mamun Huda, Awlad Hossain, Mohammad Shafiul Alam, Dinesh Mondal, Wasif Ali Khan, Mohammod Khalequzzaman, and Rashidul Haque. "Malaria prevalence in endemic districts of Bangladesh." PLoS One 4, no. 8 (2009): e6737

    work page 2009

  5. [5]

    Barber, Zulfiqar A

    Wang, Haidong, Mohsen Naghavi, Christine Allen, Ryan M. Barber, Zulfiqar A. Bhutta, Austin Carter, Daniel C. Casey et al. "Global, regional, and national life expectancy, all - cause mortality, and cause -specific mortality for 249 causes of death, 1980 –2015: a systematic analysis for the Global Burden of Disease Study 2015." The lancet 388, no. 10053 (2...

    work page 1980

  6. [6]

    Malaria rapid diagnostic tests

    Wilson, Michael L. "Malaria rapid diagnostic tests." Clinical infectious diseases 54, no. 11 (2012): 1637-1641

    work page 2012

  7. [7]

    Malaria: An update for physicians

    Nadjm, Behzad, and Ron H. Behrens. "Malaria: An update for physicians." Infectious Disease Clinics 26, no. 2 (2012): 243-259

    work page 2012

  8. [8]

    Malaria: Disease impacts and long -run income differences

    Gollin, Douglas, and Christian Zimmermann. "Malaria: Disease impacts and long -run income differences." (2007). 29

    work page 2007

  9. [9]

    Shortage of pathologists inhibits progress on UHC.The Daily Star, Retrieved from URL

    Star Health Desk, (2018, March 18 Published). Shortage of pathologists inhibits progress on UHC.The Daily Star, Retrieved from URL. https://www.thedailystar.net/health/shortage-pathologists-inhibits-progress-uhc-1549609

    work page 2018

  10. [10]

    Chaity, A. Z. (2017, December 13 Published). Bangladeshis flock to Indian, Thai hospitals in huge numbers. Dhaka Tribune, Retrieved from URL. https://www.dhakatribune.com/feature/health-wellness/2017/11/30/doctor-trust- bangladesh

    work page 2017

  11. [11]

    A survey on deep learning in medical image analysis

    Litjens, Geert, Thijs Kooi, Babak Ehteshami Bejnordi, Arnaud Arindra Adiyoso Setio, Francesco Ciompi, Mohsen Ghafoorian, Jeroen Awm Van Der Laak, Bram Van Ginneken, and Clara I. Sánchez. "A survey on deep learning in medical image analysis." Medical image analysis 42 (2017): 60-88

    work page 2017

  12. [12]

    CNN -based image analysis for malaria diagnosis

    Liang, Zhaohui, Andrew Powell, Ilker Ersoy, Mahdieh Poostchi, Kamolrat Silamut, Kannappan Palaniappan, Peng Guo et al. "CNN -based image analysis for malaria diagnosis." In 2016 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pp. 493-496. IEEE, 2016

    work page 2016

  13. [13]

    Convolutional deep belief ne tworks on cifar - 10

    Krizhevsky, Alex, and Geoff Hinton. "Convolutional deep belief ne tworks on cifar - 10." Unpublished manuscript 40, no. 7 (2010)

    work page 2010

  14. [14]

    Evaluations of deep convolutional neural networks fo r automatic identification of malaria infected cells

    Dong, Yuhang, Zhuocheng Jiang, Hongda Shen, W. David Pan, Lance A. Williams, Vishnu VB Reddy, William H. Benjamin, and Allen W. Bryan. "Evaluations of deep convolutional neural networks fo r automatic identification of malaria infected cells." In 2017 IEEE EMBS International Conference on Biomedical & Health Informatics (BHI), pp. 101-104. IEEE, 2017

    work page 2017

  15. [15]

    LeNet -5, convolutional neural networks

    LeCun, Yann. "LeNet -5, convolutional neural networks." URL: http://yann. lecun. com/exdb/lenet 20 (2015)

    work page 2015

  16. [16]

    Imagenet classification with deep convolutional neural networks

    Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E. Hinton. "Imagenet classification with deep convolutional neural networks." In Advances in neural information processing systems, pp. 1097-1105. 2012

    work page 2012

  17. [17]

    Going deeper with convolutions

    Szegedy, Christian, Wei Li u, Yangqing Jia, Pierre Sermanet, Scott Reed, Dragomir Anguelov, Dumitru Erhan, Vincent Vanhoucke, and Andrew Rabinovich. "Going deeper with convolutions." In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 1-9. 2015

    work page 2015

  18. [18]

    Applying faster R -CNN for object detection on malaria images

    Hung, Jane, and Anne Carpenter. "Applying faster R -CNN for object detection on malaria images." In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, pp. 56-61. 2017. 30

    work page 2017

  19. [19]

    Imagenet: A large-scale hierarchical image database

    Deng, Jia, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. "Imagenet: A large-scale hierarchical image database." In 2009 IEEE conference on computer vision and pattern recognition, pp. 248-255. Ieee, 2009

    work page 2009

  20. [20]

    Malaria parasite detect ion from peripheral blood smear images using deep belief networks

    Bibin, Dhanya, Madhu S. Nair, and P. Punitha. "Malaria parasite detect ion from peripheral blood smear images using deep belief networks." IEEE Access 5 (2017): 9099- 9108

    work page 2017

  21. [21]

    Convolutional deep belief networks for scalable unsupervised learning of hierarchical representations

    Lee, Honglak, Roger Grosse, Rajesh Ranganath, and Andrew Y. Ng. "Convolutional deep belief networks for scalable unsupervised learning of hierarchical representations." In Proceedings of the 26th annual international conference on machine learning, pp. 609-616. ACM, 2009

    work page 2009

  22. [22]

    Deep boltzmann machines

    Salakhutdinov, Ruslan, and Geoffrey Hinton. "Deep boltzmann machines." In Artificial intelligence and statistics, pp. 448-455. 2009

    work page 2009

  23. [23]

    On contrastive divergence learning

    Carreira -Perpinan, Miguel A., and Geoffrey E. Hinton. "On contrastive divergence learning." In Aistats, vol. 10, pp. 33-40. 2005

    work page 2005

  24. [24]

    Microscopic blood smear segmentation and classification using deep con tour aware cnn and extreme machine learning

    Razzak, Muhammad Imran, and Saeeda Naz. "Microscopic blood smear segmentation and classification using deep con tour aware cnn and extreme machine learning." In 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), pp. 801-807. IEEE, 2017

    work page 2017

  25. [25]

    Contextlocnet: Context-aware deep network models for weakly supervised localization

    Kantorov, Vadim, Maxime Oquab, Minsu Cho, and Ivan Laptev. "Contextlocnet: Context-aware deep network models for weakly supervised localization." In European Conference on Computer Vision, pp. 350-365. Springer, Cham, 2016

    work page 2016

  26. [26]

    Extreme learning machine: theory and applications

    Huang, Guang-Bin, Qin-Yu Zhu, and Chee-Kheong Siew. "Extreme learning machine: theory and applications." Neurocomputing 70, no. 1-3 (2006): 489-501

    work page 2006

  27. [27]

    Computer -automated malaria diagnosis and quantitation using convolutional neural networks

    Mehanian, Courosh, Mayoore Jaiswal, Charles Delahunt, Clay Thompson, Matt Horning, Liming Hu, Travis Ostbye et al. "Computer -automated malaria diagnosis and quantitation using convolutional neural networks." In Proceeding s of the IEEE International Conference on Computer Vision, pp. 116-125. 2017

    work page 2017

  28. [28]

    Malaria Parasite Detection with Deep Transfer Learning

    Var, Esra, and F. Boray Tek. "Malaria Parasite Detection with Deep Transfer Learning." In 2018 3rd International Conference on Computer Science and Engineering (UBMK), pp. 298-302. IEEE, 2018

    work page 2018

  29. [29]

    Pre - trained convolutional neural networks as feature extractors toward improv ed malaria parasite detection in thin blood smear images

    Rajaraman, Sivaramakrishnan, Sameer K. Antani, Mahdieh Poostchi, Kamolrat Silamut, Md A. Hossain, Richard J. Maude, Stefan Jaeger, and George R. Thoma. "Pre - trained convolutional neural networks as feature extractors toward improv ed malaria parasite detection in thin blood smear images." PeerJ 6 (2018): e4568. 31

    work page 2018

  30. [30]

    Lossless compression of curated erythrocyte images using deep autoencoders for malaria infection diagnosis

    Shen, Hongda, W. David Pan, Yuhang Dong, and Mohammad Alim. "Lossless compression of curated erythrocyte images using deep autoencoders for malaria infection diagnosis." In 2016 Picture Coding Symposium (PCS), pp. 1-5. IEEE, 2016

    work page 2016

  31. [31]

    Automatic Detection of Malaria Parasites Using Unsupervised Techniques

    Mohanty, Itishree, P. A. Pattanaik, and Tripti Swarnkar. "Automatic Detection of Malaria Parasites Using Unsupervised Techniques." In International Conference on ISMAC in Computational Vision and Bio-Engineering, pp. 41-49. Springer, Cham, 2018

    work page 2018

  32. [32]

    Machine learning approach for automated screening of malaria parasite using light microscopic images

    Das, Dev Kumar, Madhumala Ghosh, Mallika Pal, Asok K. Maiti, and Chandan Chakraborty. "Machine learning approach for automated screening of malaria parasite using light microscopic images." Micron 45 (2013): 97-106

    work page 2013

  33. [33]

    Automated detec tion of P. falciparum using machine learning algorithms with quantitative phase images of unstained cells

    Park, Han Sang, Matthew T. Rinehart, Katelyn A. Walzer, Jen -Tsan Ashley Chi, and Adam Wax. "Automated detec tion of P. falciparum using machine learning algorithms with quantitative phase images of unstained cells." PloS one 11, no. 9 (2016): e0163045

    work page 2016

  34. [34]

    Comparison of Detection Method on Malaria Cell Images

    Mustafa, Wan Azani, Ragunathan Santiagoo, Irfan Jamaluddin, Nurul S. Othman, Wan Khairunizam, and M. N. K. H. Rohani. "Comparison of Detection Method on Malaria Cell Images." In 2018 International Conference on Computational Approach in Smart Systems Design and Applications (ICASSDA), pp. 1-6. IEEE, 2018

    work page 2018

  35. [35]

    FCM: The fuzzy c -means clustering algorithm

    Bezdek, James C., Robert Ehrlich, and William Fu ll. "FCM: The fuzzy c -means clustering algorithm." Computers & Geosciences 10, no. 2-3 (1984): 191-203

    work page 1984

  36. [36]

    Periodicity, directionality, and randomness: Wold features for image modeling and retrieval

    Liu, Fang, and Rosalind W. Picard. "Periodicity, directionality, and randomness: Wold features for image modeling and retrieval." IEEE transactions on pattern analysis and machine intelligence 18, no. 7 (1996): 722-733

    work page 1996

  37. [37]

    Method for determining stimulation parameters

    Bradley, Kerry. "Method for determining stimulation parameters." U.S. Patent 7,174,215, issued February 6, 2007

    work page 2007

  38. [38]

    Adaptation of Brensen s Thresholding Algorithm for Sketched Line Drawings

    Bartolo, Alexandra, Kenneth P. Camilleri, Philip J. Farrugia , and Jonathan C. Borg. "Adaptation of Brensen s Thresholding Algorithm for Sketched Line Drawings." In SBM, pp. 81-90. 2004

    work page 2004

  39. [39]

    Blood film preparation and staining procedures

    Houwen, Berend. "Blood film preparation and staining procedures." Clinics in laboratory medicine 22, no. 1 (2002): 1-14

    work page 2002

  40. [40]

    The Wright -Giemsa stain: secrets revealed

    Woronzoff-Dashkoff, Kristine Krafts. "The Wright -Giemsa stain: secrets revealed." Clinics in laboratory medicine 22, no. 1 (2002): 15-23

    work page 2002

  41. [41]

    A nonlinear mapping approach to stain nor malization in digital histopathology images using image - specific color deconvolution

    Khan, Adnan Mujahid, Nasir Rajpoot, Darren Treanor, and Derek Magee. "A nonlinear mapping approach to stain nor malization in digital histopathology images using image - specific color deconvolution." IEEE Transactions on Biomedical Engineering 61, no. 6 (2014): 1729-1738. 32

    work page 2014

  42. [42]

    The importance of stain normalization in colorectal tissue classification with convolutional networks

    Ciompi, Francesco, Oscar Geessink, Babak Ehteshami Bejnordi, Gabriel Silva De Souza, Alexi Baidoshvili, Geert Litjens, Bram Van Ginneken, Iris Nagtegaal, and Jeroen Van Der Laak. "The importance of stain normalization in colorectal tissue classification with convolutional networks." In 2017 IEEE 14th International Symposium on Biomedical Imaging (ISBI 201...

    work page 2017

  43. [43]

    Classification of breast cancer histology images using convolutional neural networks

    Araújo, Teresa, Guilherme Aresta, Eduardo Castro, José Rouco, Paulo Aguiar, Catarina Eloy, António Polónia, and Aurélio Campilho. "Classification of breast cancer histology images using convolutional neural networks." PloS one 12, no. 6 (2017): e0177544

    work page 2017

  44. [44]

    Synthetic data augmentation using GAN for improved liver lesion classification

    Frid -Adar, Maayan, Eyal Klang, Michal Amitai, Jacob Goldberger, and Hayit Greenspan. "Synthetic data augmentation using GAN for improved liver lesion classification." In 2018 IEEE 15th International Symposium on Biomedical Imaging (ISBI 2018), pp. 289-293. IEEE, 2018

    work page 2018

  45. [45]

    Very Deep Convolutional Networks for Large-Scale Image Recognition

    Simonyan, Karen, and Andrew Zisserman. "Very deep convolutional networks for large-scale image recognition." arXiv preprint arXiv:1409.1556 (2014)

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  46. [46]

    ADADELTA: An Adaptive Learning Rate Method

    Zeiler, Matthew D. "ADADELTA: an adaptive learning rate method." arXiv preprint arXiv:1212.5701 (2012)

    work page internal anchor Pith review Pith/arXiv arXiv 2012

  47. [47]

    Understanding the difficulty of training deep feedforward neural networks

    Glorot, Xavier, and Yoshua Bengio. "Understanding the difficulty of training deep feedforward neural networks." In Proceedings of the thirteenth internat ional conference on artificial intelligence and statistics, pp. 249-256. 2010

    work page 2010

  48. [48]

    Dropout: a simple way to prevent neural networks from overfitting

    Srivastava, Nitish, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. "Dropout: a simple way to prevent neural networks from overfitting." The Journal of Machine Learning Research 15, no. 1 (2014): 1929-1958

    work page 2014

  49. [49]

    Cross entropy error function in neural networks: Forecasting gasoline demand

    Nasr, George E., E. A. Badr, and C. Joun. "Cross entropy error function in neural networks: Forecasting gasoline demand." In FLAIRS conference, pp. 381-384. 2002

    work page 2002

  50. [50]

    Unconventional Wisdom: A New Transfer Learning Approach Applied to Bengali Numeral Classification

    36. Zunair, Hasib, N abeel Mohammed, and Sifat Momen. "Unconventional Wisdom: A New Transfer Learning Approach Applied to Bengali Numeral Classification." In 2018 International Conference on Bangla Speech and Language Processing (ICBSLP), pp. 1-6. IEEE, 2018

    work page 2018

  51. [51]

    Support -vector networks

    42. Cortes, Corinna, and Vladimir Vapnik. "Support -vector networks." Machine learning 20, no. 3 (1995): 273-297

    work page 1995

  52. [52]

    A novel hybrid CNN –SVM classifier for recognizing handwritten digits

    Niu, Xiao -Xiao, and Ching Y. Suen. "A novel hybrid CNN –SVM classifier for recognizing handwritten digits." Pattern Recognition 45, no. 4 (2012): 1318-1325

    work page 2012

  53. [53]

    An Architecture Combining Convolutional Neural Network (CNN) and Support Vector Machine (SVM) for Image Classification

    Agarap, Abien Fred. "An architecture combining convolutional neural network (CNN) and support vector machine (SVM) for image classification." arXiv preprint arXiv:1712.03541 (2017). 33

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  54. [54]

    Deep convolutional neural networks and support vector machines for gender recognition

    van de Wolfshaar, Jos, Mahir F. Karaaba, and Marco A. Wiering. "Deep convolutional neural networks and support vector machines for gender recognition." In 2015 IEEE Symposium Series on Computational Intelligence, pp. 188-195. IEEE, 2015

    work page 2015

  55. [55]

    Senti ment classification: The contribution of ensemble learning

    Wang, Gang, Jianshan Sun, Jian Ma, Kaiquan Xu, and Jibao Gu. "Senti ment classification: The contribution of ensemble learning." Decision support systems 57 (2014): 77-93

    work page 2014

  56. [56]

    Ensemble machine learning on gene expression data for cancer classification

    Tan, Aik Choon, and David Gilbert. "Ensemble machine learning on gene expression data for cancer classification." (2003)

    work page 2003

  57. [57]

    Ensemble learning

    Polikar, Robi. "Ensemble learning." In Ensemble machine learning, pp. 1-34. Springer, Boston, MA, 2012

    work page 2012

  58. [58]

    Snapshot Ensembles: Train 1, get M for free

    Huang, Gao, Yixuan Li, Geoff Pleiss, Zhuang Liu, John E. Hopcroft, and Kilian Q. Weinberger. "Snapshot ensembles: Train 1, get m for free." arXiv preprint arXiv:1704.00109 (2017)

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  59. [59]

    The Effectiveness of Data Augmentation in Image Classification using Deep Learning

    Perez, Luis, and Jason Wang. "The effectiveness of data augmentation in image classification using deep learning." arXiv preprint arXiv:1712.04621 (2017)

    work page internal anchor Pith review Pith/arXiv arXiv 2017