Pith. sign in

REVIEW 4 major objections 5 minor 30 references

CUDA optimized Neural Network predicts blood glucose control from quantified joint mobility and anthropometrics

T0 review · 4 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper claims that non-invasive finger-joint mobility and anthropometric readings can separate good from poor glucose control in men, with a CUDA-trained network reaching 86.67% test accuracy, and that the same network trains 50 times…

desk verdict A genuinely new small clinical dataset and a plausible-but-unproven male screening signal; statistically too fragile to defend as stated. read the letter →

arxiv 1908.07847 v1 pith:PIRU5YHS submitted 2019-08-19 cs.LG q-bio.QMstat.ML

Sterling Ramroach , Andrew Dhanoo , Brian Cockburn , Ajay Joshi This is my paper · ORCID
classification cs.LGq-bio.QMstat.ML
keywords NeuralNetworkCUDAParallelProcessingLimitedJointMobilityDiabetesMellitusHbA1cpredictionNon-invasivescreeningBackpropagation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that limited joint mobility of the fingers, measured with a goniometer and combined with ordinary body measurements, can separate good from poor glucose control using non-invasive readings. The evidence comes from a single-hidden-layer backpropagation network trained on 33 features from 120 participants at diabetes awareness drives, with men and women modeled separately. On the male subset the network reached 86.67% accuracy on held-out test cases, while the female model overfit and fell to 66.67%. The paper also shows that moving the same training computation to a GPU with CUDA makes it about 50 times faster than the CPU version. If the male result holds, a cheap screen for diabetes control could be built from joint angles and a tape measure.

What carries the argument

The mechanism that carries the argument has two parts. The clinical part is quantified limited joint mobility: with palms pressed together and wrists maximally flexed, a finger goniometer records the angle at the MCP, PIP, and DIP joints of each finger, and the IP joint of the thumb, where negative values mean hyper-extension. The computational part is a multilayer perceptron with one hidden layer, sigmoid activation, gradient-descent backpropagation, learning rate 0.1 and no momentum, parallelized with CUDA in a topological data-parallel scheme: one copy of the network lives on the GPU, each neuron runs on its own thread, and inputs and weights are stored as flat float arrays. Those joint angles are the non-invasive signal the network learns from, and the CUDA scheme is what makes 100,000 epochs of training practical.

What would settle it

Run the same network on the male subset under repeated random 75/25 splits or leave-one-out cross-validation and inspect the distribution of test accuracy: if the 86.67% figure is not reproduced or collapses toward chance on many splits, the non-invasive male-screening claim is not supported by this dataset.

Watch

Extended reading notes

Core claim

The central claim, stated in the conclusion, is that a differentiation of good from poor glucose control can be made on males via non-invasive readings that include limited joint mobility. The supporting result is a feed-forward neural network with one hidden layer, trained by backpropagation on 33 features collected from finger-joint goniometry, demographics, and anthropometrics; at 100,000 epochs it classified the male test set with 86.67% accuracy against 95.65% on training data. The same architecture on the female subset reached 97.73% training accuracy but only 66.67% test accuracy, which the authors attribute to overfitting and read as no supporting evidence for females. As a secondary result, the CUDA-parallelized network trained about 50 times faster on the GPU than the corresponding CPU implementation.

Load-bearing premise

The male screening result depends on the assumption that one 75/25 split of 61 participants is representative; with only about 15 male test cases, a different split could change the 86.67% accuracy materially.

Editorial extensions

If this is right

  • A validated version of the male model would give diabetes outreach programs a first-line screening step that needs only a goniometer, a tape measure, and a trained network.
  • Modeling men and women separately, as done here, is a workable design choice, but it leaves female screening unsupported until a model that generalizes is found.
  • Because the GPU version trains the same network 50 times faster, larger cohorts and longer training runs become affordable without changing the model.
  • The method can be deployed wherever HbA1c testing is expensive, since the predictive inputs are already easy to record at awareness drives.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The 86.67% male test accuracy comes from one split of about 15 test cases; repeated random splits or cross-validation would likely show a wide range, and the stability of that figure is untested in the paper.
  • The female failure could stem from sample size or class imbalance rather than a true absence of signal; a larger female cohort and stratified training might recover a usable model.
  • A graded outcome (predicted HbA1c range or risk category) would be more clinically useful than the binary good/poor split and would test whether joint angles track glycemic control continuously.
  • The claimed speedup is measured at one small fixed dataset; whether the advantage grows with dataset size is a prediction the paper makes but does not measure.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper presents a CUDA-accelerated feed-forward neural network with one hidden layer, trained to classify male and female participants into good versus poor glycemic control based on HbA1c from non-invasive markers, including limited joint mobility (LJM) and anthropometrics. On a single 75/25 split of data from 120 participants (61 male, 59 female), the authors report male test accuracy of 86.67% in the abstract and 91.7% in the results section, female test accuracy of 66.67%, and a 50x training speedup from the CUDA implementation. The authors conclude that the results support using LJM-based readings for screening male participants only.

Significance. If the central claim held, the finding that finger-joint mobility combined with anthropometrics can non-invasively screen for poor glycemic control would be practically useful in resource-limited settings. The paper also provides a concrete CUDA implementation with a claimed 50x training speedup over a CPU version. However, the statistical evaluation of the central claim is thin: the male screening conclusion rests on a single small test split with no confidence intervals, no cross-validation, no baseline comparison, and an internal inconsistency in the reported accuracy. These gaps directly affect whether the stated screening claim is supported, so the paper's main contribution is currently a promising but unverified proof-of-concept.

major comments (4)
  1. [§2.2 and Figures 3–4] The male screening claim rests on a single 75/25 split that yields only about 15 male test cases. No repeated splits, cross-validation, confidence intervals, or significance tests are reported. With n≈15, the standard error of a 86.7% accuracy is roughly 8.8 percentage points, so a different split could plausibly produce a materially different accuracy. The authors should report cross-validated performance or, at minimum, the distribution of test accuracy over multiple random splits.
  2. [Abstract vs. §3, first paragraph] The abstract reports a male test accuracy of 86.67%, while the results section states that at 100,000 epochs the male testing set accuracy was 91.7%. The manuscript does not explain which number is final, what epoch count or split seed produced the abstract value, or how the discrepancy arises. The exact training configuration and the reconciled test accuracy must be stated clearly.
  3. [§2.2/§3, class balance and baselines] No class balance or confusion matrix is given for the binarized HbA1c target on either the male or female subset. The paper notes that LJM was observed in 86.6% of all participants (§3, first paragraph), which indicates that high base rates are plausible in this clinic population. If poor glucose control is the majority class in the male test set, a trivial majority-class classifier could already attain accuracy close to the reported 86.7%. The authors should report the class distribution and compare against a majority-class or other simple baseline.
  4. [§2.1/§3, feature contribution] The network is trained on 33 features that include the binary 'onMed' medication flag. Because the target, HbA1c, is directly related to diabetes medication, the network could achieve high accuracy by relying on this flag rather than on LJM and anthropometric readings. Without a feature-ablation study or a baseline trained on non-medication features alone, the conclusion that non-invasive readings which include limited joint mobility readings drive the classification is unsupported.
minor comments (5)
  1. [§2.1, data availability] The manuscript references 'Supplementary Table 1' in §2.1, but the arXiv submission does not include the table; the data should be made available or explicitly cited as available elsewhere.
  2. [Figures 3–6] The text descriptions of Figures 3–6 do not state whether axis labels and legends are included; the authors should ensure the figures clearly distinguish training and test accuracy and label the CPU/GPU curves in Figures 5 and 6.
  3. [Abstract and throughout] The manuscript contains typographical spacing errors and similar artifacts (e.g., 'pro cessing' in the abstract) that should be corrected through a careful proofread.
  4. [§2.2, parallelization description] The 'topological data parallel approach' is not defined; please clarify how neurons are topologically mapped to threads and blocks, and how the weight updates are synchronized during backpropagation.
  5. [§3, speedup claim] The 50x speedup claim should state what is measured (wall-clock time for a fixed number of epochs, including host-device data transfer) and whether the CPU implementation is itself optimized, so that the comparison is fair and reproducible.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the held-out neural network test accuracy is a generalization measurement, not a quantity derived from its own inputs.

full rationale

The paper's central claim is that a CUDA-accelerated neural network can differentiate good from poor glucose control in males using non-invasive readings, with test accuracy reported on a held-out 25% of the male dataset. This is a supervised generalization measurement: the network is trained on non-invasive features (finger joint angles, anthropometrics, demographics, and the onMed flag) to predict HbA1c-derived labels, and the reported test accuracy is computed on cases not used for training. The target is not included among the predictors, no parameter is fitted to the test set and then renamed as a prediction, and no equation in the paper reduces the prediction to its input by construction. The only self-citation is reference [2] (Sooknanan and Joshi), used as background for CUDA-based GPU computing; it is not load-bearing for the clinical screening claim. The valid concerns about this paper—a single 75/25 split with roughly 15 male test cases, no class-balance or confusion-matrix reporting, no baseline comparison, and the inconsistency between the abstract's 86.67% and the results section's 91.7% male test accuracy—are statistical-evidence and reporting issues, not circularity. They do not make the derivation self-referential, so the appropriate circularity score is 0.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central claim rests on a small single-split dataset and hand-chosen hyperparameters. No new particles, forces, or entities are introduced, so the invented-entity burden is zero.

free parameters (4)
  • learning_rate = 0.1
    Set to 0.1 with no momentum; affects convergence and final accuracy, no search reported.
  • number_of_epochs = 100,000
    Accuracy is reported at this epoch count; no early stopping or epoch tuning is described.
  • hidden_layer_size = not reported
    The one hidden layer's neuron count is never stated; it must be chosen by hand and strongly affects overfitting.
  • train_test_split = 75/25, seed not reported
    A single random split determines the test accuracy; no repeated splits or cross-validation are performed.
assumptions (4)
  • standard math Backpropagation with sigmoid activation and learning rate 0.1 converges to a useful local minimum for this task.
    Section 2.2 assumes gradient descent training works as intended without discussing convergence guarantees.
  • domain assumption Finger goniometer measurements and point-of-care HbA1c readings are accurate and reliable.
    Section 2.1 treats these measurements as ground truth for training and evaluation.
  • domain assumption The 120 participants at diabetes awareness drives represent the population to be screened.
    Section 2.1 includes all attendees without exclusion criteria, so selection bias is possible.
  • domain assumption The HbA1c threshold for poor glucose control is fixed and clinically meaningful.
    Section 1 cites HbA1c greater than 6.5% as the diagnosis threshold, and this is used as the classification target.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CUDA optimized Neural Network predicts blood glucose control from quantified joint mobility and anthropometrics." pith.science (2026). https://pith.science/paper/PIRU5YHS

@misc{pith2026190807847,
  author       = {Pith},
  title        = {Pith review of: CUDA optimized Neural Network predicts blood glucose control from quantified joint mobility and anthropometrics},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PIRU5YHS}},
  note         = {Machine review of arXiv:1908.07847}
}
read the original abstract

Neural network training entails heavy computation with obvious bottlenecks. The Compute Unified Device Architecture (CUDA) programming model allows us to accelerate computation by passing the processing workload from the CPU to the graphics processing unit (GPU). In this paper, we leveraged the power of Nvidia GPUs to parallelize all of the computation involved in training, to accelerate a backpropagation feed-forward neural network with one hidden layer using CUDA and C++. This optimized neural network was tasked with predicting the level of glycated hemoglobin (HbA1c) from non-invasive markers. The rate of increase in the prevalence of Diabetes Mellitus has resulted in an urgent need for early detection and accurate diagnosis. However, due to the invasiveness and limitations of conventional tests, alternate means are being considered. Limited Joint Mobility (LJM) has been reported as an indicator for poor glycemic control. LJM of the fingers is quantified and its link to HbA1c is investigated along with other potential non-invasive markers of HbA1c. We collected readings of 33 potential markers from 120 participants at a clinic in south Trinidad. Our neural network achieved 95.65% accuracy on the training and 86.67% accuracy on the testing set for male participants and 97.73% and 66.67% accuracy on the training and testing sets for female participants. Using 960 CUDA cores from a Nvidia GeForce GTX 660, our parallelized neural network was trained 50 times faster on both subsets, than its corresponding CPU implementation on an Intel Core (TM) i7-3630QM 2.40 GHz CPU.

Figures

Figures reproduced from arXiv: 1908.07847 by the authors.

Figure 4
Figure 4. Accuracy of the neural network predictions on the [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 30 canonical work pages

  1. [1]

    INTRODUCTION An Artificial Neural Network (NN) is a model that was inspired by the biological neural network from the brains of animals [1]. The NN is made up of various layers (input, hidden, and output) of neurons where each neuron performs a calculation using inputs from the data source or the outputs of earlier neurons. Neurons are connected by weight...

    work page

  2. [2]

    All attendees w ere included in the study

    Methods 2.1 Data Collection Data an d samples were collected from 12 0 participants at diabetes awareness drives at a p ublic primary care health center in Trinidad and Tobago. All attendees w ere included in the study. Ethics approval was granted by the campus ethics committee of the University of the West Indies, Faculty of Medical Sciences, St. Augusti...

    work page

  3. [3]

    LJM may act as an independent marker for HbA1c levels

    Results and Discussion The relationship between musculoskeletal complications and HbA1c is still being explored. LJM may act as an independent marker for HbA1c levels. The investigation of this hypothesis is hindered by many factors, including : inaccurate reporting of time since diagnosis due to late detection, and the size of the dataset . Some studies ...

    work page

  4. [4]

    Our neural network was unable to successful ly classify the female subset of participants

    Conclusion These results indicate that a differentiation of good from poor glucose control can be made on males via non-invasive readings which include limited joint mobility readings. Our neural network was unable to successful ly classify the female subset of participants. This can be due to the size of the dataset. Further research is needed to determi...

    work page

  5. [5]

    Data-based parallelism distributes large datasets over different nodes

    and model-based [6]. Data-based parallelism distributes large datasets over different nodes. Model-based parallelism splits complex models across different nodes. Learning in these networks must account for coordination and communication among processors. Figure 1. CPU vs GPU architecture and design philosophy [7]. The CUDA programming model facilitates t...

    work page 2025

  6. [6]

    Divya, U. V. and Prasad, P. S. Hashing Supported Iterative MapReduce Based Scalable SBE Reduct Computation. Springer, City, 2018

    work page 2018

  7. [7]

    Special thanks to Mr

    ACKNOWLEDGMENTS Our thanks to the Diabetes Association of Trinidad and Tobago for organizing the clinics for data collection. Special thanks to Mr. Craig Subnarine for his scholarly advice

    work page

  8. [8]

    K., Mao, J

    Jain, A. K., Mao, J. and Mohiuddin, K. M. Artificial neural networks: A tutorial. Computer, 29, 3 (1996), 31-44

    work page 1996

Show all 30 references
  1. [9]

    Sooknanan, D. J. and Joshi, A. GPU computing using CUDA in the deployment of smart grids. IEEE, City, 2016

    2016

  2. [10]

    and Chu, X

    Mei, X. and Chu, X. Dissecting GPU memory hierarchy through microbenchmarking. IEEE Transactions on Parallel and Distributed Systems, 28, 1 (2017), 72-86

    2017

  3. [11]

    and Kandrot, E

    Sanders, J. and Kandrot, E. CUDA by example: an introduction to general-purpose GPU programming. Addison-Wesley Professional, 2010

    2010

  4. [12]

    and correspondingly cannot be used with enough accuracy for the early detection of diabetes complications [13]. HbA1c as well as most other common methods of diabetes testing requires a blood sample to be collected, which may present a barrier to some patients as the willingne...

  5. [13]

    and Richmond, P

    Heywood, P., Maddock, S., Casas, J., Garcia, D., Brackstone, M. and Richmond, P. Data-parallel agent-based microscopic road network simulation using graphics processing units. Simulation Modelling Practice and Theory, 83 (2018), 188- 200

    2018

  6. [14]

    and Pottier, B

    Lounis, M., Bounceur, A., Laga, A. and Pottier, B. GPU- based parallel computing of energy consumption in wireless sensor networks. IEEE, City, 2015

    2015

  7. [15]

    Nvidia cuda c programming guide

    Nvidia, C. Nvidia cuda c programming guide. Nvidia Corporation, 120, 18 (2011), 8

    2011

  8. [16]

    and Denny, J

    Mani, S., Chen, Y., Elasy, T., Clayton, W. and Denny, J. Type 2 diabetes risk forecasting from EMR data using machine learning. American Medical Informatics Association, City, 2012

    2012

  9. [17]

    Inzucchi, S. E. and Sherwin, R. S. The prevention of type 2 diabetes mellitus. Endocrinol Metab Clin North Am, 34, 1 (Mar 2005), 199-219, viii

    2005

  10. [18]

    Standards of Medical Care in Diabetes-2018 Abridged for Primary Care Providers

    American Diabetes, A. Standards of Medical Care in Diabetes-2018 Abridged for Primary Care Providers. Clin Diabetes, 36, 1 (Jan 2018), 14-37

    2018

  11. [19]

    and Soda, P

    Cordelli, E., Maulucci, G., De Spirito, M., Rizzi, A., Pitocco, D. and Soda, P. A decision support system for type 1 diabetes mellitus diagnostics based on dual channel analysis of red blood cell membrane fluidity. Computer methods and programs in biomedicine, 162 (2018), 263-271

    2018

  12. [20]

    HbA1c as a diagnostic test for diabetes mellitus–reviewing the evidence

    Florkowski, C. HbA1c as a diagnostic test for diabetes mellitus–reviewing the evidence. The Clinical Biochemist Reviews, 34, 2 (2013), 75

    2013

  13. [21]

    and Wright, G

    Wright, S., Yelland, M., Heathcote, K., Ng, S.-K. and Wright, G. Fear of needles-nature and prevalence in general practice. Australian family physician, 38, 3 (2009), 172

    2009

  14. [22]

    and Koehler, T

    Cemeroglu, A., Can, A., Davis, A., Cemeroglu, O., Kleis, L., Daniel, M., Bustraan, J. and Koehler, T. Fear of needles in children with type 1 diabetes mellitus on multiple daily injections and continuous subcutaneous insulin infusion. Endocrine Practice, 21, 1 (2014), 46-53

    2014

  15. [23]

    L., Grgic, A

    Rosenbloom, A. L., Grgic, A. and Frias, J. L. Diabetes mellitus, short stature and joint stiffness—a new syndrome. Pediatric Research, 8, 4 (1974), 441

    1974

  16. [24]

    and Qidwai, S

    Pandey, A., Usman, K., Reddy, H., Gutch, M., Jain, N. and Qidwai, S. Prevalence of hand disorders in type 2 diabetes mellitus and its correlation with microvascular complications. Annals of medical and health sciences research, 3, 3 (2013), 349-354

    2013

  17. [25]

    and Noacco, C

    Benedetti, A. and Noacco, C. Juvenile diabetic cheiroarthropathy. Acta diabetologia latina, 13, 1-2 (1976), 54-67

    1976

  18. [26]

    G., Landman, G

    Gerrits, E. G., Landman, G. W., Nijenhuis-Rosien, L. and Bilo, H. J. Limited joint mobility syndrome in diabetes mellitus: A minireview. World journal of diabetes, 6, 9 (2015), 1108

    2015

  19. [27]

    and Irsigler, K

    Sauseng, S., Kästenbauer, T. and Irsigler, K. Limited joint mobility in selected hand and foot joints in patients with type 1 diabetes mellitus: a methodology comparison. Diabetes, nutrition & metabolism, 15, 1 (2002), 1-6

    2002

  20. [28]

    Gale, E. A. and Gillespie, K. M. Diabetes and gender. Diabetologia, 44, 1 (2001), 3-15

    2001

  21. [29]

    and Guney, E

    Aydeniz, A., Gursoy, S. and Guney, E. Which musculoskeletal complications are most frequently seen in type 2 diabetes mellitus? Journal of International Medical Research, 36, 3 (2008), 505-511

    2008

  22. [30]

    N., Khader, Y

    Mustafa, K. N., Khader, Y. S., Bsoul, A. K. and Ajlouni, K. Musculoskeletal disorders of the hand in type 2 diabetes mellitus: prevalence and its associated factors. International journal of rheumatic diseases, 19, 7 (2016), 730-735

    2016

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.