Segmentation and Identification of Vertebrae in CT Scans using CNN, k-means Clustering and k-NN

Overview

Segmentation and Identification of Vertebrae in CT Scans using CNN, k-means Clustering and k-NN

If you use this code for your research, please cite our paper:

@Article{informatics8020040,
AUTHOR = {Altini, Nicola and De Giosa, Giuseppe and Fragasso, Nicola and Coscia, Claudia and Sibilano, Elena and Prencipe, Berardino and Hussain, Sardar Mehboob and Brunetti, Antonio and Buongiorno, Domenico and Guerriero, Andrea and Tatò, Ilaria Sabina and Brunetti, Gioacchino and Triggiani, Vito and Bevilacqua, Vitoantonio},
TITLE = {Segmentation and Identification of Vertebrae in CT Scans Using CNN, k-Means Clustering and k-NN},
JOURNAL = {Informatics},
VOLUME = {8},
YEAR = {2021},
NUMBER = {2},
ARTICLE-NUMBER = {40},
URL = {https://www.mdpi.com/2227-9709/8/2/40},
ISSN = {2227-9709},
DOI = {10.3390/informatics8020040}
}

Graphical Abstract: GraphicalAbstract


Materials

Dataset can be downloaded for free at this URL.


Configuration and pre-processing

Configure the file config/paths.py according to paths in your computer. Kindly note that base_dataset_dir should be an absolute path which points to the directory which contains the subfolders with images and labels for training and validating the algorithms present in this repository.

In order to perform pre-processing, execute the following scripts in the given order.

  1. Perform Train / Test split:
python run/task0/split.py --original-training-images=OTI --original-training-labels=OTL \ 
                          --original-validation-images=OVI --original-validation-labels=OVL

Where:

  • OTI is the path with the CT scan from the original dataset (downloaded from VerSe challenge, see link above);
  • OTL is the path with the labels related to the original dataset;
  • OVI is the path where test images will be put;
  • OVL is the path where test labels will be put.
  1. Cropping the splitted datasets:
python run/task0/crop_mask.py --original-training-images=OTI --original-training-labels=OTL \ 
                              --original-validation-images=OVI --original-validation-labels=OVL

Where the arguments are the same of 1).

  1. Pre-processing the cropped datasets (see also Payer et al. pre-processing):
python run/task0/pre_processing.py

Binary Segmentation

In order to perform this stage, 3D V-Net has been exploited. The followed workflow for binary segmentation is depicted in the following figure:

BinarySegmentationWorkflowImage

Training

To perform the training, the syntax is as follows:

python run/task1/train.py --epochs=NUM_EPOCHS --batch=BATCH_SIZE --workers=NUM_WORKERS \
                          --lr=LR --val_epochs=VAL_EPOCHS

Where:

  • NUM_EPOCHS is the number of epochs for which training the CNN (we often used 500 or 1000 in our experiments);
  • BATCH_SIZE is the batch size (we often used 8 in our experiments, in order to benefit from BatchNormalization layers);
  • NUM_WORKERS is the number of workers in the data loading (see PyTorch documentation);
  • LR is the learning rate,
  • VAL_EPOCHS is the number of epochs after which performing validation during training (a checkpoint model is also saved every VAL_EPOCHS epochs).

Inference

To perform the inference, the syntax is as follows:

python run/task1/segm_bin.py --path_image_in=PATH_IMAGE_IN --path_mask_out=PATH_MASK_OUT

Where:

  • PATH_IMAGE_IN is the folder with input images;
  • PATH_MASK_OUT is the folder where to write output masks.

An example inference result is depicted in the following figure:

BinarySegmentationInferenceImage

Metrics Calculation

In order to calculate binary segmentation metrics, the syntax is as follows:

python run/task1/metrics.py

Multiclass Segmentation

The followed workflow for multiclass segmentation is depicted in the following figure:

MultiClassSegmentationWorkflow

To perform the Multiclass Segmentation (can be performed only on binary segmentation output), the syntax is as follows:

python run/task2/multiclass_segmentation.py --input-path=INPUT_PATH \
                                            --gt-path=GT_PATH \
                                            --output-path=OUTPUT_PATH \
                                            --use-inertia-tensor=INERTIA \
                                            --no-metrics=NOM

Where:

  • INPUT_PATH is the path to the folder containing the binary spine masks obtained in previous steps (or binary spine ground truth).
  • GT_PATH is the path to the folder containing ground truth labels.
  • OUTPUT_PATH is the path where to write the output multiclass masks.
  • INERTIA can be either 0 or 1 depending or not if you want to include inertia tensor in the feature set for discrminating between bodies and arches (useful for scoliosis cases); default is 0.
  • NOM can be either 0 or 1 depending or not if you want to skip the calculation of multi-Hausdorff distance and multi-ASSD for the vertebrae labelling (it can be very computationally expensive with this implementation); default is 1.

Figures highlighting the different steps involved in this stage follows:

  • Morphology MultiClassSegmentationMorphology

  • Connected Components MultiClassSegmentationConnectedComponents

  • Clustering and arch/body coupling MultiClassSegmentationClustering

  • Centroids computation MultiClassSegmentationCentroids

  • Mesh reconstruction MultiClassSegmentationMesh


Visualization of the Predictions

The base_dataset_dir folder also contains the outputs folders:

  • predTr contains the binary segmentation predictions performed on training set;
  • predTs contains the binary segmentation predictions performed on testing set;
  • predMulticlass contains the multiclass segmentation predictions and the JSON files containing the centroids' positions.

Direct LiDAR Odometry: Fast Localization with Dense Point Clouds

Direct LiDAR Odometry: Fast Localization with Dense Point Clouds DLO is a lightweight and computationally-efficient frontend LiDAR odometry solution w

VECTR at UCLA 369 Dec 30, 2022
Lightweight Salient Object Detection in Optical Remote Sensing Images via Feature Correlation

CorrNet This project provides the code and results for 'Lightweight Salient Object Detection in Optical Remote Sensing Images via Feature Correlation'

Gongyang Li 13 Nov 03, 2022
Online Multi-Granularity Distillation for GAN Compression (ICCV2021)

Online Multi-Granularity Distillation for GAN Compression (ICCV2021) This repository contains the pytorch codes and trained models described in the IC

Bytedance Inc. 299 Dec 16, 2022
Python implementation of ADD: Frequency Attention and Multi-View based Knowledge Distillation to Detect Low-Quality Compressed Deepfake Images, AAAI2022.

ADD: Frequency Attention and Multi-View based Knowledge Distillation to Detect Low-Quality Compressed Deepfake Images Binh M. Le & Simon S. Woo, "ADD:

2 Oct 24, 2022
PyTorch implementation for "Sharpness-aware Quantization for Deep Neural Networks".

Sharpness-aware Quantization for Deep Neural Networks Recent Update 2021.11.23: We release the source code of SAQ. Setup the environments Clone the re

Zhuang AI Group 30 Dec 19, 2022
Text Extraction Formulation + Feedback Loop for state-of-the-art WSD (EMNLP 2021)

ConSeC is a novel approach to Word Sense Disambiguation (WSD), accepted at EMNLP 2021. It frames WSD as a text extraction task and features a feedback loop strategy that allows the disambiguation of

Sapienza NLP group 36 Dec 13, 2022
Code for the SIGGRAPH 2021 paper "Consistent Depth of Moving Objects in Video".

Consistent Depth of Moving Objects in Video This repository contains training code for the SIGGRAPH 2021 paper "Consistent Depth of Moving Objects in

Google 203 Jan 05, 2023
Interactive Image Segmentation via Backpropagating Refinement Scheme

Won-Dong Jang and Chang-Su Kim, Interactive Image Segmentation via Backpropagating Refinement Scheme, CVPR 2019

Won-Dong Jang 85 Sep 15, 2022
All of the figures and notebooks for my deep learning book, for free!

"Deep Learning - A Visual Approach" by Andrew Glassner This is the official repo for my book from No Starch Press. Ordering the book My book is called

Andrew Glassner 227 Jan 04, 2023
This is an official implementation for the WTW Dataset in "Parsing Table Structures in the Wild " on table detection and table structure recognition.

WTW-Dataset This is an official implementation for the WTW Dataset in "Parsing Table Structures in the Wild " on ICCV 2021. Here, you can download the

109 Dec 29, 2022
SFD implement with pytorch

S³FD: Single Shot Scale-invariant Face Detector A PyTorch Implementation of Single Shot Scale-invariant Face Detector Description Meanwhile train hand

Jun Li 251 Dec 22, 2022
Physics-informed convolutional-recurrent neural networks for solving spatiotemporal PDEs

PhyCRNet Physics-informed convolutional-recurrent neural networks for solving spatiotemporal PDEs Paper link: [ArXiv] By: Pu Ren, Chengping Rao, Yang

Pu Ren 11 Aug 23, 2022
SalGAN: Visual Saliency Prediction with Generative Adversarial Networks

SalGAN: Visual Saliency Prediction with Adversarial Networks Junting Pan Cristian Canton Ferrer Kevin McGuinness Noel O'Connor Jordi Torres Elisa Sayr

Image Processing Group - BarcelonaTECH - UPC 347 Nov 22, 2022
Deep Implicit Moving Least-Squares Functions for 3D Reconstruction

DeepMLS: Deep Implicit Moving Least-Squares Functions for 3D Reconstruction This repository contains the implementation of the paper: Deep Implicit Mo

103 Dec 22, 2022
The PyTorch implementation of paper REST: Debiased Social Recommendation via Reconstructing Exposure Strategies

REST The PyTorch implementation of paper REST: Debiased Social Recommendation via Reconstructing Exposure Strategies. Usage Download dataset Download

DMIRLAB 2 Mar 13, 2022
Implement of "Training deep neural networks via direct loss minimization" in PyTorch for 0-1 loss

This is the implementation of "Training deep neural networks via direct loss minimization" published at ICML 2016 in PyTorch. The implementation targe

Cuong Nguyen 1 Jan 18, 2022
Husein pet projects in here!

project-suka-suka Husein pet projects in here! List of projects mysejahtera-density. Generate resolution points using meshgrid and request each points

HUSEIN ZOLKEPLI 47 Dec 09, 2022
Thermal Control of Laser Powder Bed Fusion using Deep Reinforcement Learning

This repository is the implementation of the paper "Thermal Control of Laser Powder Bed Fusion Using Deep Reinforcement Learning", linked here. The project makes use of the Deep Reinforcement Library

BaratiLab 11 Dec 27, 2022
Based on the given clinical dataset, Predict whether the patient having Heart Disease or Not having Heart Disease

Heart_Disease_Classification Based on the given clinical dataset, Predict whether the patient having Heart Disease or Not having Heart Disease Dataset

Ashish 1 Jan 30, 2022
A PyTorch Reimplementation of TecoGAN: Temporally Coherent GAN for Video Super-Resolution

TecoGAN-PyTorch Introduction This is a PyTorch reimplementation of TecoGAN: Temporally Coherent GAN for Video Super-Resolution (VSR). Please refer to

165 Dec 17, 2022