Evaluation of techniques to improve a deep learning algorithm for the automatic detection of intracranial haemorrhage on CT head imaging

BACKGROUND

Deep learning (DL) algorithms are playing an increasing role in automatic medical image analysis.

PURPOSE

To evaluate the performance of a DL model for the automatic detection of intracranial haemorrhage and its subtypes on non-contrast CT (NCCT) head studies and to compare the effects of various preprocessing and model design implementations.

METHODS

The DL algorithm was trained and externally validated on open-source, multi-centre retrospective data containing radiologist-annotated NCCT head studies. The training dataset was sourced from four research institutions across Canada, the USA and Brazil. The test dataset was sourced from a research centre in India. A convolutional neural network (CNN) was used, with its performance compared against similar models with additional implementations: (1) a recurrent neural network (RNN) attached to the CNN, (2) preprocessed CT image-windowed inputs and (3) preprocessed CT image-concatenated inputs. The area under the receiver operating characteristic curve (AUC-ROC) and microaveraged precision (mAP) score were used to evaluate and compare model performances.

RESULTS

The training and test datasets contained 21,744 and 491 NCCT head studies, respectively, with 8,882 (40.8%) and 205 (41.8%) positive for intracranial haemorrhage. Implementation of preprocessing techniques and the CNN-RNN framework increased mAP from 0.77 to 0.93 and increased AUC-ROC [95% confidence intervals] from 0.854 [0.816-0.889] to 0.966 [0.951-0.980] (p-value = 3.91 × 10^-12).

CONCLUSIONS

The deep learning model accurately detected intracranial haemorrhage and improved in performance following specific implementation techniques, demonstrating clinical potential as a decision support tool and an automated system to improve radiologist workflow efficiency.

KEY POINTS

• The deep learning model detected intracranial haemorrhages on computed tomography with high accuracy. • Image preprocessing, such as windowing, plays a large role in improving deep learning model performance. • Implementations which enable an analysis of interslice dependencies can improve deep learning model performance. • Visual saliency maps can facilitate explainable artificial intelligence systems. • Deep learning within a triage system may expedite earlier intracranial haemorrhage detection.

Digital subtraction of temporally sequential mammograms for improved detection and classification of microcalcifications

Background

Our aim was to demonstrate that automated detection and classification of breast microcalcifications, according to Breast Imaging Reporting and Data System (BI-RADS) categorisation, can be improved with the subtraction of sequential mammograms as opposed to using the most recent image only.

Methods

One hundred pairs of mammograms were retrospectively collected from two temporally sequential rounds. Fifty percent of the images included no (BI-RADS 1) or benign (BI-RADS 2) microcalcifications. The remaining exhibited suspicious findings (BI-RADS 4-5) in the recent image. Mammograms cannot be directly subtracted, due to tissue changes over time and breast deformation during mammography. To overcome this challenge, optimised preprocessing, image registration, and postprocessing procedures were developed. Machine learning techniques were employed to eliminate false positives (normal tissue misclassified as microcalcifications) and to classify the true microcalcifications as BI-RADS benign or suspicious. Ninety-six features were extracted and nine classifiers were evaluated with and without temporal subtraction. The performance was assessed by measuring sensitivity, specificity, accuracy, and area under the curve (AUC) at receiver operator characteristics analysis.

Results

Using temporal subtraction, the contrast ratio improved ~ 57 times compared to the most recent mammograms, enhancing the detection of the radiologic changes. Classifying as BI-RADS benign versus suspicious microcalcifications, resulted in 90.3% accuracy and 0.87 AUC, compared to 82.7% and 0.81 using just the most recent mammogram (p = 0.003).

Conclusion

Compared to using the most recent mammogram alone, temporal subtraction is more effective in the microcalcifications detection and classification and may play a role in automated diagnosis systems.

Supplementary Information

The online version contains supplementary material available at 10.1186/s41747-021-00238-w.