External Validation Study On The Value of Deep Learning Algorithm For The Prediction of Hematoma Expansion From Noncontrast CT Scans

doi:10.21203/rs.3.rs-1197777/v1

Download PDF

Research Article

External Validation Study On The Value of Deep Learning Algorithm For The Prediction of Hematoma Expansion From Noncontrast CT Scans

https://doi.org/10.21203/rs.3.rs-1197777/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 14 Mar, 2022

Read the published version in BMC Medical Imaging →

You are reading this latest preprint version

Background: Hematoma expansion is an independent predictor of patient outcome and mortality. The early diagnosis of hematoma expansion is crucial for selecting clinical treatment options This study aims to explore the value of a deep learning algorithm for the prediction of hematoma expansion from noncontrast Computed tomography（NCCT） scan through external validation.

Methods: 102 NCCT images of Hypertensive intracerebral hemorrhage (HICH) patients diagnosed in our hospital were retrospectively reviewed. The initial Computed tomography (CT) scan images were evaluated by a commercial Artificial intelligence (AI) software using deep learning algorithm and radiologists respectively to predict hematoma expansion and the corresponding sensitivity and specificity of the two groups were calculated and compared, Pair-wise comparisons were conducted among gold standard hematoma expansion diagnosis time, AI software diagnosis time and doctors’ reading time.

Results: Among 102 HICH patients, The sensitivity, specificity and accuracy of predicting hematoma expansion in the AI group were higher than those in the doctor group（80.0% vs 66.7%,73.6% vs 58.3%,75.5% vs 60.8%）,with statistically significant difference (p<0.05).The AI diagnosis time (2.8 ± 0.3s) and the doctors’ diagnosis time (11.7 ± 0.3s) were both significantly shorter than the gold standard diagnosis time (14.5 ± 8.8h) (p <0.05), AI diagnosis time was significantly shorter than that of doctors (p<0.05).

Conclusions: Deep learning algorithm could effectively predict hematoma expansion at an early stage from the initial CT scan images of HICH patients after onset with high sensitivity and specificity and greatly shortened diagnosis time, which provides a new, accurate, easy-to-use and fast method for the early prediction of hematoma expansion.

Nuclear Medicine & Medical Imaging

artificial intelligence

hypertensive intracerebral hemorrhage

hematoma expansion

early diagnosis

Hypertensive intracerebral hemorrhage (HICH) refers to the sudden occurrence of cerebral parenchymal hemorrhage in patients with a clear history of hypertension, excluding secondary cerebral hemorrhage caused by trauma, vascular abnormalities, coagulation dysfunction, blood diseases and other diseases，which is the most common form of stroke accounting for about 50% to 70% of spontaneous intracerebral hemorrhage[1].Spontaneous intracerebral hemorrhage (SICH) causes severe threats to patients physically and mentally because it starts suddenly and worsens rapidly resulting in a high fatality and disability rate[2]. About 30 %~50 %[3-5] SICH patients died within one month since disease onset, and about 1/3 of these patients had hematoma expansion, of which 73% had hematoma expansion within 24 hours[6-8]. Hematoma expansion is an independent predictor of patient outcome and mortality. Therefore, the early diagnosis of hematoma expansion is crucial for selecting clinical treatment options [9, 10].

Many imaging methods have been explored to predict hematoma expansion in patients with HICH However, the sensitivity of predicting hematoma expansion remains low simply by analyzing these imaging features. For example, the spot sign of Computed Tomography Angiography (CTA) has demonstrated high predictive value for hematoma expansion in patients with HICH[11, 12], several studies have shown that the sensitivity of CTA spot sign for the prediction of hematoma expansion is 26.2%~73%[10, 13] In one recent study, the sensitivity of CTA spot sign to predict hematoma expansion was merely 34.6%[14]. CTA scan increased radiation dose, and contrast agent injection may lead to allergic reactions and renal damage. In addition, many primary hospitals lack medical resources and might not be able to carry out CTA examinations limiting the broad application of CTA [15-17]. Some scholars seek to predict hematoma expansion by analyzing imaging features from NCCT images in HICH patients, and several studies have shown that the prediction sensitivity ranges from 31% to 55%[17-20]. In addition, the prediction of hematoma expansion through imaging features is subjective and the inter-reader agreement was merely moderate[16, 21]. Moreover, time is the brain, early and quick prediction of hematoma expansion significantly impacts on patient prognosis [22]. At present, hematoma expansion is generally determined by changes in hematoma volume of two CT scans within 24 hours since hospital admission and the interval time is relatively long. Therefore, it is imperative to find an early, reliable and easy-to-use method for predicting hematoma expansion.

AI is gaining increased use in the diagnosis of brain diseases and one latest study shows that the sensitivity to detect intracranial aneurysms of radiologists could be improved from 62.86% to 86.67% with the assistance of a deep learning algorithm [23]. Wang Jiwen et al. applied deep learning to segment CT images to measure hematoma volume of 1223 patients with SICH and the resulting measurement of hematoma volume by deep learning was more accurate than the ABC/2 formula[24]. Sun Haiyue [25] established a decision tree model to predict 196 patients with HICH. It was found that the outcome prediction was in good agreement with the actual outcome, which could assist the clinical treatment of HICH. However, few studies used AI to predict hematoma expansion in HICH patients. The previous study reported hematoma prediction AI model provides a time-saving, easy to implement, and subjective independent method to predict the risk of hematoma enlargement in patients with intracerebral hemorrhage based on the NCCT images [26]. This study aimed to explore the value of the deep learning algorithm for the prediction of hematoma expansion from NCCT Scans through external validation.

(1) Patient Characteristics

Brain CT images of HICH patients diagnosed in our hospital from January 2019 to October 2020 were retrospectively collected and analyzed, an initial CT scan was performed within 6 hours of onset and a follow-up CT scan was performed within 24 hours of admission. Exclusion criteria were: (1) Patients with a history of cerebral hemorrhage or had undergone brain surgery or had undergone brain surgery within 24 hours after admission. (2) Poor image quality that affects image evaluation. This study was reviewed and approved by the Institutional Review Board of our hospital

(2) Scanning Protocol

A conventional head scan was performed using 16-row spiral CT (Brilliance CT, Philips). Scanning parameters were: tube voltage120 kV, tube current 300mAs, slice thickness and increment were 4.5mm, matrix 512 × 512.

(3) Gold standard for hematoma expansion

Hematoma volumes were manually measured by two radiologists (Doctor A, Doctor B) with 5 and 6 years of experience in diagnosing cerebral hemorrhage and calculated using ABC / 2. Formula. Hematoma expansion was determined if the hematoma volume change of the two CT scan within 24 h increased by more than 12.5 ml or 33%[18]. Hematoma expansion was decided by two doctors, respectively and the diagnostic results were reviewed by a senior neuroradiologist (Doctor C, 20 years of experience in diagnosing cerebral hemorrhage). The final diagnosis of the neuroradiologist was used as the gold standard for hematoma expansion.

(4) AI diagnosis

Initial CT scan images of the brain were imported to the AI aided diagnostic system (BioMind 1.4.0, Hanalytics, Beijing, China) for automatic hematoma expansion prediction (Figure 1). The deep learning system was developed by involving 1771 hypertensive intracerebral hemorrhage patients from 84 hospitals spread among 9 provinces and 25 cities, all of which were members of the Chinese Stroke Center Alliance, The proposed model achieved a The sensitivity and specificity of the original model for hematoma expansion were of 89. 3% and 81.1%.[27].

(5) Doctor's diagnosis

In order to reduce interpretation variability by the two radiologists for predicting hematoma expansion, image interpretation criteria was applied to train the two radiologists with more than 15 years of experience in diagnosing hypertensive intracerebral hemorrhage. Imaging signs for hematoma expansion diagnosis were: ①Black hole sign: hypoattenuating area encapsulated within the hyperattenuating hematoma with a clearly defined border and the hematoma should have at least 28 Hounsfield unit (HU) difference between the 2 density regions[17]. ②Blend Sign: blending of the hypoattenuating area and hyperattenuating region with a well-defined margin, the hematoma should have at least 18 HU difference between the 2 density regions[20]. ③Island Sign༚≥3 scattered small hematomas all separate from the main hematoma[5]. Each of the three signs identified in patient CT images confirms hematoma expansion. Initial CT scan images were independently read by two radiologists and they were not aware of the clinical manifestations of the patients and the review results of the CT follow-up scan within 24 hours. The hematoma expansion was predicted by analyzing the imaging signs of the initial CT scan, when opinions are inconsistent, the consensus is reached through discussion.

(6) Statistical analysis

SPSS 27.0 (IBM Armonk, NY, USA) software was used for statistical analysis with P<0.05 as statistical significance. The final diagnosis of the neuroradiologist was used as the gold standard, the sensitivity, specificity and accuracy of the AI group and doctor group in predicting hematoma expansion were compared. Categorical variables between the two groups were compared using the χ2 test presented in percentages. Additionally, AI diagnostic time, doctor’s diagnostic time and gold standard diagnosis time of hematoma expansion were compared. Continuous variables were compared between groups with Student’s t-test expressed as the mean± standard deviation.

2.1 Patient information

Inclusion and exclusion criteria for patient recruitment in this retrospective study was summarized in figure 2. A total of 160 patients with HICH were initially collected. Among them, 38 patients underwent hematoma drainage within 24 hours after admission, Since it was impossible to determine whether hematoma expansion occurs, it was excluded, 18 patients who died after admission or transferred from the hospital did not undergo follow-up CT scan within 24 hours, and 2 patients had images with artifacts that affect image interpretation. A total of 102 cases were finally included in 71 males and 31 females, with an average age of 64.5 ± 13.2 years (age range 32-89 years).

2.2. The predictive result of AI group and doctor group for hematoma expansion in HICH patients

Among 102 HICH patients, about 29.41% (30/102) patients had hematoma expansion. The prediction sensitivity of the AI group was 80.0% (24/30), the specificity was 73.6% (53/72), and the accuracy was 75.5% (77/102) (Table 1).

The prediction sensitivity of the doctor group was 66.7% (20/30), the specificity was 58.3% 9(42/72) and the accuracy was 60.8% (62/102) (Table 2).

The sensitivity (Figure 3), specificity (Figure 4) and accuracy of hematoma expansion prediction in the AI group were higher than those in the doctor group (p<0.05).

Table 1

Confusion matrix of AI group for hematoma expansion prediction analysis
		Hematoma Expansion Gold Standard		Total
		Positive	Negative	Total
AI’s Diagnosis	Positive	24	19	43
AI’s Diagnosis	Negative	6	53	59
Total		30	72	102

Table 2

Confusion matrix of doctor group for hematoma expansion prediction analysis
		Hematoma Expansion Gold Standard		Total
		Positive	Negative	Total
Doctors’Diagnosis	Positive	20	30	50
Doctors’Diagnosis	Negative	10	42	52
Total		30	72	102

2.3 The diagnosis time of hematoma expansion in AI group and doctor group

The average review time of brain CT images was 14.5 ± 8.8 (hours), while the diagnosis time of AI was 2.8 ± 0.3s, and the diagnosis time of doctors was about 11.7±0.3s. The diagnosis time of doctors was significantly shorter than that of the gold standard (p < 0.05). The diagnosis time of AI was significantly shorter than that of doctors and gold standard (p < 0.05). The diagnosis time of AI was more than 99% shortened than that of gold standard, and about 76.3% shortened than that of doctors.

At present, there is no consensus on the definition of hematoma expansion. Some studies defined hematoma expansion as a 33% increase in the relative volume of the hematoma or an absolute volume increase of 6 ml[5, 28]. Some studies defined hematoma expansion as a 33% increase in the relative volume of the hematoma or an absolute volume increase of 12.5ml[6, 17, 18]. The latter was widely used in clinical settings, especially in large-scale clinical trials. In this study, hematoma expansion was defined as the increase of absolute volume > 12.5 ml or relative volume > 33% in two brain CT scan within 24 h after admission.

In 102 HICH patients in this study, about 29.4% (30/102) had hematoma expansion which is consistent with literature reports[19, 29–31], but lower than the reported 35.4% by Chan et al.[6]. The reason might be that 38 HICH patients who conducted hematoma drainage within 24 hours due to massive bleeding or broken into the ventricle were excluded which may potentially lower the incidence rate of hematoma expansion.

At present, several clinical indicators related to hematoma expansion have been confirmed, such as consciousness level, blood pressure, and blood glucose[9]. However, the prediction of hematoma expansion by consciousness level is only applicable to patients with mild initial consciousness[9]. Blood pressure and blood glucose are independent predictors of hematoma expansion. Whereas, most patients will undergo antihypertensive and hypoglycemic treatment after admission, the dynamic changes of blood pressure and blood glucose make hematoma expansion prediction impossible.

Many imaging-based methods for deciding hematoma expansion have been proposed, and some studies had found that the spot sign of CTA had a high predictive value for hematoma expansion[11, 12]. Certain studies have shown that the sensitivity of CTA spot sign to hematoma expansion was about 26.2%~73%[10, 13]. In a recent study, the sensitivity of CTA spot sign to predict hematoma expansion was 34.6%[14]. In addition, spot sign often requires patients to undergo brain CTA immediately within the first several hours of onset. It has been shown that the CTA spot sign had the highest sensitivity to predict hematoma expansion if CTA was conducted within 2 hours of onset, while the sensitivity was only 60.0%[17, 32]. CTA is not the priority for diagnosing HICH, and it could be carried out in some areas with insufficient medical resources. Therefore, most HICH patients could not perform CTA within 2 hours after onset and CTA scan increased radiation dose and the injection of contrast agent may lead to allergic reactions and renal damage. These reasons limit the widespread applications of CTA[15, 16]. In another study, Fu et al. performed spectral CTA and explored iodine sign to predict hematoma expansion. The sensitivity of iodine sign to predict hematoma expansion was 91.5%, which was significantly higher than that of spot sign (61.8%) and iodine sign had a higher accuracy (85.7vs75.8%) for hematoma expansion prediction than spot sign[33]. Although the iodine sign had a higher sensitivity and accuracy for hematoma expansion prediction, dual-energy spiral CT is needed and lacks in areas with insufficient medical resources. Therefore, the use of iodine sign to predict hematoma expansion was not applicable in all medical institutions.

More and more scholars started to predict the expansion of hematoma from the perspectives of the density, shape and edge of the hematoma on NCCT scan image[5]. Barras et al.[18] studied the brain CT of 90 HICH patients and found that hematoma heterogeneity could be an independent predictor of hematoma expansion. The more heterogeneous the density was, the more likely the hematoma expansion. However, the definition of hematoma heterogeneity was not clear, the judgment of hematoma heterogeneity was subjective, and the inter-reader agreement was merely moderate[21]. In other several studies, Li proposed to use "black hole sign", "blend sign", and "island sign" to predict hematoma expansion. The study found that the sensitivity of the above three signs to predict hematoma expansion were 31.9%, 39.3% and 44.7%[5, 17, 20]. In addition, in a study of 200 HICH patients, it was found that the sensitivity of black hole sign to predict hematoma expansion was 33.8% which is lower than that (46.5%) of swirl sign. After multivariate analysis, only the black hole sign was an independent predictor of hematoma expansion[34]. In another study, the sensitivity of the black hole sign and blend sign to predict hematoma expansion were 23.1% and 31.7%, respectively. In a study of 307 HICH patients, Deng[31] found that the island sign had higher specificity than the satellite sign in predicting hematoma expansion (98.1% vs 57.9%), and the satellite sign had a higher sensitivity than the island sign (65.6% vs 45.2%). In general, the sensitivity of NCCT scans alone to predict hematoma expansion is relatively low.

In this study, the sensitivity and specificity of AI software in predicting hematoma expansion were 80.0% and 73.6%, respectively, which were significantly higher than the sensitivity (66.7%) and specificity (58.3%) of the doctors using NCCT imaging signs. Prediction accuracy of hematoma expansion is crucial for clinical treatment and in some large randomized controlled trials, the failure to improve patient prognosis by inhibiting hematoma expansion is mainly due to the low accuracy of hematoma expansion prediction[35]. In this study, the accuracy of AI to predict hematoma expansion was 75.5% which was significantly higher than that (60.8%) of the doctors. In addition, the diagnosis time of hematoma expansion is crucial for the treatment[22]. A follow-up brain CT scan is often performed in the clinical workflow within 24 hours after admission and hematoma expansion is diagnosed by comparing the volume change of hematoma of two CT scans. The interval time is long, during which the HICH patients may deteriorate rapidly. In this study, the average interval time of brain CT was 14.5±8.8 (hours), the AI diagnosis time was 2.8±0.3s and the doctor diagnosis time was about 11.7±0.3s. The AI diagnosis time was significantly shorter than those of doctors and the gold standard diagnosis time (p<0.05), AI diagnosis time was shortened by more than 99% and 76.3% compared with the gold standard time and the doctor diagnosis time, respectively. It could be concluded that AI shortens the diagnosis time of hematoma expansion at an early stage while a high sensitivity specificity and accuracy were maintained. In addition, the AI software based on a deep learning is objective and eliminates inter-reader variability, which is helpful for clinical promotion.

In summary, we have externally validated that AI software based on deep learning algorithm could effectively predict hematoma expansion at an early stage from the initial CT scan images of HICH patients after hemorrhage onset. It has a relatively high sensitivity specificity and accuracy could significantly shorten the determination time of hematoma expansion and helps to guide the selection of clinical treatment options. In addition, this method does not rely on CTA scanning and has greater applicability, especially for primary hospitals that lack medical resources and cannot carry out CTA. In addition, this method does not require the injection of contrast agents to avoid adverse reactions caused by contrast agents. The prediction of hematoma expansion by AI software provides a new, accurate, easy to use and fast method for the early prediction of hematoma expansion, which may potentially change the diagnostic process of hematoma expansion.

However, there are several limitations of this study. First, the sample size is relatively small, In the future, the sample size could be expanded, to explore further the prediction efficiency of AI software for hematoma expansion. Secondly, in this study, the gap of experience levels of the two radiologists was minor and the inter-reader prediction difference of the two radiologists was not investigated. Thirdly, the improvement of AI software for doctors’ diagnostic efficacy could be put into future work.

Deep learning algorithm could effectively predict hematoma expansion at an early stage from the initial CT scan images of HICH patients after onset with high sensitivity and specificity and significantly shortened diagnosis time, which provides a new, accurate, easy-to-use and fast method for the early prediction of hematoma expansion.

NCCT: noncontrast Computed tomography; HICH: Hypertensive intracerebral hemorrhage;

CT: Computed tomography; AI: Artificial intelligence

SICH: Spontaneous intracerebral hemorrhage;

CTA: Computed Tomography Angiography

Acknowledgements

Thanks are due to Yifeng Zheng, Jian He for assistance with the study and to Jun Gu, Hai Rui Chu for valuable discussion and to Na Dong for data analysis. We thank all of the participants of this study.

Authors’ contributions

DCG designed the study, conducted the experiments, analyzed data,and prepared the manuscript. JG and HRC took part in valuable discussions and provided linguistic assistance. ND involved in acquisition of data. JH and YFZ took part in critical revision of the manuscript. All authors read and approved the final manuscript.

Source of Funding

This research was funded by the Public Welfare Applied Research Project of Huzhou Science and Technology Bureau (No. 2020GY20).

Availability of data and materials

All data generated or analyzed during this study are included in this published article.

Ethics approval and consent to participate

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. The informed consent was obtained from all subjects or, if subjects are under 16, from a parent and/or legal guardian. The study was approved by the Ethical Committee of Huzhou Central Hospital prior to study conduct.

Consent for publication

Not applicable.

Competing interests

The authors declared no potential conflicts of interest. We declare that there is

no any conflict of interest with the company that makes the software (Biomind Technology Co., Ltd. Beijing, China).

Author details

1 From the Department of Radiology, Huzhou Central Hospital, Affiliated Central Hospital of Huzhou University, Huzhou 313000, Zhejiang Province, China

2 Institute of Clinical Research, Biomind Technology, Beijing 100050, China

He ML: Attention should be paid to the regulation of hypertension in the early stage of intracerebral hemorrhage Chinese Journal of Neurology 2018, 51(04):244–246.
Scherer M, Cordes J, Younsi A, Sahin YA, Götz M, Möhlenbruch M, Stock C, Bösel J, Unterberg A, Maier-Hein K et al: Development and Validation of an Automatic Segmentation Algorithm for Quantification of Intracerebral Hemorrhage. Stroke 2016, 47(11):2776–2782.
Liu WD: Progress in diagnosis and treatment of hypertensive intracerebral hemorrhage. Shanghai Medical Journal 2017, 40(11):646–649.
Davis SM, Broderick J, Hennerici M, Brun NC, Diringer MN, Mayer SA, Begtrup K, Steiner T: Hematoma growth is a determinant of mortality and poor outcome after intracerebral hemorrhage. Neurology 2006, 66(8):1175–1181.
Li Q, Liu QJ, Yang WS, Wang XC, Zhao LB, Xiong X, Li R, Cao D, Zhu D, Wei X et al: Island Sign: An Imaging Predictor for Early Hematoma Expansion and Poor Outcome in Patients With Intracerebral Hemorrhage. Stroke 2017, 48(11):3019–3025.
Chan S, Conell C, Veerina KT, Rao VA, Flint AC: Prediction of intracerebral haemorrhage expansion with clinical, laboratory, pharmacologic, and noncontrast radiographic variables. International journal of stroke: official journal of the International Stroke Society 2015, 10(7):1057–1061.
Mayer SA, Brun NC, Begtrup K, Broderick J, Davis S, Diringer MN, Skolnick BE, Steiner T: Recombinant activated factor VII for acute intracerebral hemorrhage. The New England journal of medicine 2005, 352(8):777–785.
Yang YB, Xiao L, Wang SH: Analysis of Factors Associated Hematoma Enlargement in Hypertensive Cerebral Hemorrhage. Journal of Clinical Radiology 2020, 39(08):1467–1472.
Chen S, Zhao B, Wang W, Shi L, Reis C, Zhang J: Predictors of hematoma expansion predictors after intracerebral hemorrhage. Oncotarget 2017, 8(51):89348–89363.
Del Giudice A, D'Amico D, Sobesky J, Wellwood I: Accuracy of the spot sign on computed tomography angiography as a predictor of haematoma enlargement after acute spontaneous intracerebral haemorrhage: a systematic review. Cerebrovascular diseases (Basel, Switzerland) 2014, 37(4):268–276.
Orito K, Hirohata M, Nakamura Y, Takeshige N, Aoki T, Hattori G, Sakata K, Abe T, Uchiyama Y, Sakamoto T et al: Leakage Sign for Primary Intracerebral Hemorrhage: A Novel Predictor of Hematoma Growth. Stroke 2016, 47(4):958–963.
Powers WJ, Rabinstein AA, Ackerson T, Adeoye OM, Bambakidis NC, Becker K, Biller J, Brown M, Demaerschalk BM, Hoh B et al: 2018 Guidelines for the Early Management of Patients With Acute Ischemic Stroke: A Guideline for Healthcare Professionals From the American Heart Association/American Stroke Association. Stroke 2018, 49(3):e46-e110.
Hostettler IC, Seiffge DJ, Werring DJ: Intracerebral hemorrhage: an update on diagnosis and treatment. Expert review of neurotherapeutics 2019, 19(7):679–694.
Sporns PB, Schwake M, Kemmling A, Minnerup J, Schwindt W, Niederstadt T, Schmidt R, Hanning U: Comparison of Spot Sign, Blend Sign and Black Hole Sign for Outcome Prediction in Patients with Intracerebral Hemorrhage. Journal of stroke 2017, 19(3):333–339.
Ng D, Churilov L, Mitchell P, Dowling R, Yan B: The CT Swirl Sign Is Associated with Hematoma Expansion in Intracerebral Hemorrhage. AJNR American journal of neuroradiology 2018, 39(2):232–237.
Li H, Wang X, Xie YL, Zhang ST, Liu YZ, Ma F: The Preliminary Study of Predicting Early Enlargement of Hypertensive Intracerebral Hemorrhage by Radiomic Methodbased on Baseline Non-Contrast CT. Journal of Clinical Radiology 2019, 38(09):1609–1613.
Li Q, Zhang G, Xiong X, Wang XC, Yang WS, Li KW, Wei X, Xie P: Black Hole Sign: Novel Imaging Marker That Predicts Hematoma Growth in Patients With Intracerebral Hemorrhage. Stroke 2016, 47(7):1777–1781.
Barras CD, Tress BM, Christensen S, MacGregor L, Collins M, Desmond PM, Skolnick BE, Mayer SA, Broderick JP, Diringer MN et al: Density and shape as CT predictors of intracerebral hemorrhage growth. Stroke 2009, 40(4):1325–1331.
Wang SH, Wang ZW, Deng LB, Liu YB, Zhang G: Hematoma Expansion within 24 hours of Hypertensive Intracerebral Hemorrhage and Its Association with Signs on Non-enhanced Computed Tomography. Acta Academiae Medicinae Sinicae 2019, 41(04):472–478.
Li Q, Zhang G, Huang YJ, Dong MX, Lv FJ, Wei X, Chen JJ, Zhang LJ, Qin XY, Xie P: Blend Sign on Computed Tomography: Novel and Reliable Predictor for Early Hematoma Growth in Patients With Intracerebral Hemorrhage. Stroke 2015, 46(8):2119–2123.
Law ZK, Ali A, Krishnan K, Bischoff A, Appleton JP, Scutt P, Woodhouse L, Pszczolkowski S, Cala LA, Dineen RA et al: Noncontrast Computed Tomography Signs as Predictors of Hematoma Expansion, Clinical Outcome, and Response to Tranexamic Acid in Acute Intracerebral Hemorrhage. Stroke 2020, 51(1):121–128.
Gomez CR: Time Is Brain: The Stroke Theory of Relativity. Journal of stroke and cerebrovascular diseases: the official journal of National Stroke Association 2018, 27(8):2214–2227.
Yang J, Xie M, Hu C, Alwalid O, Xu Y, Liu J, Jin T, Li C, Tu D, Liu X et al: Deep Learning for Detecting Cerebral Aneurysms with CT Angiography. Radiology 2021, 298(1):155–163.
Wang JW, Lin Y, Xiong JH, Yu SP, Wei W, Yang XY, Xiao FS, Wang YL, Liang KM, Wang H et al: Evaluation of spontaneous intracerebral hemorrhage by using CT image segmentation and volume assessment based on deep learning. Chinese Journal of Radiology 2019, 53(11):941–945.
Sun HY, Zhang C, Yang C: Prediction of hypertensive cerebral hemorrhage based on big data method Chinese Journal of Gerontology 2019, 39(16):3884–3888.
Teng L, Ren Q, Zhang P, Wu Z, Guo W, Ren T: Artificial Intelligence Can Effectively Predict Early Hematoma Expansion of Intracerebral Hemorrhage Analyzing Noncontrast Computed Tomography Image. Frontiers in aging neuroscience 2021, 13:632138.
Wang Y, Li Z, Zhao X, Wang C, Wang X, Wang D, Liang L, Liu L, Wang C, Li H et al: Effect of a Multifaceted Quality Improvement Intervention on Hospital Personnel Adherence to Performance Measures in Patients With Acute Ischemic Stroke in China: A Randomized Clinical Trial. Jama 2018, 320(3):245–254.
Rodriguez-Luna D, Coscojuela P, Rubiera M, Hill MD, Dowlatshahi D, Aviv RI, Silva Y, Dzialowski I, Lum C, Czlonkowska A et al: Ultraearly hematoma growth in active intracerebral hemorrhage. Neurology 2016, 87(4):357–364.
Wu Y, Ren MS, Cheng ZZ, Li HY: Risk factors of hematoma enlargement in spontaneous intracerebral hemorrhage at acute stage. Journal of Clinical Neurology 2015, 28(01):24–27.
Li H, Wang X, Zhang ST, Xie YL, Liu YZ, M F, Li Y, Li ZQ: Predictivevalue of CTtexture analysisforearlyenlargement ofhypertensiveintracerebral hemorrhage. journal of practical radiology 2019, 035(010):1564-1567,1578.
Deng L, Zhang G, Wei X, Yang WS, Li R, Shen YQ, Xie XF, Lv XN, Li YL, Zhao LB et al: Comparison of Satellite Sign and Island Sign in Predicting Hematoma Growth and Poor Outcome in Patients with Primary Intracerebral Hemorrhage. World neurosurgery 2019, 127:e818-e825.
Dowlatshahi D, Brouwers HB, Demchuk AM, Hill MD, Aviv RI, Ufholz LA, Reaume M, Wintermark M, Hemphill JC, 3rd, Murai Y et al: Predicting Intracerebral Hemorrhage Growth With the Spot Sign: The Effect of Onset-to-Scan Time. Stroke 2016, 47(3):695–700.
Fu F, Sun S, Liu L, Gu H, Su Y, Li Y: Iodine Sign as a Novel Predictor of Hematoma Expansion and Poor Outcomes in Primary Intracerebral Hemorrhage Patients. Stroke 2018, 49(9):2074–2080.
Xiong X, Li Q, Yang WS, Wei X, Hu X, Wang XC, Zhu D, Li R, Cao D, Xie P: Comparison of Swirl Sign and Black Hole Sign in Predicting Early Hematoma Growth in Patients with Spontaneous Intracerebral Hemorrhage. Medical science monitor: international medical journal of experimental and clinical research 2018, 24:567–573.
Mayer SA, Davis SM, Skolnick BE, Brun NC, Begtrup K, Broderick JP, Diringer MN, Steiner T: Can a subset of intracerebral hemorrhage patients benefit from hemostatic therapy with recombinant activated factor VII? Stroke 2009, 40(3):833–840.

No competing interests reported.

Download PDF

Journal Publication

published 14 Mar, 2022

Read the published version in BMC Medical Imaging →

Editorial decision: Major revision
17 Jan, 2022
Reviews received at journal
10 Jan, 2022
Reviewers agreed at journal
06 Jan, 2022
Reviewers agreed at journal
30 Dec, 2021
Reviewers invited by journal
30 Dec, 2021
Editor assigned by journal
30 Dec, 2021
Editor invited by journal
30 Dec, 2021
Submission checks completed at journal
29 Dec, 2021
First submitted to journal
22 Dec, 2021

You are reading this latest preprint version

External Validation Study On The Value of Deep Learning Algorithm For The Prediction of Hematoma Expansion From Noncontrast CT Scans

Status:

Journal Publication

Version 1

Abstract

Figures

Background

1 Materials And Methods

2 Results

2.1 Patient information

2.2. The predictive result of AI group and doctor group for hematoma expansion in HICH patients

2.3 The diagnosis time of hematoma expansion in AI group and doctor group

3 Discussion

Conclusion

Abbreviations

Declarations

References

Additional Declarations

Status:

Journal Publication

Version 1